Leonardo Gonzalez

@luthiraabeykoon @karpathy Wow, super impressive. I wonder if this could be used on the cloud FPGAs AWS offers

We implemented @karpathy 's MicroGPT fully on FPGA fabric. No GPU. No PyTorch. No CPU inference loop. Just a transformer burned into hardware, generating 50,000+ tokens/sec. The model is small, but the idea is not: inference does not have to live only in software 👇

Show us the Codex pets you hatched. Use /hatch to create your own Codex pet. We’ll pick 10 favorites to get 30 days of ChatGPT Pro.

@juliarturc Yes, AI is so much happier with .md I have `marker` pipelines for local PDF conversion, but for ArXiV papers it's SO easy and lightning fast to just use that hf CLI! If you ever need to convert a non-ArXiV PDF, I recommend github.com/datalab-to/mar…

@archimagos nice, automatic conversion to .md? 👀

The enshittification of AI has begun. This girl is just trying to read two measly papers. > NotebookLM: has been timing out over the past two days, doesn't respond at all > Claude app: Sorry can't upload files over 30mb > ChatGPT app: Technically works but a vomit of emojis and sentimentalities even on technical papers. Grateful for no goblins yet.

@iScienceLuvr Codex app-server and OpenClaw is one way

Claude Code has remote control which is amazing but has low rate limits. Codex has high rate limits but no remote control. How do I get the best of both worlds???

@OpenAIDevs This has been my favorite project of the year so far! It inspired me to create OpenSymphony: opensymphony.dev

📣 What if every open issue had a Codex agent? That’s the idea behind Symphony, an open-source agent orchestrator for Codex that turns task trackers into always-on systems for agentic work, letting humans focus on review and direction.

@gdb youtube.com/watch?v=kMF3kt…

what are you building with codex?

@thsottiaux I would like a changelog for each update 😉 In the terminal, mouse support for clicking and changing the cursor position would be great. once In reach a few dozen projects, it would be nice to organize in folders, sections, labels, something.

It’s the little things that matter, what are some small papercuts you have noticed in Codex? We’ll fix as many as possible in the next week.

@TelepathicPug Waiting for Rust support

I hope everyone has been implementing dependency cool downs.

@tmuxvim I hit them repeatedly while using Symphony to build OpenSymphony

the codex limits on $200 seem impossible to hit. I've been grinding so hard lately that my eyes hurt and I've used 5% of my 7 day, 6% of my 5 hour

@dangtony98 This is exciting! I'm spreading the word: aicoe.fit/agent-vault-pr…

ChatGPT Images 2.0 is a real step change in image generation: better text, stronger layout, multilingual rendering, character consistency, QR codes, panoramas, visual reasoning. Here are key demos from @OpenAIDevs launch livestream: aicoe.fit/chatgpt-images…

@theo I've been using it with OpenSymphony and love seeing its Mermaid diagrams for architecture in the PRs it creates 😍

How are people feeling actually using Kimi K2.6 for code?

@thsottiaux You've been making frequent updates. It would be great to see a changelog after updating so we know what's fixed/new 😀

Hello builders. What are we getting wrong with Codex, what can we improve?

@theo Quoted you in my analysis here: aicoe.fit/vercel-has-a-c…

Vercel got pwn’d. Here’s what I’ve managed to get from my sources: 1. Primary victim here is Vercel. Things like their Linear and GitHub got hit with majority of it 2. Env vars marked as sensitive are safe. Ones NOT marked as sensitive should be rolled out of precaution 3. The method of compromise was likely used to hit multiple companies other than Vercel. # of customers impacted is believed to be limited, but hard to know just yet. Will share more updates as I get them.

@raw_works It's open-weight! And really good, especially for multimodal reasoning & document understanding: aicoe.fit/qwen-36-open-v…

@archimagos happy to - can i borrow your api key? ;)

sorry it took me ~50 hrs! now i've got DSPy.RLM as SOTA on LongCOT (Full) by a very large margin, using... ...drumroll... Qwen 3.5 9B! 👑 Qwen3.5-9B + dspy.RLM = 15.69% on LongCoT-full 🔥 ~1.6× GPT 5.2's 9.83% on the same slice!

@raw_works Might as well use Qwen3.6 35B-A3B

27B in progress...

@petergyang Are you running the latest version? There was a specific fix for this in 2026.4.10: github.com/openclaw/openc…

I know OpenClaw isn't part of OpenAI but this feels like a mini-crisis for OpenAI if the GPT integration doesn't improve soon. The bar is GPT needs to be just as good if not better for OpenClaw as Opus.

@outsource_ @RyanLeeMiniMax I think it was ignorance/confusion by the "backlashers" due to an ambiguously-worded license, which led to MiniMax CLARIFYING their intentions, not changing them. People can congratulate themselves for activism but based on what I know about MiniMax I believe their narrative.

🚨 NEW MiniMax updates the M2.7 license after massive backlash! @RyanLeeMiniMax clarified: ✅ Self-hosted for personal coding & internal workflows = FREE & allowed ✅ Distillation fully supported for personal use, internal tools & non-profits ❌ Large-scale commercial external services still need a paid license They admitted the original was “not detailed enough” and an update is coming to make it actually dev-friendly. We really can change a narrative by complaining LOL