So ZAI is going full RSI on GLM-5.5.... Self-improving coding agents generating their own training data + internal model merges Next GLM is going to be nasty at coding💀👀 https://t.co/BuAG2qQxfN

Wow… Google just got a poem back from space Project Suncatcher’s prototype satellite successfully powered up a TPU in orbit, ran an AI prompt and beamed a full response down to Earth a haiku written from above the planet…. Blue marble below Data streams in silent grace Starlight feeds my core first “hello world” from an actual AI chip running on sunlight in space Looks like we’re not just launching rockets anymore… we’re launching compute…

GPT-6.1 Terra is basically Grok 4.7 💀 https://t.co/9fwkTrVEXf

Elon said Grok 4.8 would finish pre training in mid September and start RL It’s been 6 weeks..... still no confirmation it actually finished or not is the the new C++ stack is taking longer than expected?? https://t.co/U00xwp1h9e

We are in the final week of OpenAI’s 28 day Codex challenge.... they spent most of it shipping small fixes and resetting limits whenever there wasn’t one.. tho no new model dropped the entire time the only thing that would actually make the ending feel worth it is Terra If it lands in the next few days the resets become setup.. if doesn’t, the whole 28 days just ends with one last wipe

the Microsoft signal on Terra is louder than the usual rumors... they already switched 18 evaluation configs from GPT-5.6 Terra to gpt-6.1-terra and GitHub is now testing it inside Copilot specifically as a cheaper alternative to GPT-6.1 Sol microsoft doesn’t usually move a bunch of evals and start wiring a model into Copilot unless they expect it to be useful.... whether Terra actually beats Sol or just matches it at a lower price the fact that they are already testing it this way suggests OpenAI is close to shipping something they plan to put in front of developers

looks like OpenAI’s next drop is GPT-6.1 Terra not GPT-6.1 Astra… while people are waiting for a real response to Opus 5.5, OpenAI is about to ship the middle child instead 🤡 Terra is basically a cheaper alternative to Sol...

DeepSeek V4.1 Flash 1017 is basically Open weight GPT-6.1 Sol xHigh at GPT-6 Luna pricing OpenAI is in really bad position rn 💀 https://t.co/rMLt3xOBPy

OpenAI needs to release GPT-6.1 Astra before Fable 5.5 drops... as Anthropic is already winning the usage war and Claude Code is leading coding agents, taking the majority of public agent commits and the top spot in developer surveys... Codex is growing but still well behind Fable 5.5 showed up in Azure’s catalog means release is imminent.. If Anthropic ships it first they take the week on top of the lead they already have OpenAI finished 6.1 Astra once and pulled it waiting now just hands them another win

OpenAI Ultrafast pricing on GPT-6.1 Sol is crazyy OpenAI just put Ultrafast on GPT-6.1 Sol.... near-Astra speed and up to 8x faster than Sol Standard available on the API, Codex, and ChatGPT Work but the price is six times the standard..$12 / $60 while Standard Sol is $2 / $10 literally six times the rate for the speed is crazy also it is only on the $500 plan Plus and Pro 200 don’t get it... you don’t pick the fast seat you buy the line lol