← Dagens Vibes

Dagens Vibes — 4. juni 2026

Dagens vibe: open weights og coding-agenter presser prisen ned og ambitionerne op. Laptoppen får multimodal hjerne, stemmen får følelser, og Codex får både Hollywood-trailer og 13 timers Switch-port. Subtilt som en jetmotor i en cykelkælder.

Fra X-feedet

Det største signal i feedet var local/open-weight-bølgen: Google landede Gemma 4 12B som en multimodal agentmodel til almindelige maskiner, Miso smed en 8B voice-model ud med vægte, og Ideogram åbnede sin designmodel. Mindre “vent på API’en”, mere “klon repoet og brænd GPU’en”.

Google@Google

Today we’re introducing Gemma 4 12B — our latest open model that brings advanced agentic reasoning, vision and audio directly to your laptop. It delivers performance nearing our larger Gemma models with a much smaller total memory footprint, while being small enough to run locally with just 16GB of VRAM. It’s open and accessible for everyone to use under a permissive Apache 2.0 license. This is all made possible by our new, unified architecture that removes separate multimodal encoders. Here’s how we did it 🧵

Promotional graphic on a black background featuring the large blue text "Gemma 4 12B" above smaller white text that reads "Unified Transformer." A glowing blue ribbon containing multi-modal icons (representing images, text, and audio) flows from the left into a central point, branching out into a complex, luminous blue neural network map on the right.
♥ 6.3k↻ 841💬 165🔖 2k
https://x.com/Google/status/2062203526588088452
Aoden Teo@AodenTeoMT

Today, we’re excited to introduce Miso One, the most emotive voice model in the world. Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency. We’ve open-sourced the model weights, with API access coming soon. Hear how Miso One sounds in the thread below.

♥ 3.7k↻ 323💬 243🔖 4.4k
https://x.com/AodenTeoMT/status/2062204362102100295
Ideogram@ideogram_ai

Introducing Ideogram 4.0: the best open image model in the world.

Think it. Make it. Own it.

Download the weights, fine-tune on your own data, and run it on your hardware. Live on every Ideogram plan and the API today. https://t.co/AdH9hfSEdb

♥ 5.2k↻ 550💬 297🔖 4.1k
https://x.com/ideogram_ai/status/2062202208700313872

Codex-sporet var todelt: OpenAI teasere og udviklere, der faktisk lader agenten gøre absurd meget arbejde. Christoph Nakazawas Switch-port er ikke produktionsklar, men det er præcis den slags demo der flytter folks mentale grænse.

OpenAI@OpenAI

It's time to fly. https://t.co/ObUaCZ07EM

♥ 8k↻ 595💬 789🔖 1.5k
https://x.com/OpenAI/status/2062249312839434452
Christoph Nakazawa@cnakazawa

Codex spent 13 hours porting Athena Crisis to the Nintendo Switch. Honestly hard to believe. It supports most gameplay features correctly, is playable, and looks decent. Lots of bugs, and I will throw this away, but feeling extremely inspired right now. https://t.co/paAcGiuB2C

♥ 26↻ 0💬 3🔖 4
https://x.com/cnakazawa/status/2062330769671094428

Agent-økonomien blev også mere kontant. Uber sætter cap på token-forbrug, Gergely peger på pris som konkurrenceparameter, og ClaudeDevs måtte omdøbe “workflow” til “ultracode”, fordi almindelige ord nu åbenbart er farlige trylleformularer.

Simon Willison@simonw

Uber reportedly now caps coding agents at $1,500/month per employee per tool - seems sensible to me, but it's also an interesting hint at the value Uber thinks these tools are providing simonwillison.net/2026/Jun/3/ube…

♥ 484↻ 42💬 88🔖 167
https://x.com/simonw/status/2062143151184465964
Gergely Orosz@GergelyOrosz

It's starting. The company with the cheapest coding model at a good enough quality is set to win big, as devs + companies become more price sensitive. Of all the major players, Cursor is in a v good position with their Composer model. Plus eg Factory with smart routing

♥ 446↻ 24💬 33🔖 80
https://x.com/GergelyOrosz/status/2062318223446466611
ClaudeDevs@ClaudeDevs

We've changed the trigger word from "workflow" to "ultracode". You can still say "use a workflow for this", but when you're clearly referring to something else, Claude won't kick off a dynamic workflow. For an explicit trigger, use "ultracode". We appreciate the feedback!

♥ 3.3k↻ 169💬 185🔖 952
https://x.com/ClaudeDevs/status/2062257177788858398

Nyhedsbonus

Google AI Edge-bloggen gør Gemma-nyheden konkret: macOS Edge Gallery, lokal diktering/redigering og en OpenAI-kompatibel lokal server via LiteRT-LM. Det er den praktiske vinkel, ikke bare leaderboard-konfetti.

Bringing Gemma 4 12B to your Laptop: Unlocking Local, Agentic Workflows with Google AI Edgedevelopers.googleblog.comhttps://developers.googleblog.com/bringing-gemma-4-12b-to-your-laptop-unlocking-local-agentic-workflows-with-google-ai-edge/

Miso-bloggen er værd at læse for arkitekturen: RVQ-transformer, audio-kontekst og en model der forsøger at komme ud af TTS’ flade robotmund. Stadig halv-duplex, så den afbryder dig ikke endnu. Small mercies.

Releasing the MisoTTSmisolabs.aihttps://www.misolabs.ai/blog/miso-tts-8b

Simon Willison sætter Uber-cappet i perspektiv: $1.500 pr. medarbejder pr. coding-agent pr. måned er ikke panik — det er et datapunkt for, hvad virksomheder begynder at mene AI-assisteret udvikling er værd.

Uber Caps Usage of AI Tools Like Claude Code to Manage Costssimonwillison.nethttps://simonwillison.net/2026/Jun/3/uber-caps-usage/