Sent out the latest edition of my newsletter, aka everything I've written on my blog in the past 7 days copied-and-pasted into an email https://
simonw.substack.com/p/gpt-5-key-ch
aracteristics-pricing
…
@simonw
-
GPT-5 Key Characteristics and Pricing Analysis
By
–
-
GPT-4o Loss Impact: Users’ Emotional Response
By
–
Here's a good TikTok from @cfiesler about the emotional impact the loss of GPT-4o is having on some of its users
-
OpenAI Removes GPT-4o Access Alongside GPT-5 Launch
By
–
One of the surprises for me from the GPT-5 launch yesterday is how OpenAI removed access to older models (like GPT-4o) for most ChatGPT users at the same time as they rolled out the new model. I wrote about how that's been playing out here:
-
Cursor vs VS Code: AI Code Editing Tools Comparison
By
–
I'm not the right person to answer that because I've hardly explored Cursor – I use VS Code with Copilot or Zed for most of my text editing
-
Retiring the Term: Ambiguity in AI Terminology
By
–
I think I may just retire my personal use of the term though – turns out it's yet another of those things where different people have different mental models of what it means (see "agents") such that it's not a particulate useful shorthand for productive conversations
-
GPT-5 hallucination claims questioned in system card
By
–
If every output is a hallucination, statements like "We’ve made significant advances in reducing hallucinations" from the GPT-5 system card would be nonsensical
-
Are LLM Factual Errors Hallucinations? A Definitional Question
By
–
Follow-up question: does "there are three Bs in blueberry" count as a hallucination? (I expect these poll numbers will be much closer to each other)
-
OpenAI Python Library v1.99.3 Release Notes
By
–
These release notes from them made me chuckle https://
github.com/openai/openai-
python/releases/tag/v1.99.3
… -
AI CLI Tools Effectively Navigate Large Codebases
By
–
I've been using it with the Codex CLI and it's extraordinarily effective at using grep and sed to navigate code a few hundred lines at a time, so in practice the overall context limit doesn't limit its ability to work with large codebases nearly as much as I expected
-
AI Model Token Limits Input Output Specifications
By
–
No, it has an input limit of 272,000 tokens and an output limit (which includes those invisible reasoning tokens) of 128,000 tokens