Why do birds suddenly appear everytime you are near confirmation bias

We are in a time of technical deflation. That must change how we manage tech debt. https://t.co/Oojkl4XOH3

I was talking about XKCD 1425 with @simonw and created the impossible app without using any words. https://t.co/cBkMYA7ohB

If you're working on a startup, my book is 64% off today! 4.6 stars on amazon with 156 reviews, and 4.2 on Goodreads. Perhaps it could be helpful?

@danshapiro My name is Jason, a self taught programmer and teacher. I enjoyed playing your game Robot Turtles with my siblings as a kid, and have been programming for almost a decade now.

Somos Novios - It's Impossible ..Not to love you.....await...patiently......my love......

@danshapiro

We tested one of the most common prompting techniques: giving the AI a persona to make it more accurate We found that telling the AI "you are a great physicist" doesn't make it significantly more accurate at answering physics questions, nor does "you are a lawyer" make it worse.

We tested the classic "You are an expert" technique to see if it would make the models smarter. It did not. They just don't get smarter by telling them that they're smart. (You can, however, make them dumber...) https://t.co/GcnSwpJngJ

My friend @harper is giving coding agents was to collaborate, and they're performing better. I think "social tokens" are a genuinely new thing, as LLMs get value from each others' outputs to solve problems more efficiently.

What animal lives longest per cubic centimeter? The black garden ant. AI FTW

Tremendous discussion with @goodpplsventure on surviving the early days of startup life & scaling:

"You'll always need to look at the code the LLM wrote" will age about as well as "You'll always need to look at the code the compiler wrote"

Yesterday I accidentally ported @steveyegge's "Beads" to Windows. Modern AI is amazing. And Beads now supports Windows. https://t.co/l4ZUyafQ7i

Told Codex CLI to install @Steve_Yegge's Beads. Failed to notice that there was no Windows build. When I checked in an hour later, it had... ported it to Windows. https://t.co/2EIOtVIoac

Anthropic's new skills are an open version of GPTs with an auto-picker.

Some of the most interesting experimentation in AI are people trying to chart a way forward by developing new approaches to doing work based on the strengths and weaknesses of the models we already have. Example: @danshapiro on slot machine deployment:

The folks at the cutting edge of AI are doing something new - Slot Machine Development.

ChatGPT Agent can drive Sora. Try to use this power for good and not for evil

Great piece by @KyleOrl on our research - classic psychology findings work with LLMs too.

More coverage of the surprising findings in our research. TL;DR - the same persuasive techniques that work on people, work on AI. https://t.co/zEt6gKzKRc (with my coauthors @emollick , @angeladuckw, @LennartMeincke, Lilach Mollick, and @RobertCialdini)