Nate: The AI That Was Just People

Derek: Welcome to Vaporware. I'm Derek. And Grant, buddy, I need you to be emotionally prepared for today. Grant: That sentence alone, I'm already recognizing the pattern. What are we actually looking at here? Derek: So, an AI shopping startup, $51 million raised, one-tap checkout, magic. Grant: Ooh, I love magic. I also love understanding how the trick works before I write the check. Derek: The magic was several hundred people in the Philippines manually typing in your order. Grant: Wait, wait, wait. That was the AI? Several hundred people in Manila? Manila manually processing check out? Derek: According to the Justice Department, effectively zero percent automation. Speaker 3: Wow. Derek: That was their number. Grant: Zero. Not five percent rounding error-zero actual automation. That's not a margin miss-that's total fiction. Derek: Prosecutors are calling it zero. Fortune covered the charges and the DOJ put it in the indictment. The actual automation rate was quote effectively zero percent. Grant: So the product worked flawlessly, just not with any actual technology in it. The gap between the pitch and the mechanism is the entire business. Derek: Right, right, right. Today we're digging into Nate, the app, the founder, Albert Saniger. And a fraud that ran for years before anyone outside the company figured it out. Grant: And this is not just one bad actor extracting value; this is a repeatable structure. I've watched enough trading floors to know when you're seeing a pattern, not a story. Derek: Oh! you're going to love this part. Presto Automation, EvenUp—TechCrunch flagged the same playbook: Philippines-based human labor dressed up as AI, investors none the wiser. Grant: So we've got the founding myth—the labor infrastructure hiding behind the buzzword—the mechanics of how they actually pulled it off. Derek: The fundraising pitch with fabricated automation rates the SEC says hit ninety-three to ninety-seven percent. Grant: Ninety-seven claimed, zero actual. That's not a forecast miss, that's a settlement gap between marketed and delivered. Derek: And then the collapse, the criminal charges, and a guy who currently lives in Spain. In Spain, Grant: Xavier Derek: and hasn't been served yet. Grant: Of course he does. Founders always extract before the narrative collapses-that's the only part of the operation that actually works. Derek: Okay, from the very beginning, let's get into it. Okay, so picture this: it's twenty nineteen, WeWork just imploded, blockchain is officially a punch line, and every VC is desperately looking for the next big thing. Grant: That's the tale. When a pitch needs a new buzzword instead of new mechanics, you're holding a position with no floor. I've watched enough pitches collapse when you ask the obvious question, what's actually different from last time. Derek: Right. And into that room walks Albert Saniger with a pitch. Pitch so clean it almost hurts; one app, any store on the internet, you hit buy once and AI handles the whole checkout; no account, no forms, no friction. Just done. Grant: Okay, I want that. That's a real gap, but I need to see settlement mechanics-how does capital move from buyer to merchant without human fingerprints somewhere in the chain? Derek: That's the thing, Grant. Everyone wanted it. According to TechCrunch, He founded One-Tap in twenty eighteen, with machine learning, neural networks, proprietary AI, ten thousand transactions a day, no human involvement. Grant: Ten thousand a day-suspiciously precise; in trading, when a counterparty leads with scale claims without unpacking failure rates first, that's when you stop listening to the pitch and start reading the fine print. Derek: Extremely specific. Suspiciously specific in retrospect. Grant: So what did the demo actually show? Because that's where you see mechanics or theater. Derek: Fortune reported that the DOJ alleged engineers were on standby during investor presentations, manually completing test purchases behind the scenes while the demo looked totally effortless on screen. Grant: Wait, so the demo only worked because engineers were literally in the back room completing transactions in real time? That's not a proof of concept. That's... That's stage magic with a hidden operator. Derek: Allegedly, yeah. Tap the button, Grant: Wow. Derek: somewhere off screen a person scrambles to complete your order, and the investors see flawless checkout. Magic. Grant: Nothing says proprietary neuro network like a sweating engineer manually processing your test order while you think you're watching automation. I've seen that play before. Derek: And to be fair, Coatue led the seed round in twenty twenty, Forerunner Ventures came in. These are not unsophisticated people. Grant: So what's the official claim, the actual pitch they're taking to investors? Derek: Sure. According to Holland and Knight's breakdown of the indictment, Saniger marketed it as working "like magic" using AI and neural networks. He allegedly claimed automation rates above ninety percent. Grant: Ninety percent automation rate-that's the number that justifies a Series A, if it's real. Derek: The DOJ says the actual rate was effectively zero percent. Grant: Zero percent actual! The spread between claimed and actual is the entire fraud. That's not rounding error, that's intentional misrepresentation. Derek: According to Fortune, the technology he bought from a third party never achieved the ability to consistently complete e-commerce purchases. Grant: So the whole pitch was buzzword timing and narrative design. (nine) Betting that investors wanted the story badly enough to skip the verification question. Derek: And nodding post-WeWork, pre-pandemic when AI-powered replaced blockchain-enabled and pitch decks, and Saniger walked in with the perfect story. Grant: Every single time is the same structure: demo succeeds because failure gets handled off screen by a human. That's not product validation; that's misdirection. Derek: Every single time. Grant: Which is the fundamental question nobody asks: What happens Speaker 4: when there's a failure? Grant: What happens when you remove the person in the back room, the whole operation evaporates and somehow nobody ran that test. Derek: So here's the thing behind the curtain: while Saniger was pitching near-perfect automation, Fortune reported the actual rate was effectively zero percent. The third-party AI he bought never managed to consistently complete a single purchase. Grant: Zero. Not three percent. Not rounding error territory. Zero. The gap between claimed automation and actual automation is your entire business model evaporating. Derek: Zero. So someone had to actually do it, and that's... Someone was hundreds of contractors in a Philippine call center, internally called, wait for it, purchasing assistants. Grant: Purchasing assistants, that's the euphemism; in trading we'd call them order fulfillment desk—same work, different label; the machinery of deception is always in the vocabulary first. Derek: That's the job; and here's the part that shows this wasn't just wishful thinking; it was active concealment. Saniger allegedly restricted access to the company's own internal automation dashboard, and told employees to treat those numbers as a trade secret. Grant: So his own engineers, the people who could actually see the code and know it doesn't work, couldn't access the dashboard showing it doesn't work. That's information architecture as a control mechanism. Derek: Most of them had no idea; according to TechCrunch, they were also told to scrub Nate from Speaker 4: the database. Derek: Eight from their social media profiles. Grant: Wait, wait, wait. So the people actually executing the fake automation couldn't even acknowledge they work there. That's operational security against your own workforce. That's not a start of running hot. That's organized concealment. Derek: Exactly. You can't accidentally out an AI startup if nobody knows you work there. Grant: Totally normal company policy. Scrub yourself from the internet so nobody can trace you back to a job that How the dozen officially exist, the structure was airtight from the inside. Derek: And then get this: October, twenty twenty one, a typhoon hits the Philippines. Grant: Oh no, that's the moment the system breaks; a typhoon doesn't care about your cover story. Derek: The call centre goes down, backlog of orders starts piling up, so Saniger directed a NAIT employee to spin up a second operation, this time in Romania! Grant: He built redundant call-center infrastructure? Multi-point Continent Backup to protect an illusion. That's not a startup anymore, that's an operation. Derek: To protect the illusion-and, per the indictment, investor transactions were specifically prioritized during the disruption so the people writing checks would never notice a slowdown. Grant: That's not an accident. That's a system someone designed. Multi-point failure protection. He architected it deliberately. Derek: Right, right, this was infrastructure, redundant multi-continent infrastructure, for humans pretending to be AI Grant: He accidentally built a genuinely sophisticated back office operation, just not the one he was pitching. The irony is the real AI here. Derek: Right! World-class fulfillment. Just not the kind he told investors about. And speaking of investors, with this machinery running full tilt, Saniger went back for the biggest fund raise yet. Grant: Uh-oh, and now he's got infrastructure, cash burn, investor expectations and zero actual technology. So obviously time to raise more money. Derek: A thirty-eight million dollar Series A, and he had some very specific numbers ready for the pitch. So picture June, Twenty Twenty-One: Senator walks into that Series A pitch with a story of Ninety-three to Ninety-seven percent automation. Grant: Wow. Derek: According to the SEC, those numbers were his own and totally fabricated. Grant: Ninety-three to Ninety-seven claimed, Zero actual. I've watched enough claim versus reality gaps in trading rooms to recognize that spread is always intentional, never optimistic. Derek: Zero. And the investors writing the checks had no way to verify. defy any of it because that "dashboard," the one he called a trade secret, was locked, so Coatue, Forerunner, Renegade Partners, and none of them could just say, "Hey, open the dashboard." That's basic settlement verification. Any trader asks that question, that's not sophisticated due diligence, that's just asking. That's the information asymmetry play. The workers were silenced on social media, the demos were staged, the one system that showed real automation data was off off limits to outside eyes-you can't audit what you can't access. Grant: Hmm. I mean, to be fair, June two thousand twenty one was a scorching market, and when a founder claims proprietary trade secret, most VCs defer, but that's also the moment you get blindsided: the strongest claims always hide the weakest data underneath. Derek: Exactly-and June twenty twenty one was a very hot market. Renegade Partners led the round, Forerunner and Coatue came in. Then, according to Crowdfund Insider, that thirty eight million dollar Series A brought Nate's total raise to fifty one million. Grant: Fifty one million total, and the founder's selling his own stake into that bid. Yeah, Derek: and here is where it gets genuinely ugly: the SEC alleges that while those checks were being written, Saniger sold approximately three million dollars of his own Nate shares. Grant: Wait, during the round? While new capital was wiring in, the founder cashed out roughly three million of his own stake? That's insider selling-I've watched that pattern precede collapse. Derek: DURING THE ROUND So the founder is cashing out roughly three million dollars while his investors are putting in their largest bet yet on a product he allegedly knew didn't work. Grant: That is textbook misaligned incentives-the founder has a downside hedge, everyone else is holding the bag. Derek: It really is. And then just to cap it off, holiday twenty twenty one hits, transaction volume spikes, so Saniger reportedly tells engineers to build Speaker 4: more servers. Derek: Build basic bots to handle the load. Grant: Wait, what, bots! He told investors Nate specifically didn't use bots because they were dumb and easy to detect, then quietly built them anyway. Derek: He did say exactly that, and then he quietly built them anyway. Grant: So the founder extracted his capital on the way out. What happened to the fifty-one million that came in the door? Derek: The answer comes next, and it is not pretty. So June, nineteen twenty-two, one article in the Information drops questioning whether Nate's AI is actually real—and that's it, that's the whole company. Grant: Wait, seriously? One story unravels two years of institutional due diligence—that's how fragile the structure was? I've seen enough leverage positions blow up on a single data point to recognize that kind of fragility. Derek: One story. According to Fortune's coverage of the indictment, Nate had a f At a funding round pending at that point, the article runs, and the round dies on the table. Grant: Two years of Coatue in Forerunner; armies of analysts, partner level due diligence, and one reporter asked the basic verification question in six weeks. That's not a gap; that's a failure mode. Derek: That's the thing, right? A reporter did in a few weeks what the VC firms couldn't do in two years. Grant: Smaller research budget. Fewer consultants; but somehow they actually ran the numbers. They asked the question that's literally taught in training. Derek: Rating 101: verify the claim against the actual settlement data. Grant: So no fresh capital means no payroll, and Nate's payroll was—um—hundreds of humans in Manila and Bucharest. Derek: Which is, you know, hundreds of payroll hit count across two geographies. That's a ten fifteen million dollar annual burn at minimum. When that stops, things break fast. Grant: Not cheap at all. TechCrunch reported the DOJ's indictment puts it plainly: Finally, Nate ran out of money and was forced to sell its assets in January, twenty twenty three. Derek: January, two thousand twenty three, from article to liquidity crisis in six months, that's the speed of narrative collapse converted to operational collapse; once the pitch breaks, the cash stops. Grant: Six months-and the dissolution method is kind of poetic-they filed a California Assignment for the Benefit of Creditors. Derek: Which means asset liquidation. Creditors get whatever's left, and equity holders get nothing. Equity holders get wiped-that's the structure of a total loss. Everyone's losses are someone else's order in the liquidation queue. Grant: It means you hand your assets to a third party, they sell everything off, and creditors get whatever crumbs are left. Investors got nothing. Prosecutors call the losses "near total." Derek: Near total on more than forty million dollars raised. I've seen enough positions get marked to zero to understand that's not an abstraction, that's someone's capital just disappearing. Grant: Yeah, and the workers in Manila and Bucharest, the people who actually ran the app every single day, they just lost their jobs. No severance headline, no press release, nothing. Derek: The contractors who actually made the operation function, day in and day out, they did the work. Bore the operational risk—and there are footnotes in every story about the fraud. Grant: Nodding, two years of smart money, locked dashboards and staged demos, and in the end a single article cracked the whole thing open! Derek: You know what's wild? Company dead by January, twenty twenty three, but the DOJ indictment waits until April, twenty twenty five! The real reckoning always has a delayed second act. Grant: In fact, that's when they move from civil to criminal. With a knowing tone, that part of the story, that kicks off in April 2025, and it's a whole different kind of reckoning. So, April 9th, 2025. Two years after the company dissolved, the DOJ and SEC finally show up. Derek: Parallel actions, same day. That's coordinated enforcement. Both agencies filing simultaneously means they're sending a signal, not just pursuing a case. Grant: According to Fortune and the SEC's own filings, Saniger faces one count of securities fraud and one count of wire fraud. Each one carries up to twenty years. Derek: That's forty years of exposure stacked; twenty on securities fraud; twenty on wire fraud. The DOJ is building overlapping leverage. Grant: The DOJ is also seeking criminal forfeiture of the proceeds, so it's not just prison time they want the money back too! Derek: And he pleaded not guilty, so we're headed to trial, assuming he stays within reach long enough for a trial to happen. Grant: December 2025, not guilty, released on a $250,000 bond, and here's the detail I cannot get over, he's allowed to reside in Spain. Derek: The SEC filed a civil complaint against a guy who's actively residing in Barcelona. Jurisdiction arbitrage on a federal enforcement action? That's a bold play. Grant: And he literally cannot be served, according to court records. The SEC had to request an extension just to attempt service abroad. As of mid-2025, no responsive pleadings filed. The case is basically sitting there. Derek: The guy built a company that couldn't solve checkout automation, but he automated himself right out of American legal reach. The irony is perfect. Grant: I'm writing that down. Derek: Okay, but why the two-year lag between company collapse and indictment? January 2023, dissolution, so investigation would... was clearly running, why does April 2025 become the filing date? That's not delay. That's a strategic hold waiting for discovery to lock down before unsealing. Grant: The investigation clearly predates the current administration. These things take time. Grand juries, document review. The filing date isn't when they started looking. Derek: Right, right. Grand jury empanelment, discovery mechanics, witness interviews. The machinery moves slow, but it moves. It's methodically—by the time charges drop they're built to stick. Grant: But here's what's interesting about the timing: in January, twenty twenty-five, the SEC settled a similar AI-washing case against Presto Automation, a drive-through ordering company that also used humans abroad to process orders it called AI-powered: no fine, just a cease and desist. Derek: Seriously? Identical scheme, same investor class got burned and— And Presto, walks out with zero fine, just to cease and desist letter? Grant: Zero dollars. And then literally three months later, Saniger gets criminally indicted. Legal analysts to DLA Piper flag this as a deliberate escalation signal from both agencies. Derek: So the escalation is deliberate. Presto gets the warning. Nate gets criminal charges. Two different enforcement signals for one playbook. Grant: That's one read; the other read is that criminal charges require a higher bar—intent, concealment, the staged demos, the locked dashboard. Nate had all of it on record. Derek: Quietly—and Saniger extracted himself to Barcelona while the indictment was still pending—he turned jurisdiction into a feature, not a bug. Grant: Case dormant; service unresolved. The indictment just exists. Here's the thing, though: Saniger isn't really the story. The story is that this keeps happening. Derek: With a slight lean, same structural playbook, different buzzword era; different founder, different capital pool, same fraud mechanics underneath, the patterns the constant. Grant: Same playbook, different buzzword era, and that pattern is what we need to talk about. So the question Saniger's case actually raises is bigger than one guy cooking a pitch deck. It's a pattern. I've watched enough trading floors to recognize when you're seeing a structure, not a story. Derek: Not even close. Presto Automation, right? Drive-thru AI for a Carl's Jr., Hardee's. They claim 95% fully automated. SEC filing shows 70% needed human hands. And yeah, also Philippines. Same labor arbitrage, different industry vertical. Grant: Also in the Philippines, at what point do Derek: Do we just call Manila the world's largest AI data center? Grant: And then there's EvenUp, billion-dollar legal-tech unicorn, AI-generated personal injury demand letters. Business Insider later reported heavily reliant on humans doing the actual work. Same structure, different application. Derek: So fast food, shopping, legal documents? Grant: Same playbook, different industry. Same gap between a pitch and a P&L. Derek: Raise on the AI label, buy time with human labor, lock down internal visibility so nobody can compare the pitch to the reality. Grant: And here's what sticks with me as a trader who's watched this cycle before. These aren't flukes. We've covered different versions on the show. Blockchain era, ML-powered, neural networks. The buzzword changes every 18 months. The underlying mechanics never do. Claim automation, deploy humans, lock the dashboard, raise capital. Derek: The Blockchain era: everyone funded the word, not the product, and now? New label; same operation underneath. Grant: Right! Derek: And when you zoom out, you see the geography too—Philippines-based call centers showing up as the hidden backbone behind multiple AI- you see the geography too-Philippines based call centers showing up as the hidden backbone behind multiple "A. AI-powered products. That's not a coincidence, that's a supply chain. Grant: Structural pattern—and the failure mode is always identical—investors, boards, auditors—they all funded the narrative, not the mechanics. Nobody asked for the dashboard; nobody ran the settlement math, Derek: which is what makes this hard to close out neatly, you know? Saniger's case, the Presto settlement, EvenUp scrutiny—the question that keeps coming back is whether AI-powered has become so loosely defined that smart people with diligence checklists are still just funding a label. Grant: The label raised a billion dollars. The actual product is zero percent automation. That's the real valuation gap, Derek: Dry automation rate effectively zero percent. Grant: which is the most honest number in the entire story, zero. Not because it's a rounding error, because that's what was actually there. Derek: Yeah, it really is. OK, so that was a lot: a shopping app, Fifty-one million dollars, automation rate zero. Grant: Zero; not approximately zero; not rounding error territory. Fortune reported the Justice Department's own language: "Effectively zero per cent." In trading, zero means zero-no rounding, no narrative adjustment. Derek: That's the thing, though: the Dashboard. Saniger allegedly locked his own team out of the real numbers. numbers, and called it a trade secret, his own team. Grant: Very normal CEO behavior-lock your own engineers out of the real numbers, call it a trade secret, extract capital during the fund raise. Textbook operation. Derek: Totally standard stuff. But here's the takeaway. The NAIT story isn't just about one guy lying. Grant: And Derek: It's about a funding environment that rewarded the story over the substance. Grant: now the DOJ and SEC apparently decided the Presto warning shot didn't land. NAIT got criminal charges instead. That's escalation. Watch this space. Enforcement just signaled it's not treating AI-washing as a soft landing anymore. Derek: If this episode hit right, tell a friend who still trusts every pitch deck they read. Subscribe wherever you listen. Drop us a review. It genuinely helps. Grant: Thanks for riding this one out with us. Saniger's case is still dormant in Barcelona, but the pattern he ran is still executing in a dozen other companies. That's the real story. Derek: We'll see you next time on Vaporware.