Dagens Vibes — 30. september 2026

Fra X-feedet

DevDay fra værkstedet: Space virker, Dots snubler

Dan Shippers hands-on er dagens bedste samlede læsning: Space slipper for agentens langsomme Google Docs-klik, og hans Dot fanger mødekonflikter i ulæst Slack. Men permissions, beskeder og browserforbindelsen driller — kontorpakken er agent-native, friktionen stadig menneskelig.

Dan Shipper@danshipper

@OpenAI read our full vibe check from hands-on use of Space: https://t.co/Dl5FsEDx0N

♥ 7 · ↻ 1 · 💬 0 · 🔖 5

https://x.com/danshipper/status/2104987546895937709

Sol 6.1: næsten samme bugfangst, en femtedel af regningen

Paweł Huryn gav modellerne to repos med 105 plantede bugs: Sol 6.1 løste 44 for $6,56, Astra 45 for $33 — ét forsøgsforløb, ikke en universel sejr. Theo bruger Sol 6.1 til review og arkitekturanalyse, men beholder Opus 5.5 til selve kodningen; hans senere test understreger også, at Codex-harnesset gør en forskel.

Paweł Huryn@PawelHuryn

So I tested GPT-6.1 Sol on real work: 2 repos, 105 planted bugs. Find and fix what you can. Unlike GPT-6 Sol, which was just a nerfed GPT-5.6 Terra, this one is real. The results: - GPT-6 Astra (max): 45 for $33 - GPT-6.1 Sol (max): 44 for $6.56 - GPT-5.6 Sol (max): 43.5 for $95.35 - Opus 5.5 (max): 41.7 for $58.53 - GPT-6 Sol (max): 29.3 for $9.33 With a model like this, the 50% cut to the $200 plan doesn't matter. n=1. More runs and results at more effort levels dropping in this thread over the next few hours 🧵

♥ 1.662 · ↻ 110 · 💬 128 · 🔖 426

https://x.com/PawelHuryn/status/2105065401193279918

Theo - t3.gg@theo

To be clear, I still think Opus 5.5 is the current GOAT for coding. GPT-6.1 Sol is a great replacement for Astra at a WAY better price, but I don't default to it for code. I absolutely default to it for code reviews, architecture analysis, computer use, email management, and a ton of my other daily work.

♥ 604 · ↻ 3 · 💬 20 · 🔖 92

https://x.com/theo/status/2105001138953327067

Dots følger sagen helt til refusionen

Max Weinbach bad Dots forsøge at få to ikke-refunderbare bookinger refunderet: den fulgte op efter to dage, eskalerede til hotelchefen og opdagede en ekstra kortbetaling. Pointen er vedholdenheden mellem samtalerne; privat Pro-adgang er dog ikke lanceret i EØS, UK eller Schweiz.

Max Weinbach@mweinbach

This is kind of cool with dots I got a little bit sick in Hawaii and after an 8 1/2 hour delay last Monday, the idea of flying to San Francisco for Dev Day was simply not happening (I appreciate the invite OpenAI, it was just too much all like 5 days before) But I had dots go cancel my non-refundable hotel and flight and try to get me a refund. It emailed the hotel, who didn’t respond. After 2 days it decided to email the general manager of the hotel and request it, then got the GM to approve the cancellation with no fee. It noticed there was a $210 charge on my card after the GM agreed, so it followed up to ask why for me. It also submitted a request with Delta for me, citing the 8.5 hour delay and saying the idea of getting on a plane after that was brutal and requesting an exception. I got a call an hour ago from support saying they’ll approve the request and do a full refund. Couldn’t make it to Dev Day, but it still got me a full refund on 2 non-refundable bookings.

♥ 359 · ↻ 6 · 💬 12 · 🔖 35

https://x.com/mweinbach/status/2105049222592532581

Decisions API: Jev får konkurrence, ikke en dødsattest

OpenAI previewer hurtige valg mellem faste svarmuligheder, også med billedinput; Every målte 76/78 korrekte computer-use-valg mod Jevs 73 og lavere latenstid. Ved samtaleklassifikation var præcisionen omtrent ens, men Jev hurtigere — brugbare tidlige tests, endnu ukendt pris.

OpenAI Developers@OpenAIDevs

Give your app real-time decision-making with Decisions API, powered by GPT-6 Luna. Define questions and possible answers to classify content, route requests, or choose an agent’s next action. Available in limited preview. https://t.co/LbaD18M6Do

♥ 4.361 · ↻ 314 · 💬 156 · 🔖 1.653

https://x.com/OpenAIDevs/status/2105003318917697873

Den lille gem: lad agenten klargøre dit review

Lucas Meijers “evaluation pack” er en bedre slutleverance end “done”: browser, tmux og simulator står klar i præcis den tilstand, mennesket skal bedømme. Agenten skal ikke bare lave arbejdet, men gøre det nemt at se, om det virker.

Lucas Meijer@lucasmeijer

@pvncher That’s the whole point: let the agent prepare an “evaluation pack” with live armed browsers, tmux sessions, iOS simulators, so when you show up as human, everything is in the exact state it needs to be for human judgement to take place.

♥ 1 · ↻ 0 · 💬 1 · 🔖 1

https://x.com/lucasmeijer/status/2104952431847362738

Nyhedsbonus

ChatGPT vil være distributionskanalen — ikke bare modellen

TechCrunch samler den større DevDay-pointe: apps bliver foreslået i samtalen, får interaktive paneler og kan bruge kundens ChatGPT-login og AI-kvote. Det er en ny vej til brugerne, men endnu ikke en færdig App Store med betaling og revenue sharing.

America.gov: borgerbetjening starter med en samtale

Den amerikanske regering lancerer én chatbot på tværs af myndigheder med officielle kildelinks; formularer og sagsfremskridt er planlagt til 2027. En direkte F2X-vinkel: samtalen som indgang til offentlig service, ikke endnu en menu med 47 underpunkter.