Hack. Dev. Transcend. // Polyglot Developer | Ethical Hacker | Biohacker | Youth Tech Advocate | Certified Human Potential Coach
Australia
Joined June 2013
- Tweets7.3K
- Following794
- Followers1.6K
- Likes14.6K
Glenn 'devalias' Grant retweeted
My CPU given Codex Desktop's huge list of bugs and inefficiencies that you seem to not care about solving...
Your models and harness must be pretty weak if they can't even triage / resolve investigated bugs with solutions... or maybe its a human problem.
xitter.cf/_devalias/status/21119…
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.
Glenn 'devalias' Grant retweeted
I love Codex and rely on it for professional and personal work. I need dependable organization, phone sync, and scheduled tasks that keep running. Here’s what keeps breaking—and what I’m asking @OpenAI @OpenAIDevs to fix.
Glenn 'devalias' Grant retweeted
Sooo... when will they have it dig through the 22.9k open issues on github.com/openai/codex/issu… and actually triage/fix them...? Because there is a TON of good investigation there already, and it would do worlds of good for improving the Codex/ChatGPT desktop app stability...
Glenn 'devalias' Grant retweeted
Hey! PM working on the product.
The headline is that users now have access to local computer use inside ChatGPT.com with just an extension.
This means that free Chat users can now bring the context of their tabs into ChatGPT without having to go into Work mode or installing Codex!
This also means that people don't have to install both the Chrome extension + the Codex desktop app to benefit from driving their browser.
As part of these changes, the side chat now mirrors ChatGPT.com's full UI, so you can access your chats, spaces, files everything vs. a stripped down version before.
Finally, now that the Chrome extension is standalone, the reliability has also improved significantly and you should run into a lot less issues with connecting/bugs!
Glenn 'devalias' Grant retweeted
A few big upgrades for Chrome users today! We’re bringing the power of computer use to anyone who uses Chrome.
First up: ChatGPT can now use your tabs in Chrome to make answers more useful and get things done for you.
If you’re in Chat, it can look across your open tabs and gather info.
If you’re in Work, ChatGPT can control your Chrome tabs to get things, like sending messages, filling out forms, or downloading files. It can even help you organize and group that pile of tabs you have open.
Glenn 'devalias' Grant retweeted
A blog on reverse engineering's industrial revolution
rea.tools/blog/touhou-recons…
Glenn 'devalias' Grant retweeted
Stability and bugfixes for many long outstanding and well documented/solutions detailed bugs tracked on GitHub; the amount of time/tokens I waste having to hack around Codex's stability/memory use/etc issues is low key rediculous...
Glenn 'devalias' Grant retweeted
There are about 1.3 billion global children who have zero internet connection in their lives. This is in the age of supercomputing, Starlink, and #AGI. The best way to solve these problems is to ensure that the friction of invention is very low—the birth of IP #freedom is MAPLE.
🤖 Made with AI
Glenn 'devalias' Grant retweeted
To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high.
If you were using high reasoning efforts with Sol and were happy, I suggest you move down to low or medium for Astra.
Glenn 'devalias' Grant retweeted
There are heaps of well sourced / referenced / detailed bug reports, and often even related bug fixes described; on the GitHub issues, but for some reason the Codex team seems to consistently not care / ignore their users contributions there; then ask on Twitter; to ignore again.
Glenn 'devalias' Grant retweeted
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: huggingface.co/moonshotai/Ki…
Tech report: github.com/MoonshotAI/Kimi-K…
Tech blog: kimi.com/blog/kimi-k3
Glenn 'devalias' Grant retweeted
The Wait Calculation: a phenomenon from the study of interstellar travel(!) that is now true of far more endeavors due to the pace of AI.
Glenn 'devalias' Grant retweeted
It's honestly not just on windows; Codex Desktop on macOS is pretty abysmal as well; lots of well researched bug reports sitting rotting for weeks on end when it would take practically no time or agentic effort to implement the fixes that are already discovered.
Glenn 'devalias' Grant retweeted
Maybe Tibothy could update the repo’s AGENTS.md to encode proper issue triage, cross-linking, de-duping, acknowledgement, and follow-through, then make it a regular part of the GitHub Actions and deployment workflow.
xitter.cf/_devalias/status/20768…
Glenn 'devalias' Grant retweeted
What about a folder system where we can organize related projects together, but they keep their individual context? Just saying...
xitter.cf/IgnacioPerezR/status/2…
Really loving the Codex App. 4.56B tokens and 28 projects in, there's one simple feature I'd love to get as the number of projects I bring into it continues to grow: groupings.
Right now I'm faking folder structure with separator rows (eg. "====== WORK ======"). Obviously this isn't great, and one extra layer (collapsible user folders above projects) would clean this up instantly.
Mockup attached. The scrollbar on the left panel is the current state.
Same feature would be applicable for Conductor
What do you think @thsottiaux @charlieholtz?
