For August it's mainly the release and impact of GPT-5, plus the new wave of image editing multi-modal LLMs represented by Qwen-Image-Edit (open weights) and Gemini Nano Banana Here's the August newsletter table of contents
@simonw
-
GPT-Realtime Knowledge Cutoff Compared to GPT-4o and GPT-5
By
–
gpt-realtime has a knowledge cutoff date of Oct 01, 2023 – the same date as GPT-4o https://
platform.openai.com/docs/models/gp
t-realtime
… The GPT-5 knowledge cutoff date is Sep 30, 2024, which makes me suspect that gpt-realtime may be a more similar model to GPT-4o than to GPT-5 -
OpenAI Voice Mode FAQ Lacks Clarity on Underlying Model
By
–
The FAQ doesn't help clarify which underlying model is being used: https://
help.openai.com/en/articles/84
00625-voice-mode-faq
… -
iPhone ChatGPT Voice Mode: GPT-4o versus gpt-realtime models
By
–
Does anyone understand the relationship between iPhone ChatGPT voice mode and the underlying models? Is it still stuck with the old GPT-4o voice model, or is it using the new gpt-realtime model? How does gpt-realtime relate to GPT-5?
-
GPT-Realtime Model Supports Video Input via Image Frames
By
–
The gpt-realtime model supports image input, my guess is that you can feed it a video frame once per second or more as a static image
-
Creating AI Newsletter Editions: Process and Themes Analysis
By
–
I've only done four of them so the process is still evolving, but mainly I read through every page of https://
simonwillison.net/2025/Aug/ to spot the big themes and write based on that I start by copying over the markdown skeleton from the previous edition -
OpenAI Model Code Transparency and Verification Concerns
By
–
The lack of full visibility into the Python code the OpenAI family of models runs is baffling to me – I see this from Deep Research sometimes Being able to see the code is such an important tool for verifying the result, and I can't think of a cromulent rationale for hiding it
-
Real-world success stories beyond vendor fine-tuning tutorials
By
–
That's a tutorial from a vendor that sells fine tuning! I need to see success stories from teams that have put those patterns into action
-
Fine-tuning Models: Cost-Benefit Analysis vs Paid Solutions
By
–
I think it's slow, expensive, very hard to get right and rarely gives you a result that's better than if you paid for one of the more expensive models – and the more expensive models are mostly cheap enough that it's not worth investing the time and money in a fine tuning project
-
Vercel’s Fine-Tuned Production AI Model Strategy
By
–
OK v0 does looks like that rare instance of a company writing about fine-tuned models that they use in production – in their case the "vercel-autofixer-01" model Looks like most of their stuff still uses off-the-shelf Claude though (Cc @0xSMW
)