Join 📚 Josh Beckman's Highlights
A batch of the best highlights from what Josh's read, .
In s1, when the LLM tries to stop thinking with `"</think>"`, they force it to keep going by replacing it with `"Wait"`. It’ll then begin to second guess and double check it’s answer. They do this to trim or extend thinking time (trimming is just abruptly inserting `"</think>"`).
It’s really dumb, I love it.
S1: The $6 R1 Competitor?
Tim Kellogg
The dark irony here is that corporatized communication is what made communicating with each other not fun in the first place. And now, there are corporate-owned A.I. tools that are offering to help us no longer communicate directly at all. And competition from various corporate entities will ensure they keep getting incrementally better, while also, of course, making things more enshitted. Which means A.I. platforms aren’t just enshitting the internet, they’re helping us enshit ourselves.
Rinse, Wash, Like and Subscribe, Repeat
Garbage Day
The number one prompt engineering tip for all LLMs continues to be “use examples”—here’s Anthropic showing Claude an example of how to use its thinking and function calls together.
Highlights From the Claude 4 System Prompt
Simon Willison
...catch up on these, and many more highlights
Readwise