Dagens Vibes — 8. juni 2026

Hovedviben: agentic coding er ved at skifte fra chat-prompting til loops, worktrees og små automatiserede kollegaer. Stadig softwareudvikling — bare med flere knapper der kan sætte ild til ting.

Fra X-feedet

Agent-loops er dagens hovedspor: mindre “prompt og håb”, mere systemer der kan reproducere, tjekke og åbne PRs uden at blive efterladt alene i kælderen.

Peter Steinberger 🦞@steipete

Here’s your monthly reminder that you shouldn’t be prompting coding agents anymore. You should be designing loops that prompt your agents.

♥ 11.6k↻ 720💬 946🔖 8.3k
https://x.com/steipete/status/2063697162748260627

Codex-appen bliver mere arbejdsstation end chatvindue: automations, worktrees, skills, connectors og sub-agenter samlet samme sted.

Vaibhav (VB) Srivastav@reach_vb

You can do this directly in the Codex App: 1. Automations to run autonomous discovery and triage 2. Worktrees to work on isolated features 3. Skills to codify your project/ dev specific knowledge 4. Plugins/ Connectors to connect to tools that you work with 5. Sub-agents to ideate and verify work All of this with a simple markdown/ linear for state tracking is all you need! Bonus: you can distill this all in one thread and have it spin up projects/ threads as needed

♥ 738↻ 32💬 46🔖 1.1k
https://x.com/reach_vb/status/2063713960495558940

Den voksne version af vibe coding: byg loops, men kig stadig på arkitekturen. Ellers bliver det bare en meget hurtig slopmaskine med selvtillid.

dex@dexhorthy

been thinking about this a lot and grappling with the sheer breadth of the ai coding experience/skill spectrum Every time I see someone say “stop building the thing and instead build the thing that builds the thing” I eyeroll super hard at this engineers-playing-ant-farm slop cannon stuff And then I meet people who genuinely would benefit from steering at least 20% more in that direction, compared to their current “8 manually prompted tmux sessions on a VPS somewhere”

♥ 9↻ 0💬 1🔖 1
https://x.com/dexhorthy/status/2063817578615918655

Antirez rammer testsporet: AI ændrer ikke bare hvor hurtigt vi skriver kode, men hvor aggressivt vi kan jage regressions og edge cases.

antirez@antirez

[blog] A new era for software testing: https://t.co/J52JqzKyfh

♥ 554↻ 66💬 20🔖 490
https://x.com/antirez/status/2063559923003281836

Apple står klar til WWDC med endnu et “nu mener vi det faktisk”-forsøg på Siri og AI. Denne gang med Cook tættere på maskinrummet.

Mark Gurman@markgurman

Tim Cook has been intimately involved in Apple’s AI comeback and AI product direction since Apple Intelligence flopped and the new Siri was delayed. He’s never been this involved in product development and features as CEO. https://t.co/4hgg1PiTQp

♥ 1.3k↻ 73💬 47🔖 131
https://x.com/markgurman/status/2063636991980572688

Dagens forskningsnørd: et paper der fik feedet til at sige “insane”, hvilket i 2026 både kan betyde genialt og ulæseligt. Ofte begge.

Miles Cranmer@MilesCranmer

This is an insane paper and I love it https://t.co/DP8OR5NJf2 https://t.co/rl4Rmr0FhJ

♥ 10.8k↻ 1.2k💬 153🔖 6.2k
https://x.com/MilesCranmer/status/2063169819474546837

Lokalt model-hack: Gemma 4 MTP er landet i llama.cpp, og open source-inferens toget kører videre med alt for meget fart og alt for få bremser.

Omar Sanseviero@osanseviero

Gemma 4 MTP just got officially merged into llama.cpp This means you can use Gemma 4 QAT + MTP for a lightweight + super fast setup. Excited to see what the community builds with it https://t.co/1te7tgdi2H

♥ 879↻ 79💬 50🔖 464
https://x.com/osanseviero/status/2063676865441665426

Nyhedsbonus