Dagens Vibes — 8. september 2026

Fra X-feedet

Spotify viser den voksne version af agentorkestrering: lad den dyre model tænke, og lad billigere arbejdere læse bunker og skrive skabelonstof. 90 % er et case-tal, ikke en naturlov — men hooken er stærk.

Daniel San@dani_avila7

Spotify cut Claude Code token usage by 90% with a few simple changes Here’s the quick breakdown: - Keep Claude for tasks that actually require reasoning - Route large file reads to cheaper models like Gemini 2.5 Flash - Use PreToolUse hooks to automatically intercept expensive reads - Return only the relevant context back to Claude - Delegate boilerplate, tests, configs, and repetitive code Don’t spend frontier model tokens on work that doesn’t require frontier model intelligence Link: https://t.co/oAM150SUW1

♥ 27   ↻ 2   💬 5   ⤴ 0   🔖 42

https://x.com/dani_avila7/status/2097154860235792780

Dagens bedste Astra-vibe check er ikke endnu et benchmark. En fire minutters video, skydelære og et irriterende hårproblem blev til en parametrisk Fusion 360-del, som passede i første 3D-print.

Daniel Griesser@DanielGri

I had my “touch of AGI” moment yesterday. I recorded a video of myself talking through a problem: how to improve maintenance of tangled hair in the shower drain. It was a ~4 min video where I explained the problem and also used a digital caliper to take measurements. I asked GPT 6 Astra (Ultra) to come up with a fix for the problem. It took the video, extracted my voice, and matched the video frames to where I measured the parts. It came up with an idea for making maintenance and cleaning easier. I got an interactive sketch so I understand what it is talking about. I told it to use computer use in Fusion 360. I watched in real time as it designed the part, making it fully parameterized. While watching, I learned a couple of tricks, like how making a design read-only frees up one of the 10 free editable designs you get 😅 Anyway, it designed it and it looked solid. I told it to add fillets. I 3D printed it, tried it, and it fit perfectly on the first try. Even though it wasn’t super complex, it solved a real-world problem, probably faster than it would have taken me to wrestle with Fusion and get what I wanted.

♥ 2,801   ↻ 95   💬 56   ⤴ 40   🔖 1,616

https://x.com/DanielGri/status/2096852983468216790

Thorsten Ball bad om 2× hurtigere Monkey-interpreter og blev ved. Slutresultatet: 16× i hans egen måling. Det er et dejligt konkret resultat — og stadig én benchmark på ét projekt, ikke Moses på bjerget.

Thorsten Ball@thorstenball

For fun, I pointed Astra in Amp at the original Monkey implementation and said it should make it "2× faster". It did. Made it 2.3× faster. Then I kept pushing and we ended with... 16× faster. Aim high, I guess? https://t.co/iEg4eMI1Ya

♥ 127   ↻ 0   💬 11   ⤴ 1   🔖 22

https://x.com/thorstenball/status/2096918238680502682

Modgiften til demo-feberen: efter to brændte Pro-kvoter beskriver scaling01 Astra som stærk, men uden nok smag og initiativ. Samme model, to verdener; god prompting og verifikation er stadig jobbet.

Lisan al Gaib@scaling01

after having burnt twice through my Pro limits using Astra exclusively, I can say that I don't feel the AGI the biggest issue is that its still not opinionated and lacks taste. it doesn't really know what to do, so it just does whatever and you end up wasting time and tokens. also I don't believe the rumors that it's 10T+. If it is, then scaling laws are truly cooked. My divine impeccable vibes, which work 1 out of 7 times, tell me that it's 6-8T. overall it's still just a code monkey. although a slightly larger one and the best we have. I would like to refer you to one of my articles: LLMs aren't AGI, but it doesn't matter (we are taking off anyway)

♥ 972   ↻ 22   💬 93   ⤴ 12   🔖 272

https://x.com/scaling01/status/2096925646404403598

Pi har en enkel, rigtig idé: et agent-run bør have et projekt-cwd. Så følger værktøjer, AGENTS.md og skills med den konkrete kodebase i stedet for at drive frit rundt i maskinrummet.

Pi@pidotdev

Pi lets you pass a project directory (cwd) to createAgentSession(), to help scope where tools should work. Tools like read, grep, find and ls efficiently resolve in that project, and the session picks up the projects AGENTS.md and skills. Here’s the code to try this out https://t.co/WbxajgJXzt

♥ 339   ↻ 14   💬 10   ⤴ 0   🔖 157

https://x.com/pidotdev/status/2096929751994687950

David Fowler sætter den hårde prioritet: hermetiske testmiljøer, headless komponenter og en harness, hvor agenten kan køre, debugge og teste. Agentisk flow uden feedback-loop er bare meget hurtigt gætteri.

David Fowler@davidfowl

Building hermetic testing environments has never been more important for the agentic inner loop. Build headless components and a harness that lets the agent run / debug / test should be your first priority if you intend to build with agents.

♥ 43   ↻ 1   💬 4   ⤴ 1   🔖 15

https://x.com/davidfowl/status/2097133084231782823

Nyhedsbonus

Mistral rejser €3 mia. i en Samsung-ledet runde til en værdi over €21 mia. Pengene går til frontier-forskning og compute: Europas satsning på åbne modelvægte og kontrol over egen AI får en temmelig stor tankfuld.

AI-copiloten tager sikkerhedshjelm på: TechCrunch har talt med NavigateAI, der vil give håndværkere byggespecifikationer og manualer gennem mobil og Meta-briller. Det langsigtede spil er video af rigtigt arbejde som træningsdata til robotter — ikke bare endnu en chatbot med arbejdstøj.