AI Tech Founder Used Fake Paper Bank Statements to Steal $60M From Investors

For all the scary warnings about dangerous “deep fake” videos, and machine made ventriloquists, you have to wonder about the very low tech fraud rattling American markets.

One allegation in the indictment stands out, in particular, for its brashness. Allegedly, when an ON investor wanted to see the company’s bank statement directly from the bank, they arranged a visit with Beckman. According to the indictment, Lau had a fake statement planted and placed in an envelope at the bank; when Beckman and the investor showed up, they received the envelope and its sheet purporting a $13 million account balance. In reality, the indictment says, that account had just $25.93 and showed Beckman had recently wired $320,000 to a member of his family.

The indictment notes that if convicted, they will have to give up their… wait for it… Tesla.

Is Stargate a $500 Billion Digital Maginot Line? Strategic Missteps in History of War

The history of warfare teaches us one lesson repeatedly: fixed defenses fail catastrophically if they can’t address mobility of opponents. If you thought “build a wall” ended in corrupt failure (it did) here we are in 2025 watching the same characters prepare to pour $500 billion into the “gate” – a massive fixed AI infrastructure project that manages to ignore virtually every lesson from both military history and information security.

The Digital Fortress Fallacy

As someone who has spent many decades analyzing technological risk while immersed in global information warfare, I can state unequivocally: Stargate represents the single largest strategic miscalculation in the history of warfare technology. It’s not just a mistake – it’s a mistake of such magnitude that future military historians will likely use it as a canonical example of strategic myopia.

The parallels with historical blunders are not just striking – they’re identical in their fundamental misunderstanding of how advantage is gained and lost in technological warfare:

  • The Maginot Line cost France 3 billion francs and the illusion of security
  • British battleship programs consumed vast resources just as naval warfare was being revolutionized by carriers
  • Soviet heavy armor investments in Afghanistan were rendered obsolete by mobile insurgents with Stinger missiles
  • Libyan armored divisions were humiliated by Chadian forces in Toyota pickups

Going backwards in time has been a hallmark of Trump, so I must ask whether suddenly driving to make these same old mistakes, at a scale that dwarfs all previous examples combined, is really what Stargate represents.

This certainly would be worse than the “build a wall” rhetoric of fraud that self-defeatingly redirected American security staff away from critical airports and seaports to stand in the empty desert wondering where all their money went.

Bannon never faced federal charges because he was pardoned by then-President Trump on Trump’s final night in office during his first term. The pardon only applied to the federal case and did not preclude state charges. Bannon had told donors to the We Build the Wall campaign that their money would 100% go toward building a wall along the U.S. southern border, while prosecutors say some of the $15 million in donations was secretly funneled to himself and the campaign’s president, Brian Kolfage.

That scam wall is bascially set to repeat as a gate with 33x the scale of losses (from $15M to $500B).

Realities of Open Source Warfare

The recent DeepSeek developments out of China have abruptly exposed the obvious and fundamental flaws in Stargate’s dubious conception. Timing of the announcement is notable. A reported ability to achieve competitive AI performance at a fraction of the cost isn’t an anomaly – it’s the expected outcome of open source warfare principles that have governed technological conflict for centuries. And it means Trump already is creating catastrophic weakness.

When analysts like Gavin Baker try to dismiss DeepSeek’s $6M achievement by pointing to “prior research costs,” they’re making the same deadly mistake military planners make when they focus on R&D budgets instead of deployment effectiveness. Baker argues this cost “excludes prior research” and required “hundreds of millions in prior research” – as if that somehow diminishes the achievement. The Nazis claimed a lowly graduate student alone invented their jet engine when everyone knows German spies stole it from Cambridge, England. Today’s analysts are making the same mistake, pretending DeepSeek’s achievements don’t count because they built on existing research.

Imagine sitting in London as the V-1 falls, arguing “but Hitler didn’t account for prior research costs”. This is exactly equivalent to claiming the Mujahideen’s effectiveness against Soviet helicopters should be discounted because they didn’t invent the shoulder-mounted rocket launcher or account properly for R&D budgets.

In warfare – whether physical or digital – what matters is effective deployment, derivation and adaptation, not who paid for the original art.

American intelligence funded extremist Islamic radicalism developing into violence, disseminating leaflets like this of a giant mujahid with “God is great” written on his jacket, shown defending Islam and God from Soviet assault. The text in the top right says “Shield of God’s Religion,” implying faith of the mujahideen will protect him from bullets. Source: FP.

Historical Precedent is Asymmetric Victory

Let’s be explicitly clear about what history teaches us:

  • Mission 101 (Ethiopia, 1940): at most 20,000 irregular troops utterly routed hundreds of thousands (~300K) Italian fascist forces through clever mobility and tactical adaptation
  • The Toyota War (Chad, 1987): Pickup trucks and rocket launchers decimated Soviet-supplied armor columns
  • Ukrainian Drone and Bike Warfare (2022-present): Consumer drones and motorbikes with adaptable tactics render Russian billion-dollar air, land and water defense systems largely irrelevant

Each of these examples demonstrates how agile forces using adapted technology consistently defeat massive fixed investments. And that’s before we account for the scale of corruption fraud expected from Trump. The Stargate project ignores the whole history of warfare lessons at a scale that beggars belief.

Don’t get me started on knights in armor sinking into the mud of 1415 Agincourt or how Napoleon’s Navy repeatedly was a sitting duck of disasterous miscalculations while Nelson literally ran circles of fire around them.

Perhaps France’s infamously aggressive dictator should be referenced today more often as Mr. Napoleon Blownapart? The gargantuan French warship L’Orient explodes at 10PM. Source: National Maritime Museum, Greenwich, London

Digital War Veteran Reality Check

From an information security perspective, Stargate represents everything we know doesn’t work in cyber defense. Concentrating resources in massive overpriced fixed infrastructure of our grandparents’ fears creates:

  1. Single points of failure
  2. High-value targets
  3. Reduced adaptability
  4. Resource drain from more effective defensive measures

It’s as if we’re building the world’s most expensive castle while our opponents are already fielding artillery. It’s like celebrating WWI veteran and politician Maginot building a concrete line from his past while opponents simply launch radio directed fast attack planes and tanks of WWII. How hard is Putin laughing at Stargate right now?

Strategic Cost of Watergate Stargate

Stargate isn’t just replicating the strategic errors of fixed fortifications – it’s potentially worse because it represents the privatization of core national security compute infrastructure. We’ve seen how this story ends before with military contractors: the taxpayer gets the bill, the private entity gets the profit, and the actual security capabilities often end up compromised. Imagine if the NSA had outsourced its core computing infrastructure to a private contractor in the 1960s – that’s effectively what we’re contemplating here, but at a far larger scale and with far higher stakes. The lesson from France in the crucial years just before WWII is that every dollar poured into a digital Maginot Line is a dollar not spent on:

  • Distributed AI development capabilities
  • Asymmetric technological advantages
  • Adaptive defense systems
  • Actual technological innovation

Ask a Historian or Lose

History’s judgment of fixed fortification strategies is universally harsh. Maginot’s Line, Hitler’s Atlantic Wall, the Great Wall of China, Hadrian’s Wall, Israel’s Bar Lev Line, Israel’s Gaza Wall, South Korea’s DMZ… and countless other massive fixed defenses share one common feature: they have failed spectacularly when overconfident and unable to address asymmetric mobility. Stargate appears poised to join this list, but at a scale that would make even the most wasteful military planners of the past blush.

The truly tragic aspect of the Stargate project is that we know better in the same way that tying our shoe laces is better than tripping over them. Once you know, you know.

American special forces have demonstrated daily how distributed, adaptive approaches work. Our cyber warfare units understand the superiority of mobile defense. Our best military historians and technologists have documented these lessons repeatedly.

Yet here we are, like Teapot Dome never happened (another case of private interests corrupting national resources)…preparing to build the most expensive fixed fortification in human history. Not just concrete and steel this time, but centralized chips and software. The lessons of history could not be clearer: this way lies disaster.

Do I know whether Stargate will fail at stated objectives? Did ChoicePoint get breached? When we privatized and centralized critical data processing infrastructure without diligence or regulation, it led to catastrophic security failures. History gives us that answer with crystal clarity. The real question is how much damage this strategic misallocation of resources will do to American technological competitiveness before reality forces a course correction.

Palantir sued the U.S. Army to Force itself in and then promised as a monopoly on intelligence it would find terrorists. Instead it created them and destroyed any chance of peace and stability (a darling of Wall Street bulls known as “self licking ISIS-cream cone”). Stargate could unleash an even bigger stock pumping bullsh*t avalanche destroying society as we know it.

Past is prologue. Look at my record since 1995: history is the best predictor. When faced with Nazi Germany’s overwhelming conventional superiority, many “realists” urged Churchill to negotiate from a position of weakness. They saw only the massive fixed infrastructure of German power – the tanks, the planes, the divisions. But Churchill understood something deeper about asymmetric warfare that applies perfectly to today’s AI arms race:

A former assistant private secretary to Churchill, Crawshay-Williams had written a letter cravenly pleading with the prime minister to make terms with Hitler — ‘I’m all for winning this war if it can be done … But it does seem to me, and, I know, to others, that “if and when” an informed view of the situation shows that we’ve really not got a practical chance of actual ultimate victory, no question of prestige should stand in the way of our using our nuisance value while we have one to get the best peace terms possible. Otherwise, after losing many lives and much money, we shall merely find ourselves in the position of France — or worse. I hope this doesn’t sound defeatist; I’m not that. Only realist’. Churchill’s response is brief and brutal: ‘I am ashamed of you for writing such a letter. I return it to you — to burn & forget’. Source: Christie’s Auction

Churchill’s brutal response – “I am ashamed of you” – came from understanding that apparent power imbalances can be overcome through unconventional approaches. He knew that Britain’s real strength lay not in matching German industrial might tank-for-tank, but in mobility, adaptation, and unconventional warfare. Today’s architects of Stargate are making the same mistake as Churchill’s doubters – assuming that massive fixed infrastructure (digital rather than industrial) is the path to security.

Stargate at first blush sounds like the worst boondoggle, even worse than the fragile Cybertruck, gifting future historians the ultimate example of how not to approach technological warfare in the information age.

A little rain in 2021 destroyed a brand new high-cost Trump wall, foreshadowing the fraud known as a Tesla Cybertruck. Source: Gizmodo

Unfortunately, that huge sucking sound seems to be $500 billion of taxpayer money being flushed, along with America’s technological leadership position. The people who couldn’t build a wall without it falling down, while they stuffed its budget into their pockets, are back with an even bigger fraud.


As if to prove the main point here deeper, OpenAI just announced ChatGPT Gov – another massive centralization of government AI infrastructure into fixed, high-value targets. They’re proudly announcing 18 million messages from government agencies flowing through their system, as if concentrating sensitive government communications in private hands while deregulating safety was something to celebrate rather than a huge strategic vulnerability.

The Great DeepSeek AI Migration: From Locked Data Centers to the Freedom of Your Own Pocket

How Distributed Computing is Restoring Personal Digital Liberty

Is that AI in your pocket or are you just happy to see the shift away from Big Tech monopolistic elites raining misery into everyone’s life?

The computing landscape is undergoing a tectonic AI shift that everyone should look at closely. While headlines focus on Nvidia’s meteoric rise and its seeming stranglehold on AI market players, the perfect storm of technological breakthroughs is arriving as usual to fundamentally reshape how and where computation happens. This transformation could democratize AI in ways thought possible since the 1950s.

Nvidia Ain’t Like Intel Inside

At first glance, Nvidia’s dominance in processing chips shifted for AI might seem reminiscent of Intel’s iron grip on the PC market. But the comparison falls apart with a few observations. Intel owned its manufacturing facilities, creating true vertical integration. Nvidia doesn’t, instead needing TSMC for manufacturing. Any competitor can access the same manufacturing quality. This crucial difference opens the door for innovation from every quarter, as perhaps it should always have been seen.

Unmistakable Forces of Decentralization

  1. Efficiency Breakthroughs
    Recent developments by companies like DeepSeek, a company with just a few hundred employees, have demonstrated that AI models can run with 45x greater efficiency than current approaches. This isn’t just an incremental improvement – it’s a paradigm shift that could make AI compute feasible on much smaller, cheaper hardware. When you can do with one chip what previously required 45, the economics of AI deployment change dramatically.
  2. Software Abstraction
    Nvidia’s real exit moat (reverse moat, hotel California, also known as the silicon valley prison) has been its CUDA software ecosystem, not its hardware. But new high-level frameworks like MLX, Triton, and JAX are abstracting away hardware dependencies and restoring the market. This mirrors how programming evolved from assembly language to C++ fortunately making specific hardware less relevant. Soon, AI workloads may run efficiently on virtually any capable hardware.
  3. Custom Silicon Proliferation
    Unlike the PC era, where manufacturers were content to use standard chips, today’s tech giants are investing billions in custom AI silicon. This is about optimizing for specific workloads that unlock diverse processing needs, reducing dependence on centralized providers, more than it is about saving cost.

Put Your AI Where You Are

This convergence of forces points to a very near future realizing the concepts talked about since the heady Hadoop days of the 2010s, where AI compute moves closer to where data lives, meaning where you actually live – on your phone, your laptop, your home server, your family cloud, your work infrastructure. Consider the implications:

  • Privacy: Personal data never needs a boundary you define, such as your devices or your trusted partners
  • Cost: Dramatic reductions in compute costs make personal AI infrastructure feasible today
  • Latency: Local processing eliminates network delays, disruption and capture
  • Customization: AI models increasingly fine-tune to individual needs for higher integrity (important!) without compromising privacy

What This Means for Leaders

First, infrastructure planning requires at minimum a hybrid approach, combining edge computing with traditional data centers. We’ve known this for years with regard to key management, and it’s more true now than ever. A move entirely into cloud is like selling your home and land to live on a camel that wanders the desert in search of water it will be allowed to drink. Pastoralists built marvels of civilization nomads could only ever dream about. Second, data strategy must recognize our world is better when processing happens closer to collection points. That’s the revolution in philosophy known as enlightenment. I think, therefore I’m not shipping all my data to some weird politically toxic wizard of Oz for them to do it for me. Third, investment priorities need to consider how efficiency improvements will reshape AI infrastructure needs like a motorcycle with a shoulder fired rocket versus a Russian tank. Finally, privacy design must consider AI compute will happily grow at the edge. Castles stopped being built for some simple economic reasons, as Magna Carta tells us amidst language about power, and farmers were actually far better off for it.

The next few years will likely see an acceleration of human rights trends I’ve laid out here (assuming anti-human OpenAI Stargate nonsense doesn’t try to erase society). Cerebras, led by Andrew Feldman, and Groq, founded by Jonathan Ross (formerly of Google’s TPU team), are already shipping hardware that challenges traditional architectures. Open-source innovations from DeepSeek made headline news with efficient AI computation accessible to all because that’s what everyone wants. The real question is how quickly AI compute will decentralize and be distributed.

The message is clear from inside the data centers: the future of AI isn’t in big and central anymore than you are about to buy a mainframe to run your concentration camps. Smarter, more efficient compute that lives wherever you and your data does is the righteous and right path. Those who plan for this transition now will be best positioned to capitalize on the natural democratization that AI is finally maturing into.

Key Insights

The centralized data center model for AI is dated and unnecessarily inefficient, like a teenager refusing to learn how to feed and bathe themselves. New hardware and software innovations are enabling dramatic efficiency improvements by distribution of power. The future of AI compute is enabling more private, more personal, architectures close to data source owners. Business strategies must prepare for this shift in compute paradigm.

This obviously isn’t just about technology – it’s about democratizing processing of data in a way that preserves privacy, restores integrity, reduces costs, and puts power into the hands of individuals and businesses of all sizes. The era of having to send all your data to massive, centralized data centers may be coming to an end sooner than the dangerous monopolists want you to believe. Can you work from home? Standing in an assembly line being flogged to speed up seems so last century. Can you shop from home? Waiting for hours in the cold to get a loaf of bread isn’t anyone’s best use of time. Do you prefer owning your own home and choosing among a dozen local cafes and restaurants to a huge centrally planned barracks with cafeteria mess hall that has only gruel and hard tack to eat? See the immediate future yet? It’s in our past.

Note: as much as I was tempted above to trot out the mostly strawman argument of evil feudalism, I must admit that such a uniform system was largely constructed during the French Revolution as propaganda to characterize and criticize the ancien régime.

Its initial effects in discourse were the horrible collapse into violence, opening the door to national capture by a demented “Emperor” who destroyed reason. History actually is a much more complex historical reality with diverse local arrangements and power relationships, which means invoking dangerous polarizing disinformation narratives of feudalism isn’t a great idea unless we are ready to deal with another ruthless psychopath stepping into the political breach. That’s why…

  • I use pastoralists vs nomads to illustrate the value of sustainable local infrastructure.
  • Home ownership vs wandering camels shows the risks of complete cloud dependence.
  • Local cafes vs mess halls invokes common experiences of how decentralization enables choice that nutures quality.
  • Work-from-home vs assembly lines hopefully connects the importance of data processing location to human dignity, for everyone.