vivid 🍌

@bijanbowen anthropic ruined it for all of us :(

Me waiting patiently to give my SSN to openai so I can use GPT5.6

@CommandCodeAI Love how fast it's growing!

We are doubling the usage limits to: 5-hour limit — 30% of your monthly credits Weekly limit — 60% of your monthly credits For all plans Go/Pro/Max. Provider and extra usage has no limit. We've also added major capacity, new providers, and a lot more GPUs to our self hosted infra. - Most models now run 5x to 8x faster - p50 TTFT: 5s → 965ms If Command Code felt slow the last few weeks, give it another go. It's fast now. Bigger plan dropping soon too. If you like the Go plan, this one's gonna hit home for y'all.

@steph_palazzolo @leomschwartz @amir It's like they want china to win. Haven't used any closed models since GLM 5.2 launched, and probably never will again.

New w/ @leomschwartz @amir: The Trump admin has asked OpenAI to stagger the release of GPT-5.6 over security concerns. On Thursday, CEO Sam Altman told staff that the government will be approving access to GPT-5.6 customer by customer, a highly unusual approach.

@ai_for_success more reasons to love opensource.

It's so over, folks... First Fable 5. Now GPT-5.6. The Trump administration has reportedly asked OpenAI to stagger the release of GPT-5.6 over security concerns. Sam Altman reportedly told staff that access to GPT-5.6 will be approved on a customer-by-customer basis. At this rate, most of us aren't seeing GPT-5.6 anytime soon.

@stevibe qwen3.6 27b is such a legendary on-device model, can't wait for 3.7

This looks like a toy. It's actually the meanest little vision eval I've built. The task: look at an emoji image, then repaint it on a 16×16 grid, one pixel at a time. Just the model, a tiny canvas, and up to 2000 brushstrokes. What I didn't expect was the personalities. > some models REGRET a stroke and go back to repaint it > some get stuck looping the same little patch over and over, like they're trying to animate it > some are calm little surgeons and just nail it first try And the task is genuinely mean: it has to see the image, crush it down to 256 cells, then decide what's actually load-bearing: > the tears on 😂 but still keep the smile > the horn on 🦄 > the antenna on 🤖 and keep the soul of it with almost no resolution to spare. 5 models. 7 emojis. Best of 5 runs each. Side by side. Who's your winner?

@TimJayas Claude is a way better model, but its subscription has way less usage

Why do people say Claude is more expensive than GPT? Just look at the API pricing: Claude Opus 4.8 > Input: $5 / MTok > Output: $25 / MTok GPT-5.5 > Input: $5 / MTok > Output: $30 / MTok Yet the narrative is “Claude is too expensive”

@Ali_TongyiLab that was a pretty entertaining video

AI excels in the virtual world, but acting in the physical world is a whole different challenge. Why is it so hard to get a robot to fetch an egg without freezing? 🤔 Check out our new explainer video on Embodied Intelligence. We discuss the gap between robot "Thinking" and "Acting,"

@MatthewBerman sometimes, there is just no solution... The issue is physically impossible to fix without a breakthrough that allows models to communicate with each other while working or something like that.

AAHHHHHHH how can I get my PRs merged into main more quickly I can't figure this out!! No matter what, once I have multiple PRs queued to merge, something always happens and just blocks the rest of them.

@synthwavedd @CommandCodeAI is a good harness. It also gives you 10$ worth of credits for just 1$. The only downside is you can't connect your own provider.

so has anyone figured out what the best harness to use glm-5.2 in is

@atomic_chat_hq ngl GLM 5.2's output looks better

Sakana Fugu surprisingly performed near GLM 5.2 level but 17× more expensive! We gave the same prompt to 4 models: build a complete live Trader Desk with both frontend and backend components, real-time market data fetched from external APIs for 8 symbols, and a custom dark-theme UI. Outputs: Fugu Ultra — 22,225 t, $0.51 Opus 4.8 — 15,802 t, $0.31 GPT-5.5 — 11,474 t, $0.26 GLM 5.2 — 13,677 t, $0.03 Fugu created the most polished and feature-rich trading desk in the run. GLM 5.2 was very close behind, with a similarly complete multi-panel interface and live data, but at a much lower cost. Opus and GPT also performed well, delivering solid results with a better balance between quality and cost

@SakanaAILabs opensource soon right?

Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API. Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls. Try it: sakana.ai/fugu 🐡

GLM 5.2's reality for consumers...

I feel GPT 5.6 is optimized for 3D stuff and fancy demos and will suck at real work like Gemini. I hope I'm wrong tho. What do ya'll think?

@mirochill I feel it's optimized for 3D stuff and fancy demos and will suck at real work like gemini. I hope I'm wrong tho.

GPT 5.6 Pro me prouve encore qu'il comprend très bien la 3D. En seulement 30 minutes, il a généré une maison complète en 3D dans un unique fichier .html, en utilisant uniquement WebGL2. La structure est cohérente, les volumes sont crédibles et l'ensemble est directement navigable dans le navigateur. Il reste quelques incohérences, notamment sur certaines fenêtres, mais la progression est impressionnante. À ce rythme, ce type de défauts risque de disparaître bien plus vite qu'on ne l'imagine 👀

@chetaslua @mirochill How are ya'll testing GPT 5.6?

🚨 GPT 5.6 Pro one Shotted pokemon Game > 31 minutes with a short prompt made > Too bad there's a copyright so he didn't make it exactly like Pokémon > In theory there are 8 badges, and the Pokémon League Thanks @mirochill for running prompts

@Presidentlin and you stole mine x.com/vivid0o0/statu…

You can just steal content. It's like High School, where the kid copying your answer can sometimes do better. Daniel San was first, doesn't matter.

You can just steal content. It's like High School, where the kid copying your answer can sometimes do better. Daniel San was first, doesn't matter.

@Gaus450 .

I’m a Claude Code user, Give me one reason to switch to Codex?

@MatthewBerman congrats!

Almost at 100 stars on day 1! github.com/Forward-Future…

@hqmank wait windsurf's still alive?

Another way to try GLM-5.2: Windsurf subscribers can use it for free for a limited time.