K Health proves AI diagnoses better than doctors
K Health, an Avishai investment, has proven AI diagnoses better than most doctors, though medical research remains a poor use case for LLMs.
Snapshot · The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
20VC: Wix's Founder on What Wall St Gets Wrong About AI and Wix | Will Base44 Win the Vibe Coding Wars | The Truth About the Economics of Vibe-Coding | The Buyback Disaster: Lessons Learned with Avishai Abrahami
Explore episode
Jul 13, 2026
Where this was said
Should a 22-Year-Old Still Learn to Code in the Age of AI?
At 40:03 · chapter starts 37:00
Harry poses the question many parents and students are asking: in the age of AI, should a 22-year-old still learn to code? Avishai's answer is more measured than the AI-optimists. LLMs are not on the verge of replacing white-collar workers wholesale — they make too many silly mistakes and aren't genuinely good at reasoning from scratch. They excel at reframing and synthesizing existing data, not generating original insight. [1] — Avishai Abrahami "LLMs won't replace white-collar jobs anytime soon — they keep making silly mistakes and aren't good at reasoning. But if there are two or t…" 37:50 Harry adds a structural point: changing UK university curriculum takes three years, meaning computer science students today are learning material that predates ChatGPT. The real risk, Avishai says, isn't the current generation of models — it's the next two or three algorithmic jumps. If those happen, the equation changes dramatically. But right now, the fear is ahead of the reality.
UK university curriculum takes 3 years to change
UK university curriculum takes 3 years to update, leaving computer science graduates learning pre-ChatGPT material.
Should a 22-Year-Old Still Learn to Code?
LLMs won't replace white-collar jobs anytime soon — they keep making silly mistakes and aren't good at reasoning. But if there are two or three more algorithmic leaps, everyone becomes a bit like monkeys. The honest answer: be less afraid, but not complacent.
AI in Healthcare: Diagnosis Yes, Research No
K Health proves AI can diagnose better than most GPs. But for medical research, LLMs are dangerous — they weight a Reddit post the same as a Nature paper. AI in medicine works perfectly when diagnosing common scenarios, fails completely for original research.
Similar snapshots
High morning cortisol = cheat code
Making the first hour of your day high-cortisol and the last hour low-cortisol is described as the single most impactful daily habit for mood, focus, alertness, stress resilience, and sleep quality.
Sleep vs Ritalin attention effects equal
A Washington University in St. Louis study found that a great night's sleep produces attention-improvement effects effectively equal to Ritalin.
4–6 hours post-exercise focus boost
In the 4 to 6 hours after exercise, dopamine and norepinephrine are elevated, making you more alert and better able to focus — unless you've exercised to exhaustion.
Physiological sigh discovered 1930s
The physiological sigh — two big inhales followed by a full exhale — was discovered by physiologists in the 1930s and is the fastest way to reduce autonomic arousal.
10 deliberate exhales raise HRV significantly
Doing just 10 deliberate exhales throughout the day can significantly elevate heart rate variability by the evening, by repeatedly activating the vagal sinoatrial node pathway.
Nicotine performance dose: 3–6mg, max 2x/week
For nicotine to have a genuine cognitive-enhancing effect rather than merely rescuing withdrawal deficit, Huberman recommends using it only once or twice per week at 3–6 mg.
Vaping = higher total nicotine than smoking
Because vaping can be done anywhere and is more palatable, people consume far more nicotine via vaping than they ever did smoking, potentially making vaping more dangerous in aggregate despite being safer per puff.
BPC-157: 2 injections cleared shoulder pain
Huberman reported that two injections of approximately 100–200 micrograms of BPC-157 eliminated persistent shoulder pain that had resisted months of other interventions.
Pinealon: 2.5–3 hrs REM in 6 hrs sleep
Huberman reported that on nights he used the peptide pinealon, he achieved 2.5 to 3 hours of REM sleep inside 6 hours — an unusually high proportion normally only found in the final hours of longer sleep.
