Tweets
2025-01-26T23:07:02.754812 | 🔗
An llms.txt for Modal https://gist.github.com/davidzqhuang/4a00758b42db0e3f57b50c7782c3e396
2025-01-26T22:47:18.912704 | 🔗
Im sorry but operator is still currently stupid, im sure it will get better though. comparatively 100x more bullish on this than video gen (although video gen techniques will find HUGE use everywhere else but video gen, esp AI for science) https://operator.chatgpt.com/v/6796b9513cac819185d2932c355a350c
2025-01-26T22:44:29.485614 | 🔗
Fairly bullish on Operator/Computer Use, but also... wtf? no wonder they are losing money on pro users
2025-01-18T00:03:57.133892 | 🔗
what people are missing about "exploitative" AI agents is that they are competing for your attention.
2025-01-13T15:16:04.997420 | 🔗
Been using o1-pro for almost a month now (will cancel). Marginally more helpful than o1 for summarizing/writing/editing. Good at macro-consistency which is helpful for editing large files. Still use o1 a lot for the speed, also o1-pro still could not fix bugs in a lexer I've been working on that I had to fix manually.
2025-01-12T23:38:34.667976 | 🔗
Random thought I thought of an interesting casino game perhaps, what about craps but over a series of round where each rounds starts with each person having a set amount of money and going for a set time. At each round you earn points according to the amount of money at the end… here’s the kicker the mapping of money to points might be nonlinear and changes every round so you have to adapt your strategy to maximize your expected number of points not dollars
2025-01-12T23:34:17.951235 | 🔗
Craps except over multiple tricks where each trick has a utility curve to fit
2024-12-28T15:28:59.070584 | 🔗
There should be hotkeys for model switching on llm chat providers
2024-12-26T15:40:48.477244 | 🔗
Most tasks for "Intelligence" correspond to organizations that have enormous organizing capacity.
