sami

When thinking about #gpt3, we are probably witnessing the birth of a new Google. Like its beginning for search engines, Google was a big leap compared to other search engines but completely changed its way to index pages nowadays. 1/2 #openai

@kirillk_web3 China leading on open models was not on my bingo card

instead of watching 2 hours of Netflix tonight, watch this 40-minute masterclass from the founder of a $20B China AI company it's the clearest explanation I've seen of how Agent Swarms and AI systems actually work at scale useful whether you've never built an agent in your life or have been using Claude every day for the past year I took the key ideas and turned them into a practical guide on how to actually build with Kimi find it below

@Hikari_07_jp exactly, we are building xybrid.ai because we faced the same issue as you

Local LLM is incredibly complex. Hardware selection, quantization, harnesses, engines, tensor parallelism, unmodified models, MTP… Despite its complexity, local LLM is irresistibly fascinating. I started using X because there was almost no one close to me who could share this excitement with me.

@Capetlevrai on prépare ce futur avec xybrid.ai

👀 Pourquoi je pense que l’ia locale va totalement révolutionner notre usage de l’informatique et pourquoi ça m’obsède Hésitez pas à me dire ce que vous en pensez 👇

@maddiedreese next, implants in your brain 🤡

When I wake up in the middle of the night, I want to fully optimize my waking moments and get as much done as I can. That’s why I have Codex projected above my bed at all times. It’s the only way.

@reachmeviz @cohere sharing the feeling…

@cohere Running on 2 H100s is not “little hardware” . I got excited and then dejected

Introducing: Cohere Command A+ We’ve created our most powerful LLM yet, optimized it to run on as little hardware as possible, and released it open-source for all.

@cohere “little hardware” = 2 * H100

@antigravity Antigravity quota

Introducing Antigravity 2.0, a new standalone desktop application that delivers fully on that original glimpse of a truly agent-optimized experience. Rebuilt from the ground up with multi-agent teams, scheduled tasks, native voice and one-click integration with other Google products. Learn how to get started with Antigravity 2.0 👇

Antigravity 2.0 and Codex #codex #google

@btraut “error running remote compact task”

Now that I've joined the Codex team, it's so freakin' cool being able to fix paper cut bugs that I was running into before the switch. What paper cuts are you running into? I'll see what I can do.

@thsottiaux reliability, couple of times i had errors completely messing up a long running goal and loosing context

For those of you living inside the codex app, what should we prioritize among features, reliability or performance?

he's become fully reliant on LLMs to code. now increase the price by 1000%

@cjzafir do you face any issues with long running tasks and failing compactions? seems to be a recurring issue right now

Codex 5.5 is AGI for me. Before going to bed last night I asked Codex to autonomously fine-tune Qwen 3.5 4B model. > Created fresh Google Colab notebook > Uploaded 145M JSONL dataset to Google Drive > Pasted Colab notebook link in Codex Desktop App > Asked it to train Qwen 3.5 4B using Unsloth > It opened Colab using its Chrome Extension > Connected to my Google Drive > Downloaded Dataset from Drive > Installed Unsloth > Started running commands in sequence. > Fixed all errors brilliantly. > Completed all steps properly > I woke up to this. Interesting: Codex put itself to sleep for 30-minute interval via automation. (0 limit loss during 4-hour training cycle). Codex's chrome extension for computer use is amazing. It doesn't eat context like other MCPs and it clicks fast. I'm in love with this model. Worth $200.

@TheZachMueller @liquidai @ZyphraAI @OpenAI @Google @PrismML indeed, my bad it's a dense model, though check have a look, it's quite impressive how fast, small and good those are any reason why the focus on MoE arch ?

@sa_mous @liquidai @ZyphraAI @OpenAI @Google @PrismML Bonsai is not MoE however

List out your favorite companies doing work on SmolMoE’s (OSS, <30B) Just to name a few: - @liquidai - @ZyphraAI - @OpenAI - @Google {you’re helping my talk so TIA}

Source from @azeem : @exponentialview/note/c-255230688" target="_blank" rel="nofollow noopener">substack.com/@exponentialvi…

Goldman expects 120 quadrillion tokens/month by 2030. That's 24x today's volume. Translation: 24x your inference bill. If your product runs on gen-AI and margins matter, a hybrid inference strategy isn't on the roadmap anymore. It's the roadmap. Long live local models!

remote control codex sessions @thsottiaux plzz

Today we're releasing ZAYA1-8B, a reasoning MoE trained on @AMD and optimized for intelligence density. With <1B active params, it outperforms open-weight models many times its size on math and reasoning, closing in on DeepSeek-V3.2 and GPT-5-High with test-time compute. 🧵

By the end of the year, we’ll see a major shift toward local coding agents. Why keep paying for heavily subsidized AI coding subscriptions when local models keep getting better, cheaper, and more private? The end of the “burn VC money for infinite inference” era is coming!