Sandeep Swadia opens his latest episode, "This Skill Makes You Dangerous In The AI Era", with a case study most people think they already understand: Theranos. The blood-testing startup raised $700 million from marquee investors and hit a peak valuation of $9 billion (3:03) before its technology was exposed as fraudulent. Swadia's point is not that Elizabeth Holmes was uniquely deceptive. It's that the investors, board members, and journalists who vouched for her failed to ask the most basic questions.
Swadia describes the episode directly: Theranos was, in his words, "a $9 billion failure of critical thinking, an expensive lesson in bad judgment"(4:26). The framing sets up his central argument — that as AI handles more analytical and executional work, the human ability to scrutinize claims, stress-test assumptions, and resist social pressure becomes the scarcer and more valuable skill.
The Theranos Framework: What Nobody Asked
Swadia walks through what made Theranos so durable as a fraud: a compelling founder narrative, high-status validators, and a technology claim that was genuinely difficult to disprove without access to the lab. The investors who wrote checks — some of the most sophisticated capital allocators in the world — did not independently verify whether the core product worked. They relied on social proof and the credibility of other investors already in the deal.
That pattern, Swadia argues, is not unique to Theranos. He points to Enron, FTX, and WeWork as companies where the same dynamic played out: consensus formed around a story before the underlying facts were examined. The question he poses as the corrective is direct:
“What needs to be true for this to be real?”
That single question, Swadia says, is the entry point to critical thinking. Applied to Theranos, it would have forced anyone evaluating the company to specify exactly what physical and chemical processes would need to function for a finger-prick blood test to return accurate results across hundreds of biomarkers — and then verify whether those processes were actually in place.
Why AI Makes This Skill More Urgent, Not Less
Swadia's argument about AI is specific. Tools like ChatGPT, Claude, and Gemini are fluent, confident, and fast. They can generate persuasive arguments for almost any position. Swadia's concern is that this makes it easier than ever to produce content that sounds authoritative without being accurate — and that users who cannot evaluate the output critically will be worse off for having the tool. He has addressed the AI fluency gap previously. In a video released three weeks prior, he discussed how systems thinking — the ability to identify patterns before taking action — is the one skill that AI struggles to replicate.
The practical framework Swadia offers in this video centers on a few concrete habits. He is explicit that the goal is not contrarianism — disagreeing for its own sake — but structured skepticism applied before a decision is made.
- 1Ask "What needs to be true for this to be real?" before accepting any significant claim.
- 2Identify the assumptions underneath the consensus, not just the conclusion itself.
- 3Seek out the person in the room who disagrees — especially when everyone else agrees.
- 4Separate the number of people who believe something from the quality of the evidence behind it.
- 5Check the claim against a falsification test: what would prove it wrong, and has anyone checked?
On the consensus point, Swadia is direct: "When everyone around you agrees, go find that person who disagrees"(12:52). He follows that with a related observation — "A thousand people believing in something doesn't automatically make it true" — which he uses to reframe the Theranos story. The investors who passed on the deal, or who raised doubts internally, were not contrarians. They were doing the baseline work the others skipped.
“When everyone around you agrees, go find that person who disagrees.”
Where This Fits in Swadia's Broader Output
Swadia has built a consistent body of work around cognitive skills and AI literacy. While this is the first time he has covered Theranos directly, though he has mentioned Elizabeth Holmes briefly earlier when he discussed how people are wasting their time and and mentioned her as a reference point for sunk costs. In the current video Swadia uses the $9 billion figure not as a cautionary tale about fraud specifically, but as a measurable cost attached to the absence of a particular thinking skill — which is a different and more actionable framing.
The video also connects to his recent piece on self-education, "How To Become Dangerously Self-Educated (with AI)", where Swadia argued that reading without a purpose produces consumption rather than understanding. The critical thinking framework here is the logical extension: consuming information without evaluating it is not just inefficient, it can be actively harmful when the information is wrong and the stakes are high.
Swadia does not offer a resolution to the harder problem he raises: how to apply structured skepticism quickly enough to be useful in real decisions, where time pressure and social dynamics work against deliberate questioning. That gap — between knowing the framework and deploying it under pressure — is where Swadia's argument currently stops.
