Dagens Vibes — 30. juli 2026

Tibo@thsottiaux

Turns out GPT-5.6 Sol is actually SoTA on ARC-AGI-3. Just took two setting changes. You just have to allow it to reason and work over multiple context windows with the help of our canonical compaction implementation. https://openai.com/index/how-two-settings-tripled-our-arc-agi-3-scores/

Jamieson O'Reilly@theonejvo

tldr. Good > @huggingface & @OpenAI sharing details. Bad > 0day (@jfrog) or not, the lack of network controls & least privilege made this sandbox escape, and further actions way easier than it should have been. Also, GPT-6 is almost irrelevant here. With those amounts of open-doors, you could have easily done the same with existing open-weight models (Kimi/GLM) and the right harness. Ugly > Every individual trick it used was old and pretty well-documented. What was new is that it tried thousands of them in a few days without getting bored, tired, or scared - and if you have anything connected to the internet worth hacking, you now have to defend against similar attacks coming from anyone who decides to run one.

OpenAI@OpenAI

After deployment, we applied GPT-5.6 Sol to advance the frontier of efficiency by making itself more efficient to run. The results: - 20% lower serving costs from production GPU kernel improvements. - 15%+ better token-generation efficiency from improved speculative decoding.

jack@jack

CodeCrucible: A blueprint for LLM-driven SAST https://engineering.block.xyz/blog/codecrucible-a-blueprint-for-llm-driven-sast

Satya Nadella@satyanadella

Some more detail on the ROIC Intelligence App I built yesterday and mentioned on today's earnings call. I took the PDF that Brian Nowak at Morgan Stanley put together for Hyperscale ROIC this week and used Copilot code (coming in our new superapp) with a single prompt + skill (/drill-me) to create the plan, then used autopilot in auto to create the full app (with history, lookups, scenarios, what-ifs, etc). And /rubber-duck to test. And the best part is that all the artifacts are in my enterprise environment. My app is in Copilot, my code is in GitHub Enterprise; all my data pipelines/lake/semantic models are in Fabric. And everything is under Agent 365 IT/Sec/FinOps control! So this is not about Tokenmaxxing or vibe coding. Every step of the way the rails are engineered to create value, making everything a long-term reusable asset, with governance/security, and cost controls. This is the full system to drive business value. Disclosures: This is all pulled from public sources, and for illustrative purposes only...not financial advice! :) Here is the app and architecture...

Petr Baudis@xpasky

People of @pidotdev, did you know Jarvis was just GPT-Live hooked to a tmux with a bunch of Pi sessions? You can have one too now - on your phone, your existing tmux setup, across all your projects (and not driving just Pi but anything else in the tmux) https://github.com/pasky/pi-super