What do you think?
Project Maven: A Marine Colonel, His Team, and the Dawn of AI Warfare
The dramatic story of the secretive decade-long Pentagon campaign to bring AI-powered targeting systems onto the battlefield.
In 2017, a small crew gathered in a windowless Pentagon room to put AI at the heart of how America makes war. Led by a Marine Corps colonel haunted by the deaths of US troops and prospect of AI-enhanced rivals, the Project Maven team raced to send AI into combat, igniting controversy and forever changing the US military. Summoning the mayhem of a tech startup, the group wrestled Pentagon bureaucrats and each other, enlisted an initially reluctant Silicon Valley, and convinced US forces to deploy little-tested AI systems in hot wars. Maven fielded algorithms to identify targets at speed and scale, developed AI-infused command systems, and learned where AI fails. Today, its lessons are folded into developing autonomous technology set to be on the frontlines of future war. Based on more than 200 interviews with Project Maven insiders and opponents, this compelling narrative explains how AI warfare, once the stuff of apocalyptic science fiction, has become a reality
In 2017, a small crew gathered in a windowless Pentagon room to put AI at the heart of how America makes war. Led by a Marine Corps colonel haunted by the deaths of US troops and prospect of AI-enhanced rivals, the Project Maven team raced to send AI into combat, igniting controversy and forever changing the US military. Summoning the mayhem of a tech startup, the group wrestled Pentagon bureaucrats and each other, enlisted an initially reluctant Silicon Valley, and convinced US forces to deploy little-tested AI systems in hot wars. Maven fielded algorithms to identify targets at speed and scale, developed AI-infused command systems, and learned where AI fails. Today, its lessons are folded into developing autonomous technology set to be on the frontlines of future war. Based on more than 200 interviews with Project Maven insiders and opponents, this compelling narrative explains how AI warfare, once the stuff of apocalyptic science fiction, has become a reality
Audible Audio
Published March 24, 2026
About the author
Katrina Manson
1 book14 followersKatrina Manson is an award–winning Bloomberg reporter who covers cyber, emerging tech, and national security. Her investigations exposed details of the US military’s AI use and US–China rivalry. She was previously the Financial Times US foreign policy and defense correspondent.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 112 reviews
June 10, 2026
I had heard good things about this book but I liked it even more than I thought I would.
Why this book? I started by being interested in AI with all the hype about it. However, reading general books on various forms of AI led me either advanced coding/modelling tomes or else to the overall assessments of commentators that the area and its hype was overblown. Neither of these was interesting to me - the first due to lack of knowledge/skill and the second due to doubt regarding the skepticism - since there seemed to be something to the hype.
So…I read some materials on the firms involved, where people were investing a lot and had grand expectations regarding firms and their products. For all the involved firms, there was much to read but I narrowed in on Anthropic and Palantir, which seemed different from the normal hype around the more prominent tech firms. With Palantir (and Karp) the link to government work and the applications of AI to warfare stand out - especially given recent geopolitical developments. It does not take much effort to focus in on Drew Cukor and Project Maven.
This got me to what for me is the most interesting aspect of AI - the linking of what is incredibly general and abstract with what is as particular as it gets - the application of AI to modern combat situations and whether this application can boost the performance, efficiency, etc. of modern warfighters (and change the nature of modern warfare). The answer - persuasively argued by Katrina Manson - is a resounding yes. What more could you want from a book - an important topic issues that ties together varied intellectual strands in a way that prior books have not seemed to match? I was hooked!
What else to say? The book is well written and easy to follow. The characters (especially Cukor, but others as well) are sharply drawn and believable. The linkage of the worlds of high-level combat skills, government bureaucracy, geopolitics, and AI details/minutia works well and shows me that Ms. Manson has benefitted from some superb editing.
Overall, this is a superb book that is well worth reading.
Why this book? I started by being interested in AI with all the hype about it. However, reading general books on various forms of AI led me either advanced coding/modelling tomes or else to the overall assessments of commentators that the area and its hype was overblown. Neither of these was interesting to me - the first due to lack of knowledge/skill and the second due to doubt regarding the skepticism - since there seemed to be something to the hype.
So…I read some materials on the firms involved, where people were investing a lot and had grand expectations regarding firms and their products. For all the involved firms, there was much to read but I narrowed in on Anthropic and Palantir, which seemed different from the normal hype around the more prominent tech firms. With Palantir (and Karp) the link to government work and the applications of AI to warfare stand out - especially given recent geopolitical developments. It does not take much effort to focus in on Drew Cukor and Project Maven.
This got me to what for me is the most interesting aspect of AI - the linking of what is incredibly general and abstract with what is as particular as it gets - the application of AI to modern combat situations and whether this application can boost the performance, efficiency, etc. of modern warfighters (and change the nature of modern warfare). The answer - persuasively argued by Katrina Manson - is a resounding yes. What more could you want from a book - an important topic issues that ties together varied intellectual strands in a way that prior books have not seemed to match? I was hooked!
What else to say? The book is well written and easy to follow. The characters (especially Cukor, but others as well) are sharply drawn and believable. The linkage of the worlds of high-level combat skills, government bureaucracy, geopolitics, and AI details/minutia works well and shows me that Ms. Manson has benefitted from some superb editing.
Overall, this is a superb book that is well worth reading.
May 18, 2026
Project Maven is the story of a secret decade-long military project of mass surveillance and autonomous optimization of “kill chains” spearheaded by US Marine, Drew Cukor.
Katrina Mason delivers a masterclass in investigative reporting on the accuracy of AI weaponry and the ethics of autonomous warfare. Mason’s reporting provides deeper insight into the partnership between Silicon Valley and the US military. She provides evidence of military AI being deployed in wars abroad and to America’s own citizens (shout out the imperial boomerang!) on a tremendous scale dubiously matched by Big Tech’s invasion into politics. Mason’s critique is most definitely acute, but I appreciate how she approaches these architects of AI warfare with striking empathy. Although these men are quoted speaking about geopolitical strategy in ways I find to be devoid of humanity, Mason makes clear their intentions behind Project Maven were altruistic; these men felt it their duty to stop any more soldiers dying like so many of their friends did- as a byproduct of failures in American military technology. Like many cautionary tales, Project Maven’s story begins with a group of people eager to create something for the good of society, only for the creation to turn far more sinister when money, power, greed, and in America’s case, Alex Karp, are added to the mix. In the least fear monger way possible, I think everyone should read this book!
Katrina Mason delivers a masterclass in investigative reporting on the accuracy of AI weaponry and the ethics of autonomous warfare. Mason’s reporting provides deeper insight into the partnership between Silicon Valley and the US military. She provides evidence of military AI being deployed in wars abroad and to America’s own citizens (shout out the imperial boomerang!) on a tremendous scale dubiously matched by Big Tech’s invasion into politics. Mason’s critique is most definitely acute, but I appreciate how she approaches these architects of AI warfare with striking empathy. Although these men are quoted speaking about geopolitical strategy in ways I find to be devoid of humanity, Mason makes clear their intentions behind Project Maven were altruistic; these men felt it their duty to stop any more soldiers dying like so many of their friends did- as a byproduct of failures in American military technology. Like many cautionary tales, Project Maven’s story begins with a group of people eager to create something for the good of society, only for the creation to turn far more sinister when money, power, greed, and in America’s case, Alex Karp, are added to the mix. In the least fear monger way possible, I think everyone should read this book!
May 15, 2026
Authors need to start writing with the readers in mind. This is a very interesting topic that was bogged down in the minutiae.
April 13, 2026
Not what I was expecting. I was hoping to learn more about the use of AI in defense, especially how it relates to current events.
May 23, 2026
This is a hard book to rate. On one hand, the author gathered a lot of factual information about the development of AI for use in warfare, and seemingly tries to present it in a "neutral" fashion. However, no presentation is neutral, and the book didn't feel nuanced enough at the human level to round out this topic. The author seems to attempt, at the 11th hour, to craft an antihero/redemption arc for Cukol at the end, but it doesn't ring true.
The Tl;dr of this book is that killer robots and Skynet are already here, pioneered by Cukol and his team using a combination of military machismo and techbro hubris, oiled by savvy bureaucratic skills. The military and corporate AI interests are hopelessly entangled, with the latter exacerbating the deterioration of any ethical considerations under the guise of "saving human lives," while knowingly using targeting systems that are glitchy AF. Except, unlike in a buggy videogame release, where you get pissed-off gamers, you get dead civilians, but this is accepted as the price for iteratively improving the algorithms now dominating modern warfare, which will likely (and already have in some cases) undergo the boomerang effect and be used domestically.
P.S. One would think it would be obvious that you are working on Something Bad, when "shorten the kill chain" is the goal, and it's stated repeatedly.
P.P.S. Giant tech corporations want (to keep receiving) a piece of this multi-trillion dollar AI warfare pie, which is one reason they're interested in building data centers and making people foot the bill for AI warfare at least twice, in data centers and in taxes. Unfortunately, as we have historically seen, warfare is the driving force behind scientific and technological "advancement." No one bothers to use the political and financial will to create something the scale of the Manhattan Project to solve climate change 🤷♂️
The Tl;dr of this book is that killer robots and Skynet are already here, pioneered by Cukol and his team using a combination of military machismo and techbro hubris, oiled by savvy bureaucratic skills. The military and corporate AI interests are hopelessly entangled, with the latter exacerbating the deterioration of any ethical considerations under the guise of "saving human lives," while knowingly using targeting systems that are glitchy AF. Except, unlike in a buggy videogame release, where you get pissed-off gamers, you get dead civilians, but this is accepted as the price for iteratively improving the algorithms now dominating modern warfare, which will likely (and already have in some cases) undergo the boomerang effect and be used domestically.
P.S. One would think it would be obvious that you are working on Something Bad, when "shorten the kill chain" is the goal, and it's stated repeatedly.
P.P.S. Giant tech corporations want (to keep receiving) a piece of this multi-trillion dollar AI warfare pie, which is one reason they're interested in building data centers and making people foot the bill for AI warfare at least twice, in data centers and in taxes. Unfortunately, as we have historically seen, warfare is the driving force behind scientific and technological "advancement." No one bothers to use the political and financial will to create something the scale of the Manhattan Project to solve climate change 🤷♂️
August 21, 2026
Interesting. But different from what I expected: I bought this because I'm interested in the tech. But the book focuses more on the military and the individual soldiers involved in project Maven.
This is not meant as criticism, simply a heads-up. People reading the subtitle would be clued up to this.
As stated above, the bulk of the narrative concerns the persona of colonel Drew Cukor, the lead and driving force behind Maven, and his team of military-disrupters. It's fascinating to read how they manage to bring some of the Silicon Valley ethos to the Pentagon, and how private non-military software companies - Palantir, Microsoft, Google, Amazon AWS and many others - take up prominent roles and push the legacy military complex to the back.
At times, the book goes into too much detail, oodles of pages devoted to politicking and inter-departmental powerplays, and I would have preferred a focus on the main actors instead of including so many bitplayers.
The big question that remains unanswered is: Is AI on the battlefield, and especially in the form of autonomous weapons, overall beneficial or detrimental: On the upside, AI may reduce (human) error rates and thus minimize collateral damage. On the downside, AI makes it possible to strike many more targets, thus increasing the death toll. And by limiting or removing humans from the kill chain (and by reducing the risk to your own soldiers), any morals and hesitancy may be removed as well.
Examples where AI is already used in war are found in Ukraine and Gaza.
A hugely important and very, very fraught topic.
This is not meant as criticism, simply a heads-up. People reading the subtitle would be clued up to this.
As stated above, the bulk of the narrative concerns the persona of colonel Drew Cukor, the lead and driving force behind Maven, and his team of military-disrupters. It's fascinating to read how they manage to bring some of the Silicon Valley ethos to the Pentagon, and how private non-military software companies - Palantir, Microsoft, Google, Amazon AWS and many others - take up prominent roles and push the legacy military complex to the back.
At times, the book goes into too much detail, oodles of pages devoted to politicking and inter-departmental powerplays, and I would have preferred a focus on the main actors instead of including so many bitplayers.
The big question that remains unanswered is: Is AI on the battlefield, and especially in the form of autonomous weapons, overall beneficial or detrimental: On the upside, AI may reduce (human) error rates and thus minimize collateral damage. On the downside, AI makes it possible to strike many more targets, thus increasing the death toll. And by limiting or removing humans from the kill chain (and by reducing the risk to your own soldiers), any morals and hesitancy may be removed as well.
Examples where AI is already used in war are found in Ukraine and Gaza.
A hugely important and very, very fraught topic.
April 11, 2026
I finished Project Maven and immediately bought several extra copies to give to friends. That’s how strongly I feel about it.
This is a meaty, high-stakes story, but Manson makes it remarkably accessible. She expertly sets the scene and chronicles the journey of a small, scrappy group of data-driven outsiders tasked with solving a battlefield problem of enormous consequence. What could have been dry or overly technical instead reads with urgency and momentum.
It’s clear how deeply reported this book is. Years of research show up not just in the facts, but in the texture - how vividly each individual involved in Project Maven is drawn. The characterization is one of the book’s greatest strengths; these aren’t abstract operators, but real people navigating ambiguity, pressure, and ethical gray zones.
At the center is Drew Cukor - a figure you may find inspiring, unsettling, or both. As the driving force behind Maven, he leads a high-stakes effort to build an AI-enabled system capable of transforming modern warfare. Manson resists caricature and instead presents him as a complex, relentlessly driven leader pursuing what he sees as a necessary evolution in military capability.
What lingered with me most after finishing the book wasn’t just what Maven achieved, but what it might unleash.
Could a small team of pioneers, led by someone as determined as Cukor, create a system so efficient at identifying and neutralizing threats that it accelerates us toward something far more dangerous, an era of uncontrollable algorithmic escalation? In removing the slower, more deliberate human layer to save lives and better protect innocent civilians, are we building a shield for humanity, or engineering a Schlimmbesserung, an improvement that ultimately makes things worse?
The question feels especially urgent given emerging research suggesting that advanced AI systems, when placed in simulated conflict scenarios, can display a measurable bias toward escalatory, even catastrophic-decisions under pressure.
Is this the future we’re building toward?
This book doesn’t offer easy answers, but it forces you to confront the paradoxical question. And that alone makes it essential reading.
This is a meaty, high-stakes story, but Manson makes it remarkably accessible. She expertly sets the scene and chronicles the journey of a small, scrappy group of data-driven outsiders tasked with solving a battlefield problem of enormous consequence. What could have been dry or overly technical instead reads with urgency and momentum.
It’s clear how deeply reported this book is. Years of research show up not just in the facts, but in the texture - how vividly each individual involved in Project Maven is drawn. The characterization is one of the book’s greatest strengths; these aren’t abstract operators, but real people navigating ambiguity, pressure, and ethical gray zones.
At the center is Drew Cukor - a figure you may find inspiring, unsettling, or both. As the driving force behind Maven, he leads a high-stakes effort to build an AI-enabled system capable of transforming modern warfare. Manson resists caricature and instead presents him as a complex, relentlessly driven leader pursuing what he sees as a necessary evolution in military capability.
What lingered with me most after finishing the book wasn’t just what Maven achieved, but what it might unleash.
Could a small team of pioneers, led by someone as determined as Cukor, create a system so efficient at identifying and neutralizing threats that it accelerates us toward something far more dangerous, an era of uncontrollable algorithmic escalation? In removing the slower, more deliberate human layer to save lives and better protect innocent civilians, are we building a shield for humanity, or engineering a Schlimmbesserung, an improvement that ultimately makes things worse?
The question feels especially urgent given emerging research suggesting that advanced AI systems, when placed in simulated conflict scenarios, can display a measurable bias toward escalatory, even catastrophic-decisions under pressure.
Is this the future we’re building toward?
This book doesn’t offer easy answers, but it forces you to confront the paradoxical question. And that alone makes it essential reading.
May 10, 2026
BAHAHAHAHAHA Katrina Manson you're brilliant.
This book was the final straw in me realizing I want absolutely nothing to do with the world of security and defence anymore!
Somewhat an exposé (as much as it can be without anyone losing their jobs or being disappeared), somewhat a mystifying character study, and part cultural study on new big tech culture.
There's a vast gender gap in the AI resistance space, especially in the realm of security. I can understand women's (im using 'women' as a blanket and an inclusive term here) hesitancy to embrace AI in other contexts such as in communications, education, advertising, general surveillance etc, but I would love to talk to people about what specifically it is in the world of defence which creates a gendered hesitancy.
We have three copies of this book at work right now, and I'm waiting for my favourite customer to come in again so I can recommend it to him.
This book was the final straw in me realizing I want absolutely nothing to do with the world of security and defence anymore!
Somewhat an exposé (as much as it can be without anyone losing their jobs or being disappeared), somewhat a mystifying character study, and part cultural study on new big tech culture.
There's a vast gender gap in the AI resistance space, especially in the realm of security. I can understand women's (im using 'women' as a blanket and an inclusive term here) hesitancy to embrace AI in other contexts such as in communications, education, advertising, general surveillance etc, but I would love to talk to people about what specifically it is in the world of defence which creates a gendered hesitancy.
We have three copies of this book at work right now, and I'm waiting for my favourite customer to come in again so I can recommend it to him.
May 23, 2026
Let's not beat about the bush.
This book is about how the United States has acquired a new ability to mete out death on an unprecedented scale.
News reports these days do not suggest that wars initiated by the US or any other nation are controlled by autonomous weapons systems issuing the targets and pulling the trigger, but it's pretty hard to escape the notion that something's afoot.
In the first days of the current US-Israel attacks on Iran in excess of 5,000 targets were hit by the aggressors. In Israel wave after wave of attacks on Gaza leveled large populous areas. On the high seas US ships are eliminating suspected drug shipments and the humans piloting those boats.
Project Maven started as the dream child of Marine Colonel Drew Cukor to automate the sharing of intelligence across US defense services, mutated into an automated system for targeting enemy fire, and now seems to be the foundation for lethal autonomous weapons systems.
Author Katrina Manson got my attention right out of the gate when she tells us that Project Maven is only one of an astounding 800 AI-related projects at the Pentagon. What we learn further on in the story is that Maven is now not only accepted by the generals in all the services, but is now a serious part of NATO, individual members of NATO, and integrated into the NORAD infrastructure guarding my home country Canada.
A salient part of this story is how these initially far out ideas of fighting wars gained credence with the military. It wasn't as though the Pentagon immediately turned on the spigots and the money started flowing. It was a relatively long, painful process for Colonel Cukor.
And it was a long, painful process for Palantir and a few other tech firms to become the chosen developers and suppliers of key computer algorithms to power Maven. But one once investors realized enormous profits were to be made from defense contracts, public capital flowed in help hire top talent and spoon-fed advanced technology into the Pentagon.
And that seems to be how it goes these days: regardless of the initial safety concerns around the development of artificial intelligence - and this book doesn't even begin to discuss the implications of artificial general intelligence -- the rich financial incentives kick in and the stuff gets built...before safety concerns are adequately addressed.
You may remember a few years ago when a lot of Google Brain employees flat out refused to work on military defense contracts, in particular Project Maven. Google was forced to back off the project.
What happened?
Microsoft stepped in. Palantir stepped in. The stuff got built.
I'm a little surprised some enterprising Google shareholders didn't sue management for walking away from these lucrative opportunities.
I was drawn to the discussion of international attempts to regulate the construction and use of lethal autonomous weapons.
The US, China, and Russia were all content to let the discussions drone on and on with no resolution in sight.
But is an "autonomous" system of death really autonomous? Somebody is responsible for getting them up and running, for the hits and the misses.
This book is about how the United States has acquired a new ability to mete out death on an unprecedented scale.
News reports these days do not suggest that wars initiated by the US or any other nation are controlled by autonomous weapons systems issuing the targets and pulling the trigger, but it's pretty hard to escape the notion that something's afoot.
In the first days of the current US-Israel attacks on Iran in excess of 5,000 targets were hit by the aggressors. In Israel wave after wave of attacks on Gaza leveled large populous areas. On the high seas US ships are eliminating suspected drug shipments and the humans piloting those boats.
Project Maven started as the dream child of Marine Colonel Drew Cukor to automate the sharing of intelligence across US defense services, mutated into an automated system for targeting enemy fire, and now seems to be the foundation for lethal autonomous weapons systems.
Author Katrina Manson got my attention right out of the gate when she tells us that Project Maven is only one of an astounding 800 AI-related projects at the Pentagon. What we learn further on in the story is that Maven is now not only accepted by the generals in all the services, but is now a serious part of NATO, individual members of NATO, and integrated into the NORAD infrastructure guarding my home country Canada.
A salient part of this story is how these initially far out ideas of fighting wars gained credence with the military. It wasn't as though the Pentagon immediately turned on the spigots and the money started flowing. It was a relatively long, painful process for Colonel Cukor.
And it was a long, painful process for Palantir and a few other tech firms to become the chosen developers and suppliers of key computer algorithms to power Maven. But one once investors realized enormous profits were to be made from defense contracts, public capital flowed in help hire top talent and spoon-fed advanced technology into the Pentagon.
And that seems to be how it goes these days: regardless of the initial safety concerns around the development of artificial intelligence - and this book doesn't even begin to discuss the implications of artificial general intelligence -- the rich financial incentives kick in and the stuff gets built...before safety concerns are adequately addressed.
You may remember a few years ago when a lot of Google Brain employees flat out refused to work on military defense contracts, in particular Project Maven. Google was forced to back off the project.
What happened?
Microsoft stepped in. Palantir stepped in. The stuff got built.
I'm a little surprised some enterprising Google shareholders didn't sue management for walking away from these lucrative opportunities.
I was drawn to the discussion of international attempts to regulate the construction and use of lethal autonomous weapons.
The US, China, and Russia were all content to let the discussions drone on and on with no resolution in sight.
But is an "autonomous" system of death really autonomous? Somebody is responsible for getting them up and running, for the hits and the misses.
May 3, 2026
К счастью для всех нас, в богоспасаемом краю духовных скреп перевод этой книги не опубликуют никогда
April 10, 2026
AI in the 60s
Project Maven brings back some memories. Our small scientific software company was under contract with a major defense contractor in Orlando Florida. The task was to support development of a helicopter borne system to recognize tanks and other large targets using IR imagery.
A hybrid analog-digital computer provided the speed required for the helo to rise, take an IR picture, analyze the scene and then fire or not and then descend to safety. It was a very interesting job.
Later, we were the prime contractor to provide the US Air Force with a complete algorithm development system at Eglin AFB in Florida.
Project Maven brings back some memories. Our small scientific software company was under contract with a major defense contractor in Orlando Florida. The task was to support development of a helicopter borne system to recognize tanks and other large targets using IR imagery.
A hybrid analog-digital computer provided the speed required for the helo to rise, take an IR picture, analyze the scene and then fire or not and then descend to safety. It was a very interesting job.
Later, we were the prime contractor to provide the US Air Force with a complete algorithm development system at Eglin AFB in Florida.
April 4, 2026
New Tech are we up for it.
Good review of the growth of how to use algorithms and AI tech for warfare. Not as clean as we think.
I find the lack of hubris very interesting as the world moves forward with technology. The author presents this as we see how the AI tech unfolds in the Middle Esat.
Good review of the growth of how to use algorithms and AI tech for warfare. Not as clean as we think.
I find the lack of hubris very interesting as the world moves forward with technology. The author presents this as we see how the AI tech unfolds in the Middle Esat.
September 16, 2026
In this book, Katarina Manson presents the history of how a Marine officedr, Drew Cukor, created and developed Project Maven into one of the most important AI targeting systems being used now by the US military, as part of the revolution in current military technology. The story presented is not just about the technology but largely how Cukor fought government bureaucracy, military hierarchy, and numerous competing private technology interests, like Microsoft, Palantir and Anduril, to create this system. Along the way, critical moral questions were raised by some about the morality of developing a system that could identify and destroy targets without human intervention. Errors by Maven and other targeting systems are described where erroneous deaths resulted, including some deaths by US servicemen and civilians in Afghanistan and Iraq. The style of writing however makes the book somewhat difficult to read; the multitude of quotes causes the development of the themes to drag unnecessarily. At the end of the book, questions remain, as normal rules of engagement have become blurred resulting in fears about how deadly future wars will become.
June 8, 2026
There’s a lot of information in this book. It’s the secretive history of the Pentagon’s effort to integrate and apply AI into the military and war. The author did an exhaustive amount of research and presents it in a very straightforward manner. There’s a lot to absorb and it doesn’t read quickly but she does a good job of separating the hype from reality and associating individuals along the way. There are a lot of names, government agencies and acronyms to track along the way. It’s a fascinating examination of the future of war as we will know it.
July 18, 2026
I learned a lot from this book about the nature of modern warfare and about the wars my country has been fighting throughout my adult life. Because they’ve happened on the other side of the world, I’ve had the luxury of not knowing much about them. The chapter on the USA’s use of AI to assist Ukraine was especially interesting. I did get a bit lost in the jumble of different AI technologies, military commands and departments.
June 30, 2026
Artificial intelligence has inspired no shortage of books promising to explain how tomorrow's wars will be fought. Many of them read like technical manuals interrupted by occasional existential crises. Katrina Manson takes a decidedly different approach. Project Maven is less interested in algorithms than in the wonderfully imperfect humans who decided to build them. If you were hoping for a deep technical examination of the software, prepare instead for a guided tour through the ambitions, rivalries, convictions, and anxieties of the people standing behind the keyboard.
That is not necessarily a criticism. It is, in fact, the book's greatest strength.
Manson tells the story of Project Maven largely through the personalities who shaped it, particularly Marine Colonel Drew Cukor and the eclectic coalition of military officers, technologists, bureaucrats, skeptics, and industry leaders who transformed an obscure Pentagon initiative into one of the Department of Defense's most consequential artificial intelligence programs. Because much of Maven's operational history remains classified, this emphasis is understandable. One can hardly describe in detail what remains hidden behind secure doors, so Manson instead illuminates the people carrying the keys.
Readers looking for a detailed chronology of every technical breakthrough, operational deployment, or classified capability may leave slightly hungry. The book frequently gestures toward remarkable accomplishments before politely reminding the audience that national security still exists. It is rather like being invited to dinner at Area 51 and discovering the menu consists primarily of fascinating conversations about the chef.
Fortunately, Manson is an engaging guide through those conversations.
Where Project Maven truly distinguishes itself is its sustained exploration of the ethical questions surrounding military artificial intelligence. Every chapter seems to ask a variation of the same uncomfortable question: What happens when machines increasingly assist, influence, or eventually participate in decisions involving human life? Rather than presenting caricatures of reckless technologists or technophobic idealists, Manson treats nearly every participant as someone wrestling with genuine moral uncertainty. She resists the temptation to assign heroes and villains, allowing readers to appreciate that many of the architects of military AI were motivated not by fantasies of autonomous killer robots but by a sincere desire to reduce friendly casualties, process overwhelming amounts of intelligence, and make battlefield decisions more accurate.
That balance is refreshing. The people involved are neither saints nor comic-book antagonists. They are professionals attempting to solve difficult problems while carrying the uncomfortable knowledge that every solution introduces new risks.
Yet this same focus also reveals the book's principal limitation.
Manson devotes enormous intellectual energy to examining the moral implications of American AI development. The ethical concerns receive careful, thoughtful treatment, and deservedly so. But the discussion often unfolds as though the greatest variable in the strategic equation is whether the United States chooses to proceed.
The uncomfortable reality, however, is that strategic competition rarely grants the luxury of unilateral moral reflection.
Readers searching for a comparable examination of Chinese, Russian, or other adversarial AI programs will find comparatively little. Those competitors appear largely as motivations driving Project Maven forward rather than as fully developed actors pursuing their own technological revolutions. This creates an odd asymmetry. We spend hundreds of pages interrogating the conscience of democratic societies while only briefly considering what happens if authoritarian competitors simply decline to participate in the same ethical debate.
To her credit, Manson rarely states outright what conclusions readers must reach. She generally avoids overt advocacy, allowing her subjects to explain their own motivations and decisions. Yet neutrality in tone is not always neutrality in construction. The selection of evidence, the amount of narrative space devoted to particular arguments, and the questions repeatedly posed all gently, but unmistakably, steer the reader toward a particular interpretation of military AI and its consequences.
That becomes most apparent in the treatment of ethical concerns. The book thoughtfully presents the moral dilemmas surrounding Project Maven, but it often affords those concerns substantially greater evidentiary weight than the strategic arguments supporting the program. Readers are encouraged to contemplate the dangers of AI-enabled warfare in considerable depth, while comparatively less attention is devoted to the consequences of not developing such capabilities in an increasingly competitive international environment. The result is less an overt argument than a carefully landscaped path whose destination becomes increasingly difficult to miss.
The irony is almost amusing. The book repeatedly asks whether democracies should proceed cautiously with military AI, but spends comparatively little time examining what happens when authoritarian competitors answer the same question with a collective shrug and a larger research budget. Ethical restraint is unquestionably admirable. It is considerably less reassuring when your competitors view it as a scheduling inconvenience.
None of this diminishes the importance of the ethical questions Manson raises. They deserve thoughtful examination and serious public debate. But the case presented occasionally feels incomplete. The narrative asks readers to weigh profound moral costs without devoting equal attention to the strategic costs of abstention, leaving the overall argument feeling more persuasive rhetorically than it is fully supported analytically.
In that sense, Project Maven is less a history of a Pentagon program than a meditation on responsibility in the age of artificial intelligence.
For readers seeking an operational history of AI-enabled military capabilities, the book will likely feel incomplete. Those looking for technical discussions of machine learning models, targeting systems, or intelligence workflows may find themselves wishing the classification guides had been just a little less enthusiastic. But for readers interested in understanding how institutions change, how personalities shape national security, and why ethical debates surrounding military AI refuse to stay confined to military circles, Manson has written a compelling and highly readable account.
Ultimately, Project Maven is less about artificial intelligence than about human intelligence confronting its own creation. It succeeds in introducing readers to the remarkable individuals who shepherded one of the Pentagon's most transformative initiatives while thoughtfully exploring the ethical questions that inevitably followed. At the same time, its treatment of those questions reflects a perspective that gives considerably more attention to the moral hazards of American innovation than to the strategic consequences of falling behind less constrained competitors. That imbalance does not make the book unworthy; indeed, it is well worth reading precisely because it provokes these questions. But readers should recognize that while Manson invites them to think for themselves, she has carefully arranged the furniture so that they are more likely to end up sitting in one particular chair.
That is not necessarily a criticism. It is, in fact, the book's greatest strength.
Manson tells the story of Project Maven largely through the personalities who shaped it, particularly Marine Colonel Drew Cukor and the eclectic coalition of military officers, technologists, bureaucrats, skeptics, and industry leaders who transformed an obscure Pentagon initiative into one of the Department of Defense's most consequential artificial intelligence programs. Because much of Maven's operational history remains classified, this emphasis is understandable. One can hardly describe in detail what remains hidden behind secure doors, so Manson instead illuminates the people carrying the keys.
Readers looking for a detailed chronology of every technical breakthrough, operational deployment, or classified capability may leave slightly hungry. The book frequently gestures toward remarkable accomplishments before politely reminding the audience that national security still exists. It is rather like being invited to dinner at Area 51 and discovering the menu consists primarily of fascinating conversations about the chef.
Fortunately, Manson is an engaging guide through those conversations.
Where Project Maven truly distinguishes itself is its sustained exploration of the ethical questions surrounding military artificial intelligence. Every chapter seems to ask a variation of the same uncomfortable question: What happens when machines increasingly assist, influence, or eventually participate in decisions involving human life? Rather than presenting caricatures of reckless technologists or technophobic idealists, Manson treats nearly every participant as someone wrestling with genuine moral uncertainty. She resists the temptation to assign heroes and villains, allowing readers to appreciate that many of the architects of military AI were motivated not by fantasies of autonomous killer robots but by a sincere desire to reduce friendly casualties, process overwhelming amounts of intelligence, and make battlefield decisions more accurate.
That balance is refreshing. The people involved are neither saints nor comic-book antagonists. They are professionals attempting to solve difficult problems while carrying the uncomfortable knowledge that every solution introduces new risks.
Yet this same focus also reveals the book's principal limitation.
Manson devotes enormous intellectual energy to examining the moral implications of American AI development. The ethical concerns receive careful, thoughtful treatment, and deservedly so. But the discussion often unfolds as though the greatest variable in the strategic equation is whether the United States chooses to proceed.
The uncomfortable reality, however, is that strategic competition rarely grants the luxury of unilateral moral reflection.
Readers searching for a comparable examination of Chinese, Russian, or other adversarial AI programs will find comparatively little. Those competitors appear largely as motivations driving Project Maven forward rather than as fully developed actors pursuing their own technological revolutions. This creates an odd asymmetry. We spend hundreds of pages interrogating the conscience of democratic societies while only briefly considering what happens if authoritarian competitors simply decline to participate in the same ethical debate.
To her credit, Manson rarely states outright what conclusions readers must reach. She generally avoids overt advocacy, allowing her subjects to explain their own motivations and decisions. Yet neutrality in tone is not always neutrality in construction. The selection of evidence, the amount of narrative space devoted to particular arguments, and the questions repeatedly posed all gently, but unmistakably, steer the reader toward a particular interpretation of military AI and its consequences.
That becomes most apparent in the treatment of ethical concerns. The book thoughtfully presents the moral dilemmas surrounding Project Maven, but it often affords those concerns substantially greater evidentiary weight than the strategic arguments supporting the program. Readers are encouraged to contemplate the dangers of AI-enabled warfare in considerable depth, while comparatively less attention is devoted to the consequences of not developing such capabilities in an increasingly competitive international environment. The result is less an overt argument than a carefully landscaped path whose destination becomes increasingly difficult to miss.
The irony is almost amusing. The book repeatedly asks whether democracies should proceed cautiously with military AI, but spends comparatively little time examining what happens when authoritarian competitors answer the same question with a collective shrug and a larger research budget. Ethical restraint is unquestionably admirable. It is considerably less reassuring when your competitors view it as a scheduling inconvenience.
None of this diminishes the importance of the ethical questions Manson raises. They deserve thoughtful examination and serious public debate. But the case presented occasionally feels incomplete. The narrative asks readers to weigh profound moral costs without devoting equal attention to the strategic costs of abstention, leaving the overall argument feeling more persuasive rhetorically than it is fully supported analytically.
In that sense, Project Maven is less a history of a Pentagon program than a meditation on responsibility in the age of artificial intelligence.
For readers seeking an operational history of AI-enabled military capabilities, the book will likely feel incomplete. Those looking for technical discussions of machine learning models, targeting systems, or intelligence workflows may find themselves wishing the classification guides had been just a little less enthusiastic. But for readers interested in understanding how institutions change, how personalities shape national security, and why ethical debates surrounding military AI refuse to stay confined to military circles, Manson has written a compelling and highly readable account.
Ultimately, Project Maven is less about artificial intelligence than about human intelligence confronting its own creation. It succeeds in introducing readers to the remarkable individuals who shepherded one of the Pentagon's most transformative initiatives while thoughtfully exploring the ethical questions that inevitably followed. At the same time, its treatment of those questions reflects a perspective that gives considerably more attention to the moral hazards of American innovation than to the strategic consequences of falling behind less constrained competitors. That imbalance does not make the book unworthy; indeed, it is well worth reading precisely because it provokes these questions. But readers should recognize that while Manson invites them to think for themselves, she has carefully arranged the furniture so that they are more likely to end up sitting in one particular chair.
October 5, 2026
With AI in the headlines, especially with it deploying in the military, this book is a great way to dive into how it started and where it is going through the eyes of Project Maven. Marine Colonel Cukor drove the initial efforts to bring AI into the intelligence and planning at the operational level. What I found was a rushed effort to push AI algorithms into as many places within the US military’s commands. I think there are a lot of flaws they didn’t grasp, as they didn’t understand the fundamentals of the technology they thought was the holy grail of computer magic.
Project Maven was started so they could manage the thousands of hours of drone footage, having the machine understand what is there so a human doesn’t have to watch it 24/7. The goal was to have it help target people and places of interest. Sometimes it could do this well and other times it caused a huge mess.
One problem was the training data. I remember reading about an effort to teach a computer what was a tank so it could identify it. What the researchers in the mid-1990’s didn’t understand was how the neural net taught itself. The result of that project was the computer knowing when it was sunny or cloudy. The images used had a tank during sunny days (if I remember correctly) and an empty field was cloudy. This is a fundamental problem with current AI systems. They can learn, but what happens within the black box of the model is not fully understood. Colonel Cukor does put a huge effort in very clear & well done labelling of training images. There is a chapter on how he fought to keep it away from other groups in the Pentagon.
But the limitations of the models came to the fore when the Russian invasion of Ukraine occurred. The models up to that time were trained on images from Iraq & Afghanistan. Using the same models to find vehicles and soldiers, it failed. Different environment, different objects. Once the model is outside of its bounded box, it’ll flop, making any type of targeting useless.
Most recently, Project Maven was involved with setting targets in Iran, notably the strike on the girls school by a Tomahawk. Oh how the US military tried to brush it off, that it was Iran who did the strike or that it was a legitimate target. Only when the evidence became overwhelming did the US military say it was due to old data in the targeting system. Maven was blamed, which somehow made it better. But its target list wasn’t vetted. Examples of this are shown in the book, where the system can produce a huge number of targets an officer then clicks yes or no to.
Palantir is front and center of Project Maven. They are an arogant bunch that oversold what they had and rushed to build something that worked. I found the task of building a video player from scratch a bit funny. They looked to Netflix as an example. But there was already commercial software that had a lot of what they wanted to do. For a time I was in the security camera and recording system space. Multiple camera feeds, follow people, intelligence at the edge, motion tracking, etc. This was table stakes more than 10 years ago (more like 15-20). I think the environment everyone was placed in under Colonel Cukor was like the hell of working in Silicon Valley. Except without the stock options.
At one point in the book, when some of the teams were telling him to slow down on the demands, he says “American taxpayers don’t pay you to rest.” Oh, how he really believed that working most of your waking hours results in super output. It might for him, as he didn’t seem to slow down or care that people needed breaks. Most human minds break under that kind of strain and working more hours causes more mistakes. Silicon Valley is infected with this type of thinking and it destroys people. The quote really pushes into the vibe of “I don’t care about you, just results.” It is a wonder it works at all if everyone involved is super exhausted and stressed.
It is deployed across the entire US military, for good or bad. There is a lack of training on the sytem, especially of how it can go wrong. The models are not 100% accurate, which Cukor acknowledges. But the military commands treat its output as accurate. They want to believe the machine and blame bad data if it causes things to go wrong. Instead of it working along side humans as Cukor envisioned, it is positioned to shorten the kill chain to the point where the machine does the action.
The chapters where Ukraine was involved are very telling from the US military point of view. They thought the Ukrainian data systems were not secure, but they were more secure than most of the US military. The secure network the US military was using was so poorly designed, it had packets bouncing back and forth several times across the Atlantic before finding where they actually needed to be. Someone thought it was a good design at some point. The effort required to fix all of it was emense, especially when those involved didn’t go through proper channels. The Ukrainians, on the other hand, were building out their own drone targeting systems that didn’t need the fancy American hardware. The US military was still thinking big, while Ukraine was thinking fast and flexible. The US military still hasn’t understood the lessons of Ukraine in the future of warfare. It isn’t “AI everywhere” or “muscled warfighters.” Its being nimble and using cheap drones to dominate.
I do have a bone to pick with the author. She mentions the Winchester Mystery House, as Kristen Cukor brings up the mansion and its crazed builder. I think it is a good reflection of Kristen Cukor to ponder if she is a modern day Sarah Winchester. The author can’t get the right place in California. She says it is “...the curious 1880s Santa Cruz mansion...” No, it is in San Jose & Sarah Winchester kept building until she died in 1922. Please get your landmarks right, especially when one of the people interviewed is making a point of the house (it is also the title of Chapter 26).
I think the author did a wonderful job in illustrating the ins and outs of Project Maven. From the book and what I know of AI models, I don’t believe the folks in the military fully comprehend how such systems can go wrong, which is often. Current models will always product a large amount of wrong answers. It is built right into how the models act. But the military is embracing AI as really good, having it make targeting decisions that are instantly executed. At one point there is a warning about how technologies developed for overseas always come back to be used in the homeland. Maven or something like it could easily be retasked. Flock cameras are in the theme of Maven, tracking targets without regard to privacy. I believe the use of AI will cause more problems than it solves, making the kill chain so short a lot of innocent people are harmed.
Project Maven was started so they could manage the thousands of hours of drone footage, having the machine understand what is there so a human doesn’t have to watch it 24/7. The goal was to have it help target people and places of interest. Sometimes it could do this well and other times it caused a huge mess.
One problem was the training data. I remember reading about an effort to teach a computer what was a tank so it could identify it. What the researchers in the mid-1990’s didn’t understand was how the neural net taught itself. The result of that project was the computer knowing when it was sunny or cloudy. The images used had a tank during sunny days (if I remember correctly) and an empty field was cloudy. This is a fundamental problem with current AI systems. They can learn, but what happens within the black box of the model is not fully understood. Colonel Cukor does put a huge effort in very clear & well done labelling of training images. There is a chapter on how he fought to keep it away from other groups in the Pentagon.
But the limitations of the models came to the fore when the Russian invasion of Ukraine occurred. The models up to that time were trained on images from Iraq & Afghanistan. Using the same models to find vehicles and soldiers, it failed. Different environment, different objects. Once the model is outside of its bounded box, it’ll flop, making any type of targeting useless.
Most recently, Project Maven was involved with setting targets in Iran, notably the strike on the girls school by a Tomahawk. Oh how the US military tried to brush it off, that it was Iran who did the strike or that it was a legitimate target. Only when the evidence became overwhelming did the US military say it was due to old data in the targeting system. Maven was blamed, which somehow made it better. But its target list wasn’t vetted. Examples of this are shown in the book, where the system can produce a huge number of targets an officer then clicks yes or no to.
Palantir is front and center of Project Maven. They are an arogant bunch that oversold what they had and rushed to build something that worked. I found the task of building a video player from scratch a bit funny. They looked to Netflix as an example. But there was already commercial software that had a lot of what they wanted to do. For a time I was in the security camera and recording system space. Multiple camera feeds, follow people, intelligence at the edge, motion tracking, etc. This was table stakes more than 10 years ago (more like 15-20). I think the environment everyone was placed in under Colonel Cukor was like the hell of working in Silicon Valley. Except without the stock options.
At one point in the book, when some of the teams were telling him to slow down on the demands, he says “American taxpayers don’t pay you to rest.” Oh, how he really believed that working most of your waking hours results in super output. It might for him, as he didn’t seem to slow down or care that people needed breaks. Most human minds break under that kind of strain and working more hours causes more mistakes. Silicon Valley is infected with this type of thinking and it destroys people. The quote really pushes into the vibe of “I don’t care about you, just results.” It is a wonder it works at all if everyone involved is super exhausted and stressed.
It is deployed across the entire US military, for good or bad. There is a lack of training on the sytem, especially of how it can go wrong. The models are not 100% accurate, which Cukor acknowledges. But the military commands treat its output as accurate. They want to believe the machine and blame bad data if it causes things to go wrong. Instead of it working along side humans as Cukor envisioned, it is positioned to shorten the kill chain to the point where the machine does the action.
The chapters where Ukraine was involved are very telling from the US military point of view. They thought the Ukrainian data systems were not secure, but they were more secure than most of the US military. The secure network the US military was using was so poorly designed, it had packets bouncing back and forth several times across the Atlantic before finding where they actually needed to be. Someone thought it was a good design at some point. The effort required to fix all of it was emense, especially when those involved didn’t go through proper channels. The Ukrainians, on the other hand, were building out their own drone targeting systems that didn’t need the fancy American hardware. The US military was still thinking big, while Ukraine was thinking fast and flexible. The US military still hasn’t understood the lessons of Ukraine in the future of warfare. It isn’t “AI everywhere” or “muscled warfighters.” Its being nimble and using cheap drones to dominate.
I do have a bone to pick with the author. She mentions the Winchester Mystery House, as Kristen Cukor brings up the mansion and its crazed builder. I think it is a good reflection of Kristen Cukor to ponder if she is a modern day Sarah Winchester. The author can’t get the right place in California. She says it is “...the curious 1880s Santa Cruz mansion...” No, it is in San Jose & Sarah Winchester kept building until she died in 1922. Please get your landmarks right, especially when one of the people interviewed is making a point of the house (it is also the title of Chapter 26).
I think the author did a wonderful job in illustrating the ins and outs of Project Maven. From the book and what I know of AI models, I don’t believe the folks in the military fully comprehend how such systems can go wrong, which is often. Current models will always product a large amount of wrong answers. It is built right into how the models act. But the military is embracing AI as really good, having it make targeting decisions that are instantly executed. At one point there is a warning about how technologies developed for overseas always come back to be used in the homeland. Maven or something like it could easily be retasked. Flock cameras are in the theme of Maven, tracking targets without regard to privacy. I believe the use of AI will cause more problems than it solves, making the kill chain so short a lot of innocent people are harmed.
June 5, 2026
Karen Manson’s book “Project Maven: A Marine Colonel (Drew Cukor) His Team, and the Dawn of AI Warfare” was released in 2026 by Lakeside Book Company. Drew is a retired Marine Colonel; and a former military intelligence officer who at the time of the book’s release was working for Chief Executive Jamie Dimond at J.P. Morgan Stanley Bank. His military leadership mission was to work with Marine Corp leadership as they implemented artificial intelligence’s rapidly evolving technology in the U.S. military.
Author Katrina Manson in 2017 was the U.S. foreign policy and defense correspondent for the Financial Times. This was the same year that Project Maven started in the Marine Corp. She watched Afghanistan and Iraq “embrace AI as a shortcut to sustaining military dominance.” The Department of Defense in her words became “devoted to AI and autonomy.” Five years later in 2022 she became a Bloomberg correspondent covering technology advances, chatbots, and AI agents. She writes about the ineptitude of military intelligence practices and according to Drew Cukor an over reliance on archaic “maneuver to contact strategies.” In 2016 the central corp team created a breakfast club to eat and also discuss electronic military strategies that included AI. Drew joined the club and helped direct the strategy discussions. The club met in a Pentagon room very close to the Secretary of Defense’s office. In the 21st century U.S. military perspectives were rapidly aligning with technologies like autonomous Mission Control and cloud computing power, with an increasing focus on military internet technologies. Amazon, Google, IBM, and Palantir were very involved in the emerging AI revolution. The Breakfast Club convener Greg Grant and Drew Cukor became close friends, and they werestrong advocates for using drone AI warfare.
“Maven” means “one who understands”; and AI has amazing abilities to understand inter technological connections. The commitment to implement military AI became the Project Maven initiative for US military leadership. In the summer 2018 ECS Federal joined the Maven initiative. They provided the Maven team with AI cloud, cybersecurity, and engineering to the military and many other Federal government entities including the postal services and Department of Homeland Security organizations. As the Maven Project took life the initiative had many new recruits who built many innovative and forward-looking information technology services. Maven quickly deployed to the Navy, Air Force, and Army. In 2018 Ai algorithms were in three different drone initiatives, 200+ Air Force MQ-9s, and bug-eyed sensors called “Gordon Stare.” AI special operations data security was also built. The last three chapters in Part 1 of the book feature math challenges, evolutionary events, and moral social outrage nonmilitary Americans had about AI. Google was contracted to develop a “pilot project” to identify the location and vulnerabilities of drones and other machine learning military projects. The project was for 18 months, and Google experienced an extreme backlash from their employment base who insisted Google not become involved “in the AI business of war.”
The second part of the book is titled “Fix” and has seven chapters that discuss AI military algorithms, submarine search locators, Ai arms race initiatives, AI arms dealer Alex Karp, and the AI initiatives of Plantir Corporation’s terrorist leader Abu Bakr al-Baghdadi’s activities. The second part also features the data pipelines in the U.S. military. The book describes in detail Maven’s ability in the Ukraine-Russia warfare to track Russian military vehicles and equipment including transporter erector launchers (TELs) The system documents in real-time data from drones, satellites, and radar to locate warfare objects.
Part three is titled “Finish” and Part four is titled “Feedback.” The “Finish” chapters discuss the Maven’s team search for spy satellites, U2 planes, and AI strategies to locate rocket launchers as well as Scarlet Dragon’s targeting enemy locations. This part of the book also discusses in detail the role Rear Admiral Frank “Trey” Whitworth had in developing and using AI. He was head of intelligence for the U.S. MilitaryJoint Staff. Part 3 ends with a discussion of Taliban kill chains, Ukraine’s use of AI, and OPLAN battle strategies for the U.S. algorithmic warriors defense of Europe.
Part four is titled Feedback and it has three “end of book” chapters about why machines “shouldn’t kill people”, AI intelligent robots, and super intelligent computers called Skynet. The Project Maven book has a wonderful Epilogue and an exceptional section of 363 notes on reference sources along with a page index. (P)
Author Katrina Manson in 2017 was the U.S. foreign policy and defense correspondent for the Financial Times. This was the same year that Project Maven started in the Marine Corp. She watched Afghanistan and Iraq “embrace AI as a shortcut to sustaining military dominance.” The Department of Defense in her words became “devoted to AI and autonomy.” Five years later in 2022 she became a Bloomberg correspondent covering technology advances, chatbots, and AI agents. She writes about the ineptitude of military intelligence practices and according to Drew Cukor an over reliance on archaic “maneuver to contact strategies.” In 2016 the central corp team created a breakfast club to eat and also discuss electronic military strategies that included AI. Drew joined the club and helped direct the strategy discussions. The club met in a Pentagon room very close to the Secretary of Defense’s office. In the 21st century U.S. military perspectives were rapidly aligning with technologies like autonomous Mission Control and cloud computing power, with an increasing focus on military internet technologies. Amazon, Google, IBM, and Palantir were very involved in the emerging AI revolution. The Breakfast Club convener Greg Grant and Drew Cukor became close friends, and they werestrong advocates for using drone AI warfare.
“Maven” means “one who understands”; and AI has amazing abilities to understand inter technological connections. The commitment to implement military AI became the Project Maven initiative for US military leadership. In the summer 2018 ECS Federal joined the Maven initiative. They provided the Maven team with AI cloud, cybersecurity, and engineering to the military and many other Federal government entities including the postal services and Department of Homeland Security organizations. As the Maven Project took life the initiative had many new recruits who built many innovative and forward-looking information technology services. Maven quickly deployed to the Navy, Air Force, and Army. In 2018 Ai algorithms were in three different drone initiatives, 200+ Air Force MQ-9s, and bug-eyed sensors called “Gordon Stare.” AI special operations data security was also built. The last three chapters in Part 1 of the book feature math challenges, evolutionary events, and moral social outrage nonmilitary Americans had about AI. Google was contracted to develop a “pilot project” to identify the location and vulnerabilities of drones and other machine learning military projects. The project was for 18 months, and Google experienced an extreme backlash from their employment base who insisted Google not become involved “in the AI business of war.”
The second part of the book is titled “Fix” and has seven chapters that discuss AI military algorithms, submarine search locators, Ai arms race initiatives, AI arms dealer Alex Karp, and the AI initiatives of Plantir Corporation’s terrorist leader Abu Bakr al-Baghdadi’s activities. The second part also features the data pipelines in the U.S. military. The book describes in detail Maven’s ability in the Ukraine-Russia warfare to track Russian military vehicles and equipment including transporter erector launchers (TELs) The system documents in real-time data from drones, satellites, and radar to locate warfare objects.
Part three is titled “Finish” and Part four is titled “Feedback.” The “Finish” chapters discuss the Maven’s team search for spy satellites, U2 planes, and AI strategies to locate rocket launchers as well as Scarlet Dragon’s targeting enemy locations. This part of the book also discusses in detail the role Rear Admiral Frank “Trey” Whitworth had in developing and using AI. He was head of intelligence for the U.S. MilitaryJoint Staff. Part 3 ends with a discussion of Taliban kill chains, Ukraine’s use of AI, and OPLAN battle strategies for the U.S. algorithmic warriors defense of Europe.
Part four is titled Feedback and it has three “end of book” chapters about why machines “shouldn’t kill people”, AI intelligent robots, and super intelligent computers called Skynet. The Project Maven book has a wonderful Epilogue and an exceptional section of 363 notes on reference sources along with a page index. (P)
June 5, 2026
Project Maven
A Marine Colonel, His Team, and the Dawn of AI Warfare
by Katrina Manson
W. W. Norton, 2026
Katrina Manson, an award-winning Bloomberg reporter on cyber and national security, has produced a worthwhile but ultimately disappointing account of how the Pentagon dragged itself into the age of AI warfare. The book is built around Drew Cukor, the unyielding Marine Corps colonel who, from a windowless Pentagon office in 2017, led the small team that put artificial intelligence into the American kill chain. The Maven Smart System Cukor built is now deployed in every branch of the US military, in space, and in submarines, and the story of how it got there ought to be a riveting one.
Manson had real access. She interviewed more than 200 insiders, and the texture of life inside Cukor's operation comes through. We see a workaholic colonel who held morning meetings standing up, had not checked his voicemail since 2005, and worked so hard he had no time for it anyway. We see the bureaucratic permafrost he had to crack: the jealous six-service Pentagon, the budget process where almost no money is available for genuinely new spending because everything is committed to continuing prior programs, the Navy meeting AI not just with inertia but with downright hostility. We see Googfle’s Chairman, Eric Schmidt, telling a four-star general that the Pentagon absolutely sucks at machine learning. We see Google employees revolting against military work while their company stayed cheerfully engaged with China. We see Special Operations Command, in contrast to the rest of the military, wanting AI on everything from the beginning.
There are real revelations. The early algorithms identified trees as people, rocks as buildings, and buildings as cars. Worse than useless, it was dangerous. By 2019 the same system was providing operational pictures in every active war zone. By the Ukraine war, AI could find thousands of targets within moments, and the bottleneck shifted from finding targets to humans being unable to keep up with what the machines were producing. Interesting fact: By May 2025, Ukraine was consuming 4,000 drones a day, the same number the Defense Department managed to purchase in a year.
These are the facts that ought to keep policymakers awake at night, and Manson lays them out.
The book also captures something important about Cukor himself. He was relentless, he succeeded, and along the way he made enough enemies that he never made general. He retired as a full colonel in October 2021. By the end, having spent years building data, Cukor refused to share it with the Joint AI Center, becoming exactly the sort of office bureaucrat protecting his own turf that he had always railed against. That irony is one of the more honest moments in the book.
So why three stars rather than four?
Because the book is at least 100 pages too long, and shows it. There is a great deal of repetition. Manson lingers on interpersonal squabbles within the Maven team in a way that pads the final third without adding insight. The editing is loose throughout, and the prose would have benefited from a firmer hand.
More irritating is Manson's persistent need to slip in zingers calculated to appeal to the anti-military left. The reporting itself is honest, but the asides keep tilting the picture. At the close she gives a respectful platform to UN Secretary-General António Guterres and the International Committee for Robot Arms Control on the proposition that machines should not kill people. Well, of course, but in my opinion people should not kill people either. That is the whole problem of war, and the moral seriousness of the question deserves more than a slogan. There are also small unforced errors and questionable claims throughout, the kind a competent editor catches before publication.
Read this book if you care about AI in warfare, about how the Pentagon actually works, or about how Silicon Valley and the military uneasily found each other. The subject is genuinely important, and Manson has more access than anyone else currently writing on it. Just be prepared to skim, and to read past the editorializing to get at the reporting underneath. A mediocre book on a vital topic.
A Marine Colonel, His Team, and the Dawn of AI Warfare
by Katrina Manson
W. W. Norton, 2026
Katrina Manson, an award-winning Bloomberg reporter on cyber and national security, has produced a worthwhile but ultimately disappointing account of how the Pentagon dragged itself into the age of AI warfare. The book is built around Drew Cukor, the unyielding Marine Corps colonel who, from a windowless Pentagon office in 2017, led the small team that put artificial intelligence into the American kill chain. The Maven Smart System Cukor built is now deployed in every branch of the US military, in space, and in submarines, and the story of how it got there ought to be a riveting one.
Manson had real access. She interviewed more than 200 insiders, and the texture of life inside Cukor's operation comes through. We see a workaholic colonel who held morning meetings standing up, had not checked his voicemail since 2005, and worked so hard he had no time for it anyway. We see the bureaucratic permafrost he had to crack: the jealous six-service Pentagon, the budget process where almost no money is available for genuinely new spending because everything is committed to continuing prior programs, the Navy meeting AI not just with inertia but with downright hostility. We see Googfle’s Chairman, Eric Schmidt, telling a four-star general that the Pentagon absolutely sucks at machine learning. We see Google employees revolting against military work while their company stayed cheerfully engaged with China. We see Special Operations Command, in contrast to the rest of the military, wanting AI on everything from the beginning.
There are real revelations. The early algorithms identified trees as people, rocks as buildings, and buildings as cars. Worse than useless, it was dangerous. By 2019 the same system was providing operational pictures in every active war zone. By the Ukraine war, AI could find thousands of targets within moments, and the bottleneck shifted from finding targets to humans being unable to keep up with what the machines were producing. Interesting fact: By May 2025, Ukraine was consuming 4,000 drones a day, the same number the Defense Department managed to purchase in a year.
These are the facts that ought to keep policymakers awake at night, and Manson lays them out.
The book also captures something important about Cukor himself. He was relentless, he succeeded, and along the way he made enough enemies that he never made general. He retired as a full colonel in October 2021. By the end, having spent years building data, Cukor refused to share it with the Joint AI Center, becoming exactly the sort of office bureaucrat protecting his own turf that he had always railed against. That irony is one of the more honest moments in the book.
So why three stars rather than four?
Because the book is at least 100 pages too long, and shows it. There is a great deal of repetition. Manson lingers on interpersonal squabbles within the Maven team in a way that pads the final third without adding insight. The editing is loose throughout, and the prose would have benefited from a firmer hand.
More irritating is Manson's persistent need to slip in zingers calculated to appeal to the anti-military left. The reporting itself is honest, but the asides keep tilting the picture. At the close she gives a respectful platform to UN Secretary-General António Guterres and the International Committee for Robot Arms Control on the proposition that machines should not kill people. Well, of course, but in my opinion people should not kill people either. That is the whole problem of war, and the moral seriousness of the question deserves more than a slogan. There are also small unforced errors and questionable claims throughout, the kind a competent editor catches before publication.
Read this book if you care about AI in warfare, about how the Pentagon actually works, or about how Silicon Valley and the military uneasily found each other. The subject is genuinely important, and Manson has more access than anyone else currently writing on it. Just be prepared to skim, and to read past the editorializing to get at the reporting underneath. A mediocre book on a vital topic.
August 1, 2026
Katrina Manson wants to make sure you know that writing a book about Project Maven was a challenging nut to crack, since very little about the program has been declassified (though some aspects of it have been trumpeted, sometimes quite loudly, by companies and agencies involved). There is a bigger challenge Manson faced, however. Too many people believe conspiratorial, scary legends about Maven, usually focusing on how the innocent Google, the first contractor, was replaced by the evil and treacherous Palantir. Although Palantir may well live up to its evil billing, it was never at the center of Maven. Manson tells us in a logical, sequential way how the program for AI-based targeting grew up in the Marine Corps, due largely to a true believer named Drew Cukor, and was only incidentally about the contractors that signed up for the ride of normalizing Maven across all major air-sea-sky-space platforms, in several different theaters of war.
The real scoop Manson tells us in incidental fashion, toward the end of the book, is that after Maven had been bounced around many joint intelligence and combat agencies, it ended up as a project of the nation's newest intelligence agency, the National Geospatial Intelligence Agency, or NGA. In fact, the majority of NGA's missions are now Maven-based. So if there's a 2020s center to Maven, following its Marine Corps years, it resides with NGA, not Palantir. Manson also tells us that the Office of Naval Research is the odd bird out, since it has always believed in highly-classified AI targeting projects based on the agency's proprietary in-house software. These programs, kept in deep black in the classified portions of the defense budget, include Goalkeeper, a broad-based program for smart naval munitions, and Whiplash, a more targeted program to put smart munitions on board jet skis. Yeah, you heard right. Manson also reveals that when the U.S. supplied Ukraine with Hi-MARS rockets, those weapons were targeted solely by U.S. forces, making the U.S. an active weapons-bearing participant in the Ukraine war since at least late 2022.
Manson tries to keep her reporter's eye as neutral as possible, even as she interviews people who are proud to say the real Maven goal is to kill as many people as possible. In the early days, Cukor tried to get Maven partners to talk about target acquisition, but it becomes more and more fashionable during the two Trump administrations to say the software developed for multiple platforms is an active-targeting search-and-destroy tool.
Manson does not hesitate to name Palantir's Alex Karp as one of the most evil people alive, but she makes the point that few within the Pentagon like Karp or Palantir either. In fact, many people at various agencies and partner-contractors tried to limit Palantir's involvement because the company is universally hated. She later shows that many in the targeting AI realm, such as Alex Wang and Parker Luckey, are little Karp clones, and as evil as their hero and his financier, Peter Thiel. But what do we do about the fact that the evil ones are utterly in the thrall of Pete Hegseth's Pentagon, where he rails about "woke AI" and wants robots that "kill, kill, and kill again" regardless of the consequences.
At the end of the book, Cukor comes to a partial resolution of morals after he leaves the military. He thinks that Maven was the right kind of project, but he says that the trends now in place in generative AI, agentic AI, and artificial general intelligence, are very ominous and seem incapable of being stopped. But not to worry. We have people like Sam Altman and Elon Musk in commercial A.I., and people like Karp, Wang, and Luckey in military A.I. In other words, we are most rightly and fully screwed.
The real scoop Manson tells us in incidental fashion, toward the end of the book, is that after Maven had been bounced around many joint intelligence and combat agencies, it ended up as a project of the nation's newest intelligence agency, the National Geospatial Intelligence Agency, or NGA. In fact, the majority of NGA's missions are now Maven-based. So if there's a 2020s center to Maven, following its Marine Corps years, it resides with NGA, not Palantir. Manson also tells us that the Office of Naval Research is the odd bird out, since it has always believed in highly-classified AI targeting projects based on the agency's proprietary in-house software. These programs, kept in deep black in the classified portions of the defense budget, include Goalkeeper, a broad-based program for smart naval munitions, and Whiplash, a more targeted program to put smart munitions on board jet skis. Yeah, you heard right. Manson also reveals that when the U.S. supplied Ukraine with Hi-MARS rockets, those weapons were targeted solely by U.S. forces, making the U.S. an active weapons-bearing participant in the Ukraine war since at least late 2022.
Manson tries to keep her reporter's eye as neutral as possible, even as she interviews people who are proud to say the real Maven goal is to kill as many people as possible. In the early days, Cukor tried to get Maven partners to talk about target acquisition, but it becomes more and more fashionable during the two Trump administrations to say the software developed for multiple platforms is an active-targeting search-and-destroy tool.
Manson does not hesitate to name Palantir's Alex Karp as one of the most evil people alive, but she makes the point that few within the Pentagon like Karp or Palantir either. In fact, many people at various agencies and partner-contractors tried to limit Palantir's involvement because the company is universally hated. She later shows that many in the targeting AI realm, such as Alex Wang and Parker Luckey, are little Karp clones, and as evil as their hero and his financier, Peter Thiel. But what do we do about the fact that the evil ones are utterly in the thrall of Pete Hegseth's Pentagon, where he rails about "woke AI" and wants robots that "kill, kill, and kill again" regardless of the consequences.
At the end of the book, Cukor comes to a partial resolution of morals after he leaves the military. He thinks that Maven was the right kind of project, but he says that the trends now in place in generative AI, agentic AI, and artificial general intelligence, are very ominous and seem incapable of being stopped. But not to worry. We have people like Sam Altman and Elon Musk in commercial A.I., and people like Karp, Wang, and Luckey in military A.I. In other words, we are most rightly and fully screwed.
August 16, 2026
What .... a ...... difficult book.
Regarding the information and revelations about the use of the advanced targeting technology especially in Eastern Europe (and thus confirming direct participation of NATO in the war - I will never understand these book reveals while officials constantly say no, no, we are just providing political support; Right ...... I mean why do you lie for f**k sake when every person with two eyes and something between them sees what happens) this is a gold mine. I still like how they mention HIMARS as operated by locals while book dedicated to HIMARS systems (fight against ISIS) explicitly says that without direct system link with targeting software that is proprietary and under control of US Army it just does not work. I mean all that talk about building integration software out of the blue..... right, directly from people who never worked on any software integration. Again ..... whatever, disappointing, it is completely ball-less and pu**y-footing but again that's West these days it seems.
Second thing is that book shows the only workable business model for LLMs - sifting through mass of data to provide targeting solutions in near real time. Not reasoning, not actual intelligence as such but quick analytical tools for automation of target identification and selection. What I especially liked is that book clearly shows that these types of automation existed for years (as can also be seen in any literature on sensor integration in last 30 years) but in 2010's and 2020's pairing of brute force approach and very capable hardware (which caused brute force approach) enabled rise of modern LLMs which are just same category of algorithms but running on a very fast hardware and if required very close to target area - data centers built by IT corporations that are spread all over the world.
Chapters on the way IT companies became part of US military industry using public infrastructure shows how shortsightedness when it comes to strategic decisions results in a double edged sword solutions [as current Middle East conflict shows]. This means that all satellite networks, cables and any data centers in vicinity are legit targets for anyone. Which again means we are maybe simple escalation away of complete breakup of world computer networks. Now that would be very interesting, imagine people returning to local data centers and local networks :D I mean that is just natural step from current state where normal non-AI computing is so much sidelined that it seems that calculators will be most advanced things for anything practical (phones being used solely for any Tok software).
Rest of the book is author talking with people, being given ominous hints over the shoulder and being told by people interviewed about juvenile behavior of project teams, weird social dynamics (that special type of psychopathic behavior lauded as engine of industry) and decisions made. Out of all this gibberish I only understood actions of military because they were asking for solutions that can be applied now-and-here.
Interesting book, but due to these chapters explaining inter-personal relations (that are slog, terrible, terrible slog - imagine reading about assholes that like terrorizing others to meet their goals and justifying it with requirement to achieve goals as soon as possible ..... for hundreds of pages ...... ) this was a very long read. Unusually long when I think about it.
But for information it provided, it is recommended book. Interesting things ahead, unfortunately in that proverbial meaning.
Regarding the information and revelations about the use of the advanced targeting technology especially in Eastern Europe (and thus confirming direct participation of NATO in the war - I will never understand these book reveals while officials constantly say no, no, we are just providing political support; Right ...... I mean why do you lie for f**k sake when every person with two eyes and something between them sees what happens) this is a gold mine. I still like how they mention HIMARS as operated by locals while book dedicated to HIMARS systems (fight against ISIS) explicitly says that without direct system link with targeting software that is proprietary and under control of US Army it just does not work. I mean all that talk about building integration software out of the blue..... right, directly from people who never worked on any software integration. Again ..... whatever, disappointing, it is completely ball-less and pu**y-footing but again that's West these days it seems.
Second thing is that book shows the only workable business model for LLMs - sifting through mass of data to provide targeting solutions in near real time. Not reasoning, not actual intelligence as such but quick analytical tools for automation of target identification and selection. What I especially liked is that book clearly shows that these types of automation existed for years (as can also be seen in any literature on sensor integration in last 30 years) but in 2010's and 2020's pairing of brute force approach and very capable hardware (which caused brute force approach) enabled rise of modern LLMs which are just same category of algorithms but running on a very fast hardware and if required very close to target area - data centers built by IT corporations that are spread all over the world.
Chapters on the way IT companies became part of US military industry using public infrastructure shows how shortsightedness when it comes to strategic decisions results in a double edged sword solutions [as current Middle East conflict shows]. This means that all satellite networks, cables and any data centers in vicinity are legit targets for anyone. Which again means we are maybe simple escalation away of complete breakup of world computer networks. Now that would be very interesting, imagine people returning to local data centers and local networks :D I mean that is just natural step from current state where normal non-AI computing is so much sidelined that it seems that calculators will be most advanced things for anything practical (phones being used solely for any Tok software).
Rest of the book is author talking with people, being given ominous hints over the shoulder and being told by people interviewed about juvenile behavior of project teams, weird social dynamics (that special type of psychopathic behavior lauded as engine of industry) and decisions made. Out of all this gibberish I only understood actions of military because they were asking for solutions that can be applied now-and-here.
Interesting book, but due to these chapters explaining inter-personal relations (that are slog, terrible, terrible slog - imagine reading about assholes that like terrorizing others to meet their goals and justifying it with requirement to achieve goals as soon as possible ..... for hundreds of pages ...... ) this was a very long read. Unusually long when I think about it.
But for information it provided, it is recommended book. Interesting things ahead, unfortunately in that proverbial meaning.
Review of advance copy received from Publisher
#ad much love for my finished copy @wwnorton
Project Maven
Available: March 24, 2026
I’m sure some of you have heard the rumblings of AI robot killers designed for war. Our future is coming at us straight from a sci-fi horror movie, only it’s not fiction. It’s here. And it is only growing in every branch of the military - here and abroad. It. Is. Coming. It. Is. Here.
This book tells the story about how we got here and where we are going. It’s a crucial read. An alarming read. It asks important questions and analyzes what future wars might look like.
Taken from hundreds of interviews, documents, and more Katrina Mandon brings a straight-forward, well researched, investigative book packed with information that may shock you. It goes beyond mass surveillance.
Those for it say it means less loss of life.
Those against say it could mean more mass deaths.
Project Maven is a timely book that’s being released at exactly the right time. As the Trump administration continues to pour money into AI advancement in the military, along with autonomy, and the UN pushes for a ban on killer robots, Manson gives clear and crucial information within the AI debate. It’s an informative read.
Manson gives us the facts while keeping her personal feelings out of it. Both sides are given space and are explored equally. It’s a fascinating read. But also kind of terrifying because we know they’re still working on this and it will only get worse - in every aspect of life.
It can be used as a tool and do good work. But what the founder had in mind when he first started his crusade for AI technology in warfare, isn’t where things are now headed. It’s out of his hands now anyways.
Who ever thought that Google, Meta, and other companies would ever have their hands in war. It’s kind of insane to think about.
Final Thoughts:
They’ve already admitted AI lies on purpose - so let’s use it for wars .. algorithms deciding who dies and who lives. Sounds great ……… GOD. HELP. US. ALL.
“this book investigates whether Al will improve accuracy and save lives or if a fundamentally unreliable black-box technology will unleash mistakes and atrocities at scale.”
Project Maven
Available: March 24, 2026
I’m sure some of you have heard the rumblings of AI robot killers designed for war. Our future is coming at us straight from a sci-fi horror movie, only it’s not fiction. It’s here. And it is only growing in every branch of the military - here and abroad. It. Is. Coming. It. Is. Here.
This book tells the story about how we got here and where we are going. It’s a crucial read. An alarming read. It asks important questions and analyzes what future wars might look like.
Taken from hundreds of interviews, documents, and more Katrina Mandon brings a straight-forward, well researched, investigative book packed with information that may shock you. It goes beyond mass surveillance.
Those for it say it means less loss of life.
Those against say it could mean more mass deaths.
Project Maven is a timely book that’s being released at exactly the right time. As the Trump administration continues to pour money into AI advancement in the military, along with autonomy, and the UN pushes for a ban on killer robots, Manson gives clear and crucial information within the AI debate. It’s an informative read.
Manson gives us the facts while keeping her personal feelings out of it. Both sides are given space and are explored equally. It’s a fascinating read. But also kind of terrifying because we know they’re still working on this and it will only get worse - in every aspect of life.
It can be used as a tool and do good work. But what the founder had in mind when he first started his crusade for AI technology in warfare, isn’t where things are now headed. It’s out of his hands now anyways.
Who ever thought that Google, Meta, and other companies would ever have their hands in war. It’s kind of insane to think about.
Final Thoughts:
They’ve already admitted AI lies on purpose - so let’s use it for wars .. algorithms deciding who dies and who lives. Sounds great ……… GOD. HELP. US. ALL.
“this book investigates whether Al will improve accuracy and save lives or if a fundamentally unreliable black-box technology will unleash mistakes and atrocities at scale.”
July 8, 2026
Independent of your personal views on the military, artificial intelligence, or their growing intersection, this book is a must read. Aren't you curious how a large portion of your earning are being used?
This is not a detailed technical analysis of the AI systems rather broad brush stokes on capabilities and reactions by those involved. In some respects it is a biography of the main character, Colonel Drew Cukor, USMC. It is also a view from both sides of the argument to use machines for assistance in warfare.
It’s tempting to dismiss early attempts at using AI in warfare as crude, error-prone, and unreliable. Yet those shortcomings must be weighed against the flawed human systems they aim to improve or replace. What stands out is how a small group of battle-tested Marines—adept at cutting through bureaucracy and institutional hubris—drove the project forward. Their determination not only delivered results but also seeded numerous companies through initial contracts.
I particularly appreciated the leaders’ philosophy of deploying early and often. Sending engineers directly into the field (whether metaphorical or literal) proved to be the fastest way to iterate and refine a complex system under real conditions.
As the story progresses, it raises fascinating questions about how the military and government will respond as private companies increasingly control the core capabilities that have traditionally defined military power.
The progression from AI as copilot to full autonomy—and ultimately to lethal decision-making—is undeniably unsettling. However, the book rightly challenges readers to compare these systems not against an impossible standard of perfection, but against today’s reality of human-controlled operations. A new system doesn’t need to be flawless; it only needs to be meaningfully better than what exists now.
My one minor criticism is the book’s heavy use of in-text hyperlinks to source documents. While thorough, they can make reading cumbersome, as it’s easy to accidentally jump into the extensive reference section while scrolling.
I am reminded that those few individuals who make a difference have the right mixture of vision, leadership and drive. A missing ingredient in this cocktail results in failure.
It is our politicians who set the tone of how to work within the Geneva Convention, choose your leaders wisely. Leadership is all about character.
This is not a detailed technical analysis of the AI systems rather broad brush stokes on capabilities and reactions by those involved. In some respects it is a biography of the main character, Colonel Drew Cukor, USMC. It is also a view from both sides of the argument to use machines for assistance in warfare.
It’s tempting to dismiss early attempts at using AI in warfare as crude, error-prone, and unreliable. Yet those shortcomings must be weighed against the flawed human systems they aim to improve or replace. What stands out is how a small group of battle-tested Marines—adept at cutting through bureaucracy and institutional hubris—drove the project forward. Their determination not only delivered results but also seeded numerous companies through initial contracts.
I particularly appreciated the leaders’ philosophy of deploying early and often. Sending engineers directly into the field (whether metaphorical or literal) proved to be the fastest way to iterate and refine a complex system under real conditions.
As the story progresses, it raises fascinating questions about how the military and government will respond as private companies increasingly control the core capabilities that have traditionally defined military power.
The progression from AI as copilot to full autonomy—and ultimately to lethal decision-making—is undeniably unsettling. However, the book rightly challenges readers to compare these systems not against an impossible standard of perfection, but against today’s reality of human-controlled operations. A new system doesn’t need to be flawless; it only needs to be meaningfully better than what exists now.
My one minor criticism is the book’s heavy use of in-text hyperlinks to source documents. While thorough, they can make reading cumbersome, as it’s easy to accidentally jump into the extensive reference section while scrolling.
I am reminded that those few individuals who make a difference have the right mixture of vision, leadership and drive. A missing ingredient in this cocktail results in failure.
It is our politicians who set the tone of how to work within the Geneva Convention, choose your leaders wisely. Leadership is all about character.
July 30, 2026
This is a very interesting book about the origins of AI targeting in the US military, though this is a precursor technology to truly automated warfare, it is an important one and the author spends a lot of time focused on this. The book is structured with the career arc and the personality of the leader of project maven, an aggressive and unusual combination of business drive and beurecratic savvy. The story of the unit overcoming obstacles both within the pentagon and without is quite interesting as is the exposition of how the technology was developed and deployed. The book is slightly marred by the biases of the author. She displays a justified concern for the dangers of the technology but she seems to blame the US for its use, and holds the country accountable for any and all errors in the technology. The fact that Maven occasionally makes errors is seen as a prohibitive factor in its development but the alternative, continued reliance on human efforts, which are even more error prone, is not really considered a negative. Additionally, much attention is given to the moral questions raised by the technology within the US, tech workers and activists seeking to ban the technology or to prevent their companies from working on it are portrayed as uniformly good. No consideration is really given to alternative scenarios where non-Democratic nations develop and deploy the technology faster than the democracies, or the fact that American weapons occasionally kill civilians in error whereas our opponents, the Iranians, the Chinese, Khamas, the Russians have made killing civilians, including their own, a key element in policy. Alas, in the modern publishing world, the progressive perspective has to be taken as baseline by more or less all reporters and the only way to get a story like this told at all is to have to wade through the biases of the author. So the book is certainly worth a read.
August 13, 2026
My career has revolved around the defense industrial complex. I've worked for many contractors and have spent some time working with programs and at locations mentioned in this book. I had some minor involvement with some of the efforts mentioned.
And this book just didn't address a lot of the interesting parts of what was going on.
Reading this as part of a reading group and at one point I indicated to the team that the book had a rating between 2 and 5 stars and that it was all over the place. And that's how I feel now that it's been finished. There were moments of brilliance and insight. And then there were moments when we heard more about hurt feelings and the transitioning of people off the team.
I think this book missed several opportunities to provide more insight into the definition of artificial intelligence (AI) in the military context and possibly any sort of quick description about what the algorithms were doing. There were brief but very interesting discussions about trying to bring speed and a more entrepreneurial mindset to the military, followed by more soap opera discussions. I felt the discussion of cross-purposes in the military was reasonably well handled and discussions of the increase in bandwidth and the lack of network preparation for modern war was fascinating. The security aspects of bringing in new and untried products, of sharing training data, etc. were all very interesting. But, in the end, it all felt rather shallow.
I appreciate the discussions about the ethics of new products in the military, but I felt that the concerns about AI were left a bit on their own outside of general concerns about warfare, though the last chapter and the epilogue did some to improve that.
Overall, I think this is an interesting book on an interesting topic, but I also felt it was a lost opportunity to really dig in and provide useful information rather than the overlong blog post it seemed to devolve into.
And this book just didn't address a lot of the interesting parts of what was going on.
Reading this as part of a reading group and at one point I indicated to the team that the book had a rating between 2 and 5 stars and that it was all over the place. And that's how I feel now that it's been finished. There were moments of brilliance and insight. And then there were moments when we heard more about hurt feelings and the transitioning of people off the team.
I think this book missed several opportunities to provide more insight into the definition of artificial intelligence (AI) in the military context and possibly any sort of quick description about what the algorithms were doing. There were brief but very interesting discussions about trying to bring speed and a more entrepreneurial mindset to the military, followed by more soap opera discussions. I felt the discussion of cross-purposes in the military was reasonably well handled and discussions of the increase in bandwidth and the lack of network preparation for modern war was fascinating. The security aspects of bringing in new and untried products, of sharing training data, etc. were all very interesting. But, in the end, it all felt rather shallow.
I appreciate the discussions about the ethics of new products in the military, but I felt that the concerns about AI were left a bit on their own outside of general concerns about warfare, though the last chapter and the epilogue did some to improve that.
Overall, I think this is an interesting book on an interesting topic, but I also felt it was a lost opportunity to really dig in and provide useful information rather than the overlong blog post it seemed to devolve into.
April 21, 2026
A very good journalistic report about how a diplomatic maverick colonel Drew Cukor single-handedly see through Project Maven, an AI warfare tooling, to its wide acceptance within the military, overcoming resistance from traditional military mindset, overcoming doubts due to the incompetence during the fledgling stage of the project. That sounds like a typical heroic story without mentioning some nuances revealed.
First, the motivational speech requires an arch enemy, fake it if must.
Second, the Maven AI tooling requires war field data that only real battle grounds can provide. Naturally, battle ground in Afghanistan became opportunities for Maven's AI training, innocent peasants, women and children in Afghanistan paid the price when AI detection made mistakes or when human operators made mistakes. When news about bombing civilians broke out repeatedly, what Maven project leaders cared about most is reputation, while shrug off the civilian damage as a must price to pay for the AI technology. To be fair, killer instinct is part of strong man nature through out human history, but with AI warfare endlessly amplifies the killing capability partly by technology and partly by separating killers from the killed, peace for plain earthlings look further remote from now on.
The use of Maven technology in Ukraine-Russia war sounds more justified because US is helping an invaded country defending itself. At the same time, Maven gained precious snow field battle ground data for training to sharpen itself. Enemies without the privilege to participate in wars are in disadvantage.
Reports like this inevitably echo the old philosophical and religious debate about good and evil. Must the good build a stronghold through evil means in order to deter evil, imagined or real? Who is evil? What is good?
First, the motivational speech requires an arch enemy, fake it if must.
Second, the Maven AI tooling requires war field data that only real battle grounds can provide. Naturally, battle ground in Afghanistan became opportunities for Maven's AI training, innocent peasants, women and children in Afghanistan paid the price when AI detection made mistakes or when human operators made mistakes. When news about bombing civilians broke out repeatedly, what Maven project leaders cared about most is reputation, while shrug off the civilian damage as a must price to pay for the AI technology. To be fair, killer instinct is part of strong man nature through out human history, but with AI warfare endlessly amplifies the killing capability partly by technology and partly by separating killers from the killed, peace for plain earthlings look further remote from now on.
The use of Maven technology in Ukraine-Russia war sounds more justified because US is helping an invaded country defending itself. At the same time, Maven gained precious snow field battle ground data for training to sharpen itself. Enemies without the privilege to participate in wars are in disadvantage.
Reports like this inevitably echo the old philosophical and religious debate about good and evil. Must the good build a stronghold through evil means in order to deter evil, imagined or real? Who is evil? What is good?
July 21, 2026
More like 2.5 stars
Project Maven is the code name for the Pentagon’s foray into AI-based weapons. It was begun in 2017 and shepherded into existence by a semi-obsessive in the Pentagon named Drew Cukor, who first got the idea while watching over and over a YouTube video of a computer defeating a world-class master at the game of Go. That’s how this book begins. It ends with a close analysis of AI’s use in the wars in Ukraine and the IDF’s brutal response to the attacks by Hamas in Gaza. Those are the two most interesting things in this book. In between, despite extensive research and careful analysis, is stultifyingly dull — lots of in-fighting in the Pentagon, among AI vendors and suppliers, between branches of the service. There is a reasonable amount of technical information about what the Maven system can do (basically, choose targets, find them and, someday, though apparently not yet, fire the weapons to dispense with those targets). There is some conversation about the ethics of using AI in war, and a bit about the measured approach under Biden and the more bonkers approach by Pete Hegseth who, among other things, has zero use for the Geneva Conventions. Great, let’s trust Pete Hegseth to make our decisions about what constitutes an ethical prosecution of war — a guy who’s only qualification to be Secretary of “War” is being a talking head on Fox “News” (oh, yeah, also displaying undying fealty to Donald Trump), but I digress. In any case, though what it has to say is important, there is an awful lot of boring minutiae in this book, whereas the core concepts would’ve made a terrific 7500 word piece in the New Yorker. Instead, I listened for over 13 hours and retain very little of the content. Sort of the textbook definition of a important book that’s a lousy listen.
Grade: C+
Project Maven is the code name for the Pentagon’s foray into AI-based weapons. It was begun in 2017 and shepherded into existence by a semi-obsessive in the Pentagon named Drew Cukor, who first got the idea while watching over and over a YouTube video of a computer defeating a world-class master at the game of Go. That’s how this book begins. It ends with a close analysis of AI’s use in the wars in Ukraine and the IDF’s brutal response to the attacks by Hamas in Gaza. Those are the two most interesting things in this book. In between, despite extensive research and careful analysis, is stultifyingly dull — lots of in-fighting in the Pentagon, among AI vendors and suppliers, between branches of the service. There is a reasonable amount of technical information about what the Maven system can do (basically, choose targets, find them and, someday, though apparently not yet, fire the weapons to dispense with those targets). There is some conversation about the ethics of using AI in war, and a bit about the measured approach under Biden and the more bonkers approach by Pete Hegseth who, among other things, has zero use for the Geneva Conventions. Great, let’s trust Pete Hegseth to make our decisions about what constitutes an ethical prosecution of war — a guy who’s only qualification to be Secretary of “War” is being a talking head on Fox “News” (oh, yeah, also displaying undying fealty to Donald Trump), but I digress. In any case, though what it has to say is important, there is an awful lot of boring minutiae in this book, whereas the core concepts would’ve made a terrific 7500 word piece in the New Yorker. Instead, I listened for over 13 hours and retain very little of the content. Sort of the textbook definition of a important book that’s a lousy listen.
Grade: C+
August 3, 2026
This was a worthwhile book for me, as it covers the intersection of two topics for which I'm very interested: (1) Department of Defense/War and (2) AI. My interest was especially piqued, though, when I found out that my brother was quoted (multiple times) by the author. The book was certainly interesting and seemed to parallel my own military career and duties at times. It did seem a bit longer than necessary, with a level of detail that at times was difficult to follow, but I appreciated the in-depth review of the topic. Drew Cukor, who was the main protagonist for this book, was a major character in another book I recently finished reading, Mobilize, by Shyam Sankar. Sankar portrays Cukor as one of several defense "mavericks," who has led the way in advancing innovation within the DoD. Manson's book goes much deeper into Cukor, painting a compelling portrait of both his personal and professional lives. Of note, I found the portion on the "Jevons paradox" particularly compelling. This law says that with any increases to resource efficiency caused by technology, consumption or use of that resource will also increase. Given this law, the increased speed and efficiency of targeting will not necessarily mean more time for commanders to make decisions, but instead, it will actually mean a massive increase in targeting and resultant killing. The pursuit of AI for defense is often billed as saving soldiers' lives, but the Jevons paradox could indicate the more likely result of even more death and destruction. AI is already changing our world, and as such, it will inevitably continue to change the Department of War, how our nation fights its wars, and how we maintain our national security well into the future.
August 11, 2026
Holy moly. I knew this wasn’t going to be a beach read, but this book was both hard to put down, scary, and at the same time, a little inspiring. One man, compelled by his beliefs, riding his team to work insane hours fighting against the pentagon and various bureaucrats, all to bring to life some seriously scary shit that will help the US win wars. Fully automated target acquisition and now, “shooting” weapons at adversaries, potentially without a human in the loop. This is the first time I have read in depth about the crossover between Silicon Valley and a lot of the AI targeting we know is happening, but don’t maybe pay attention to. The author does a great job of positioning the many characters in this book and the broad range of viewpoints on AI and its role in war, but ultimately gives quite a bit of airtime (deservedly so) to American soldiers who were/are hellbent on building automated targeting capabilities, and did so over the last 10 years. Was this a terrible idea in terms of ethics? Perhaps. Would other countries have been able to develop semi or mostly automated targeting as quickly as the US did? Probably not. Is it a strong enough deterrent to prevent china invading Taiwan or another conflict in the next decade? Who knows. The author alludes to but doesn’t go too deep into doom and gloom about the future, but does talk a lot about terminator. All I know is that quite depressingly, I’m glad the closest thing we have to Skynet is (for now) controlled by the pentagon.
June 19, 2026
AI is in the news daily, the Pope has a commercial telling of its danger, and data centers are the new industry used to process and store its accumulated information. Is it a WAR machine or an educational tool? Needless to say, the controversy continues. This book mainly deals with the technology development along with its military use.
In a short time, drone technology has become a military weapon, and when tied with AI, the speed of information communication, it now gives an actual view of any place in the world, including human observation. This raises a moral question: "Does the machine make the decision to kill?"
I have an "agricultural drone" in my garage, an observation drone with an excellent camera on site, and another drone on an especially designed trailer with all the equipment to spray crops. These tools for the farm industry can be quickly designed to be a weapon. I have one grandson who flies "a flying brick" in races, and another grandson who uses his to take photos. A new technology that is causing protests around the world. The use determines the love or hate opinion.
Katrina Manson follows a system developer through the ever-changing uses and the need for regulation. By the time this book hit the shelves, much had changed. War provides a testing field. Just last week, we observed a "boat drone" rescue two pilots in a very dangerous area. AI allowed it all to be observed on TV.
There will be much added to the facts printed in this read, and very soon.
In a short time, drone technology has become a military weapon, and when tied with AI, the speed of information communication, it now gives an actual view of any place in the world, including human observation. This raises a moral question: "Does the machine make the decision to kill?"
I have an "agricultural drone" in my garage, an observation drone with an excellent camera on site, and another drone on an especially designed trailer with all the equipment to spray crops. These tools for the farm industry can be quickly designed to be a weapon. I have one grandson who flies "a flying brick" in races, and another grandson who uses his to take photos. A new technology that is causing protests around the world. The use determines the love or hate opinion.
Katrina Manson follows a system developer through the ever-changing uses and the need for regulation. By the time this book hit the shelves, much had changed. War provides a testing field. Just last week, we observed a "boat drone" rescue two pilots in a very dangerous area. AI allowed it all to be observed on TV.
There will be much added to the facts printed in this read, and very soon.
Displaying 1 - 30 of 112 reviews