Low key embarrassing to be someone driving Codex Desktop development at @OpenAI, with such capable agents and models, and still unable to do use them to do the most basic of software engineering tasks like bug report triage / fixing; when half of them are pre-investigated anyway.
It's actually low key kind of sad/pathetic tbh.. like, in a traditional software dev environment I could respect the time it takes to fix bugs/etc; but with an army of highly capable AI agents at hand, and with users filing far more detailed bug reports (and often solutions) [..]
At API rates on GPT-6.1 Sol; that would be ~US$132.06 in wasted tokens being inefficient, when it already had a perfectly useful tool it could have called to do so better.
Thankfully 99.7% of those hit the prompt cache (US$0.10/1M), rather than being the full US$2.00/1M equiv 💀
Thanks @OpenAI Codex for 2 threads recently deciding that instead of using their existing efficient interruptible sleep tool; or keeping questions open and uncollapsed when a turn 'ends' in general; they'd go into 1min sleep loops and burn ~500mil + ~700mil tokens waiting 🙄
Honestly but.. like ignoring model quality and usage limits and all the main things people complain about; the number of SERIOUS bugs and performance issues in the ChatGPT / Codex desktop app that have sat largely ignored despite users reporting them on GitHub / etc lk pathetic
I agree here. Every couple of months they say the ChatGPT desktop app will get and has got efficiency improvements, but they never come. The app is still horrible to use, and the models themselves just don't get any work done.
I'm now having Claude manage the OpenAI models and I've moved all my projects over to Claude, which feels like it has more generous weekly usage.
We can't forget that ChatGPT has the vast majority of the market share, but in reality, the top AI lab right now is Anthropic.
I really want to see all labs and models do well, and for me particularly OpenAI. But if they haven't hit the red alert button yet, they desperately need to.
Also to read / triage the GitHub issues, where many people have reported (and sometimes fully investigated the root causes of) the myriad of problems that Codex has currently...
Introducing tsrs: a Rust port of the TypeScript compiler 🦀
We are not forking TypeScript!
◆ In close collaborating with the TypeScript team
◆ Up to 8x faster than TS 7.0 & ~70% less memory
◆ Runs in WASM, so it works in your browser & on Workers
◆ Soon part of the Oxc toolchain
npx tsrs
blog.cloudflare.com/typescri…
Lets get that codex performance finally!
But... can we have some Codex CLI / Codex Desktop version numbers where the improvements have landed please...
I've also been working to fix paper cuts in the app. Side chats should start faster, communicate better with their parents, have fewer tool call failures, and have clearer UI around activity.
If you have more feedback, please reach out!
Took a while, but I think I managed to squash all the places we were leaking memory for long-running turns. You should be able to /goal as long as you like now!
AI agents are writing more code. How can developers ensure that code is correct?
Microsoft Research’s Proof-Driven Development (PDD) Kit explores a more rigorous approach: it translates a developer’s intent into a formal specification, then uses formal verification to prove that generated code meets it.
PDD Kit works with Verus for Rust and is integrated with @GitHubNext’s Chopin, bringing formal methods into an AI-assisted development workflow for teams.
See the work in action at @github Universe, Oct. 28–29, 2026.
How do you _guarantee_ your agent's code is correct? You prove it.
Introducing Proof Driven Development Kit (PDD Kit) from @MSFTResearch.
PDD Kit helps turn your plan into a formal spec, then proves the agent's code meets that spec for every input. No vibe verification. PDD Kit is integrated with @GitHubNext's Chopin, so you can plan as a team.
PDD Kit takes an app's login flow from intent to a verified pull request:
It discusses your code, infers your intent as it goes, and proposes a formal spec in Verus, a proof language for Rust.
Once you approve the spec, the Verus verifier automatically proves the code meets it.
Where proof isn't possible, it falls back to empirical verification with tests and computer-use agents (CUAs).
It works on Rust today. No new language, no change in how you produce the underlying code, making formal methods more accessible.
The goal is a verified software factory, where engineers review specifications instead of code.
Come see it in action at GitHub Universe, Oct 28–29. Swing by the Speakeasy and ask for GitHub Next!
Saw this Beeper <-> ChatGPT Dots integration pop up on GitHub the other day and been watching it's development/integration waiting for this announcement!
That feel when @OpenAI Codex bugs/issues burn ~42-80 hours of your life (not counting lost productivity/opportunity cost/etc) on a whole pile of bugs and performance issues that could have been triaged and fixed SO much sooner if they were actually prioritised.
Then looking at my time tracking:
> The strongest figure is 42 hours 47 minutes logged against explicit Codex-related troubleshooting, across 138 entries on 34 dates. Including recovery tools and workflow work brings it to 80 hours 34 minutes, after removing overlapping timers.