My post initially complained about the lack of reasoning traces in the API, but it turns out I was wrong about that! You can get back reasoning summaries with "reasoning": {"summary": "auto"} – I've updated that section of my post to describe that here: https://
simonwillison.net/2025/Aug/7/gpt
-5/#thinking-traces-in-the-api
…
@simonw
-
API Reasoning Traces: Getting Summaries with GPT-5
By
–
-
Reasoning Summaries Now Available in API Updates
By
–
Turns out I was wrong about that! You can get reasoning summaries back after all, I updated that section:
-
GPT-5 ChatGPT API access and model configuration differences
By
–
That "chooses how much to think" bit is for GPT-5 running in ChatGPT, the API gives you direct access to the model and settings that you want
-
Insider Perspective on AI Model Testing Under NDA
By
–
Yeah I had to hold my tongue when people were sharing pelican images from those models because I recognized them from my NDA-covered test runs!
-
GPT-5 Preview: Core Features, Pricing and System Card Insights
By
–
I've had preview access to GPT-5 for a couple of weeks, so I have a lot to say about it. Here's my first post, focusing just on core characteristics, pricing (it's VERY competitively priced) and interesting details from the GPT-5 system card
-
You Must Lie to Your Computer to Get Results
By
–
You gotta lie to your computer these days if you want it to do what you want!
-
Using Codex AI to Ship Blog Enhancements from Mobile
By
–
I use Codex from my phone pretty often – I've shipped a bunch of enhancements to my blog that way https://
github.com/simonw/simonwi
llisonblog/pulls?q=ispr+isclosed+labelcodex
… -
Claude System Prompt Connection Philosophy Explored
By
–
My favorite part of the Claude system prompt is how it ends: "Claude is now being connected with a person."
-
Evaluation Methods for Prompt Engineering Changes
By
–
I would love so much to see the evals you used to figure this one out (or indeed any evals for any of these prompt changes)
-
GLM-4.5 Air Outperforms OpenAI 20B on Space Invader Prompt
By
–
I tried the same space invader prompt against the OpenAI 20B model and thought the GLM-4.5 Air result was better – though that model is 3x the memory size of OpenAI's