Stories by dnw

Using LLM in the shebang line of a script

When life gives you lemons, write better error messages

Implementation Details of Codex /Goal

Finding Zero Days with any model?

Polymarket weather bet manipulated with a hairdryer

RLMs are the new reasoning models

Hermes Agent by Nous Research

Our World in Data

Scaling Managed Agents: Decoupling the brain from the hands

Emotion concepts and their function in a large language model

Who should you subject to your vibe-coding?

Benchmark for measuring code erosion under iterative specification refinement

Claude connects the dots on sleep apnea diagnosis

Taste for Makers [February 2002]

Pike: To Exit or Not to Exit

IronCurtain: A Personal AI Assistant Built Secure from the Ground

Semi-formal reasoning helps agents reason about code without executing the code

LLMs work best when the user defines their acceptance criteria first

OpenAI resets spending expectations. Compute target is around $600B by 2030

Redefining the Software Engineering Profession for AI