The Why Man (Anand Vallamsetla) (@thewhyman) header
The Why Man (Anand Vallamsetla) (@thewhyman) profile picture

AI Fund | Applied AI Engineering Leader | Evaluation-First LLM & Agent Platforms | Reliability & Guardrail Design | ex-Google | UC Berkeley

Los Altos Hills, CA
Joined February 2011
Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)Photo by The Why Man (Anand Vallamsetla) (@thewhyman)
Ethan Mollick describes two ways of working with AI. The centaur divides the labor: you do this, the machine does that. The cyborg intertwines, back and forth, continuously. There is a third mode he does not name, and it is the one quietly winning.
1
1
52
Six weeks ago Dario Amodei said slow down. Last week Anthropic filed to go public at a reported $2T valuation. I don't think the moat is real — I think the ask for a cage is the moat. Full argument: thewhyman.blog/p/they-asked-…
1
26
Abdication does not feel like failure. It feels like leverage. Then the model goes down, or is confidently wrong in a way that matters, and you discover the thing that was true the whole time: you did not get faster. You got hollow.
1
7
Delegate the search, the draft, the first ten wrong answers. Never delegate judgment. This is the follow-up to the piece on learning how to learn — full thread two links down.
3
22
Ethan Mollick names two ways of working with AI: the centaur, who divides the labor on a clean line, and the cyborg, who intertwines with it. There is a third mode nobody named. I will give it one: abdication. The practitioner is the ghost — present in name, absent in substance.
1
14
Your brain: about twenty watts. El Capitan: about thirty megawatts. Call it a million to one. And that still does not count what your twenty watts is also doing — holding a body, a lifetime of context, and the ninety-per-cent signal.
1
6
How many times has your gut been right? Not the times you wish you had listened — the times you actually had the feeling and checked later. Most people I ask land north of ninety per cent. That is not a hunch. That is a working classifier. 🧵
1
17
2,400 years ago Socrates called himself a midwife — barren of wisdom, full of questions. A language model is the exact inverse: full of answers, has lived nothing. The method for training judgment is older than the problem.
1
3
A gut feeling is not the absence of reasoning. It is reasoning you never got the transcript of. The birds stop. Insect noise drops out. None of it individually matters. Collectively it is a pattern, and something matches it against a very old library: wrong.
1
2
If a model gave me ninety per cent on a hard classification task, I would ship it. So either everyone is flattering themselves about their own track record, or something real is happening and we have a bad name for it.
1
6