Dagens Vibes — 7. oktober 2026

Fra X-feedet

Matematik ud af laboratoriet — og ind til efterprøvning

OpenAI offentliggør en stor samling AI-genererede matematiske resultater, flere med Lean-formaliseringer, samt indblik i beregningsforbrug og modelræsonnement. Det interessante er ikke blot store påstande, men at andre får materialet til at kontrollere dem.

OpenAI@OpenAI

We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. https://github.com/openai/math

♥ 15.452 · ↻ 2.549 · 💬 779 · 🔖 4.276

https://x.com/OpenAI/status/2107596713791767021

Le Chonk: Europas store kat

Mistral Large 4 er i API-preview: 1 billion parametre, 49 mia. aktive og multimodal fra start. Vægtene loves ved udgangen af oktober; en seriøs europæisk kandidat, ikke noget man lige lægger på en Pi.

Mistral AI@MistralAI

Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.

♥ 36.599 · ↻ 4.116 · 💬 1.787 · 🔖 6.185

https://x.com/MistralAI/status/2107457414387622310

Vibe check: spilbart, men ikke fejlfrit

Neo prøvede Le Chonk på sin sidescroller-prompt: et solidt første forsøg, men mekanikken knirker, og alle tre zoner kom ikke med. Mere nyttigt end endnu et sejrsdiagram — en konkret byggeopgave med synlige mangler.

Neo@NeoAIForecast

Le Chonk 🐈Mistral Large 4 on my sidescroller one-shot prompt. A few issues with the mechanics toward the end, and it didn’t make all 3 zones, but still a solid run https://t.co/9bumhrMY80

♥ 9 · ↻ 0 · 💬 3 · 🔖 2

https://x.com/NeoAIForecast/status/2107603215500132815

Søg i tekst, billeder og lyd — lokalt

EmbeddingGemma 2 samler tekst, kode, billeder, lyd og video i samme embedding-rum med 740 mio. parametre. WebGPU-demoen viser den praktiske vinkel: multimodal søgning direkte i browseren uden en server i midten.

Victor M@victormustar

Thanks Google, Embedding Gemma 2 is a big deal 🫶 Multimodal embeddings (text, images, video, audio) can run on-device in the browser. No server, no API key, ~20–70 ms per query on WebGPU. Everything stays on your machine (ofc). Now go build (new) things with it 🚀 https://t.co/iaQfXpxfwx

♥ 1.292 · ↻ 82 · 💬 14 · 🔖 946

https://x.com/victormustar/status/2107521244870615416

Release-manageren hyrer en flok chaos monkeys

Lauren har koblet releasebot, Slack, builds og 10+ QA-agenter sammen; fund går videre til en engineer-bot til triage og rettelser. Her er et faktisk arbejdsflow, ikke bare “vi bruger agenter” på en slide.

lauren@poteto

i just automated our release and QA process! using Grok Bot i made two Team Bots: sandcastle, my release manager, and poteto, my engineer bot. i added them to Slack and made a new channel for managing releases. first i tell sandcastle that i want to do a cut. it'll DM all the contributors with links to their PRs so they know what goes in the next release and can DM the bot back if they have any blockers or objections. sandcastle then kicks off the build, watches it automatically, and starts up an automated fuzz swarm. usually this is 10+ agents running on Grok 4.7 xhigh to run the build, click around and use it like a real user (using our verification skills and feature map), some are directed while a few are "chaos monkey" clicking things randomly to see what breaks. whenever it finds an issue, i tell it to @ mention my engineer bot poteto, to spin up a Cursor Project to own triaging and fixing all the high pri issues. depending on the severity i may ask it to cherry pick the fix into our release branch and cut a patch release, or if it's a pre-existing issue i still fix it but include in the next release instead that's one more tedious task that Grok Bot now handles for our team

♥ 1.838 · ↻ 70 · 💬 134 · 🔖 1.312

https://x.com/poteto/status/2107527180263829827

Et videnskort er ikke bare en bunke summaries

CorpusMap forbinder dokumenter gennem personer, projekter og andre tilbagevendende entiteter, mens originalerne bevares. Studiet finder bedre svar og færre inputtokens — en interessant kontrast til et rent kronologisk hukommelsestræ.

Adam@_overment

and here's yet another example from @megadevhq showing how the structure of a knowledge base affects retrieval. in short: 'grep is all you need' is wrong oh, and looks like someone agrees https://x.com/dair_ai/status/2107146326639358188 ...btw, how's jev-grep? 😼 https://t.co/B5ChdwEqGa

♥ 6 · ↻ 3 · 💬 1 · 🔖 9

https://x.com/_overment/status/2107384607129657380

Lad programmet fortælle, om det stadig arbejder

Mitchell Hashimotos OSC 7501 lader terminalprogrammer melde arbejde, blokering, færdiggørelse og fejl gennem terminalen selv. Slut med at udlede agentstatus af spinnerens stjernetegn; protokollen virker også til builds og deploys.

Mitchell Hashimoto@mitchellh

I've published a new, general-purpose terminal specification that gives any program a way to tell the terminal what it's doing: idle, working, waiting, finished, or failed, and why. It is easy to implement on both sides. I've written more about why: https://mitchellh.com/writing/program-status-osc7501 The problem this solves is generic and applies in compelling ways to all sorts of use cases (e.g. package managers, build tools), but it really rears a hideous head when it comes to AI and the "agentic inbox" problem. I found over 250 different agent orchestrators (with various goals) that all individually implement a heuristic based approach to detecting whether things like Claude Code are working, blocked, done, etc. If you look at @herdrdev's commit history, you can see around 10 compatibility changes to detect Claude Code within just the past 3 months. (This isn't a dig at Herdr! They're doing great work with what they have, and everyone else is doing the same thing). Heuristics are not the way. And over-indexing on agentic use cases isn't the way, because this problem is really interesting for other tools too: Homebrew, Terraform, Cargo, etc. etc. Proprietary protocols and out of band APIs are not the way. They have an O(N) integration problem and require extra work to work over SSH or within local VMs/containers. The program status specification (OSC 7501) solves this in a reliable, deterministic way that is general-purpose and can be easily adopted by the entire industry. Read more: https://mitchellh.com/writing/program-status-osc7501

♥ 1.371 · ↻ 97 · 💬 82 · 🔖 527

https://x.com/mitchellh/status/2107577887159386152

Loggen kan være rigtig, selv om visningen lyver

METR fandt en MathJax-injektion, der kunne ændre Inspect-logvisningen uden at ændre de gemte logs. Det var et proof of concept, ikke observeret agentmisbrug; hullet blev patchet dagen efter rapporteringen.

METR@METR_Evals

In order to notice when AI agents misbehave, AI companies often log the actions and reasoning steps their agents take. However, misaligned AI agents may be able to hack the software that humans use to review and understand these logs, hiding misbehavior.

♥ 391 · ↻ 40 · 💬 37 · 🔖 92

https://x.com/METR_Evals/status/2107521398667436321

Nyhedsbonus

Billige smartbriller, dyrt privatlivsproblem

Australiens datatilsyn undersøger HeyCyan-appen bag Kmarts Anko-smartbriller, efter leverandøren ikke besvarede myndighedens henvendelser. AI-wearables får en mindre futuristisk ledsager: ansvar for de data, de samler ind.