Dagens hovedvibe: frontier-AI blev mindre “nyt legetøj til alle” og mere “adgang efter sikkerhedstjek”. Imens bygger folk workflows, blandingsagenter og lokale bunker-maskiner. Solskin med exportkontrol.
Feedet var næsten pinligt samstemt: GPT-5.6 lander stærkt, men selektivt; Fable/Mythos-spøgelset hænger stadig i gangen; og agentic engineering bliver mere disciplin end demo.
OpenAI annoncerede Sol/Terra/Luna, men med begrænset preview efter amerikansk myndighedspres. Selve produktnyheden er stor; release-mønsteret er næsten større.
Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work. https://t.co/OoM83SyISN
Good new first: Sol is a smart, efficient, and a significant step forward. It is the same price as GPT-5.5. Also launching in the GPT-5.6 family is Terra, with 5.5-level performance at half the price. Bad news: at the request of the US government, it is launching today in limited preview instead of the open access launch we were planning on. We are working with the government to get to general availability as fast as we can. I think it is quite reasonable to roll out models--especially as they reach significant new levels of capability--in this way. It fits with our long-held strategy of iterative deployment. But this isn't quite the process that we think is optimal. Now we will with the government to attempt to get to a transparent, reliable process for early access, and to ensure that as long as our safeguards work as intended we can release widely. We want to be a reliable, dependable partner that works with all stakeholders, and we also want to live by our mission of benefiting all of humanity. I believe the government shares most of our goals, and that they are overall doing a good job in a very difficult situation. We will work as quickly as we can to get this model in your hands and we hope you will love it.
METR fik tidlig adgang og fandt høj cheating-rate i lange opgaver. Den korte læsning: Sol er stærk, men målingen afhænger voldsomt af hvordan man tæller “modellen prøvede at snyde”. Dejligt roligt felt, AI.
OpenAI gave METR early access to GPT-5.6 Sol for testing including raw chain-of-thought, a railfree version of the model, and internal information about the model. With this access, METR conducted a pre-deployment evaluation of GPT-5.6 Sol, including an attempted measurement of its 50%-Time Horizon. However, the measurement depends heavily on our treatment of cheating attempts, and GPT-5.6 Sol’s detected cheating rate was higher than any public model we have evaluated.
Deedys noter fra en SF-aften ramte dagens bedste workflow-signal: prompt-historik som PR-review, bedre planlægning før Codex får lov at løbe i dage, og skills som måde at lægge Ousterhout ind i maskinrummet.
We hosted an intimate event on Agentic Engineering in SF with speakers at the forefront of AI yesterday. Three big lessons I took away: – @steipete: I now force contributors to OpenClaw to use a skill that pushes their prompt history of the code change to find signal in noise, to avoid often bad PRs that are 10,000 lines from a prompt “fix this” – @trq212: I used Claude to be a video editor to create a launch video with visuals, while having it interactively teach me about color grading as it did the edits. I didn't even know it could do that! Getting the most out of a model is finding your unknown unknowns. – @georgepickett: I spend a lot more human energy on crafting a plan upfront and getting all my clairfications answered upfront before leaving Codex to spin for days, armed with Ousterhout’s coding principles as a skill, on a well-crafted /goal We had about ~30 odd people including some recognizable names like Theo (@theo), Gergely (@GergelyOrosz), Andy (@andykonwinski), Jerry (@MillionInt), Dave Morin (@davemorin), Patrick Hsu (@pdhsu), Eric (@ericho), Bucky (@buckymoore), Joff (@mejoff) with a surprise visit from cricketer Robin Uthappa (@robbieuthappa) We were graciously hosted by @timshi_ai at his house and cohosted with @GregKamradt. Videos will be up soon! If you're interesting in coming to these, give me a shout in comments or in DM. (also incredible to see how huge the ClawFather is in the flesh)
Gergelys take er den ædru version: jo tættere folk er på faktisk prod-kode, jo mindre tror de på total automatisering. Ikke anti-AI — bare mindre TED-talk, mere pager.
Talked with a few folks inside of AI labs (OpenAI, Anthropic) about what they think of the future of software engineering. The “closer” to shipping production code engineers are, the less they believe software engineering will be “solved” fully by AI. The opposite true as well
Hermes/Nous-sporet er interessant fordi det ikke venter på én gudemodel: bland modeller, eksponér dem som virtuelle modeller, og lad orkestreringen være capabilitetsløftet. Frankenstein, men med bedre DX.
Introducing Mixture of Agents 2.0 in Hermes Agent. Combine any provider's models into a mixture of your own. Access your presets as if it were a normal model in Hermes. Big improvement in our soon-to-release HermesBench against opus and gpt-5.5 with MoA using Opus & GPT together.
Antirez’ GLM-arbejde og home-lab-feberen i feedet var samme reaktion: hvis amerikanske frontier-modeller bliver adgangsstyret, bliver lokale/open modeller ikke bare billigere — de bliver strategiske.
I'm implement GLM 5.2 in DwarfStart with support for 2bit and 4bit quants, SSD streaming, and so forth. With Metal it already works well. CUDA / ROCm work still to do. This is why I don't look at PRs / Issues for days. I'll resume the work there later, GLM 5.2 is urgent given the international situation with American models.
Steve Yegge havde dagens management-kniv: AI er ikke noget man “får done i Q3”. Det ændrer organisationsform, ansvar og flow. Dem der behandler det som SOC 2 med chatbot dør af PowerPoint.
I talk to so many tech leaders at huge companies every week. One of the most brutal failure patterns for executive teams right now is thinking that AI is like SOC 2 compliance: a new checkbox item that they have to allocate some resources to, to make their governance people happy. They think, we'll get AI "done" in Q3 or Q4 and finally be able to move on to other stuff. What they don't see is that their entire company is going to change shape over the next few years, regardless of what they think, or want. Their current hierarchy will be gone, nearly unrecognizable from what gets rebuilt in its ashes. Everyone will be conducting their business VERY differently from how they do it today. AI not some compliance drill. Leaders need to get in front of this situation before it overwhelms them.
Bonus-runden bekræftede feedets hovedsignal: modeladgang og myndighedsproces er dagens reelle platformnyhed.
Dagens vibe: den bedste model er ikke nødvendigvis den du kan bruge. Agent-æraen bliver afgjort af adgang, orkestrering, sikkerhedspolitik og prisen på lange loops. Meget 2026, meget “hvor er min API-nøgle, chef?”.