Independent AI news, with every claim linked to its primary source
Independent AI news, with every claim linked to its primary source
Slack Code opens shared channels where Claude, Devin, Copilot and Vercel agents work in front of the whole team. Slack says over 70% of those channels close within a day, and its own post lists ChatGPT as still to come.
Read More
Apple pretrains a 2.8B model it never ships, then prunes it to 1.3B under a single learning rate schedule. It reaches 46.4% on MMLU, but only just beats training the small model from scratch.
Read More
Generalist AI sold $198.2M of stock in a Form D filed August 24. TechCrunch reports a $3B valuation with 8VC leading, neither of which appears in the filing.
Read More
OpenAI shipped an Admin plugin for ChatGPT Work and Codex on August 25, letting admins read usage data and change workspace settings inside one conversation.
Read More
OpenAI published the first InferenceX benchmark results for Jalapeño, its custom inference chip, claiming 1.5 to 1.9 times more AI work per watt than the Nvidia GB200 and GB300 systems it was measured against.
Read More
Perplexity Portable Computer runs on-device on an Nvidia DGX Spark, charges no token fees for local work, and reaches Windows RTX machines in September.
Read More
Anthropic has signed a reported $45 billion, six-year deal to rent about 460 megawatts of Nvidia Vera Rubin capacity from Nscale in West Virginia. Neither company has announced it.
Read More
IBM released Granite 4.2 on August 25, 2026 in 3B, 8B and 30B sizes under Apache 2.0. Only the 8B and 30B got the agentic reinforcement learning stage.
Read More
Sentence Transformers v6.0 adds a MultiVectorEncoder class for ColBERT-style late interaction retrieval. A 14.5-hour finetune on one RTX 3090 beat every zero-shot retriever Hugging Face tested on medical search.
Read More
Nvidia told investors it expects 70% revenue growth in fiscal 2028, against a 44% analyst consensus. That implies about $673 billion in sales, and supply is the stated cap.
Read More
Apple M5 Ultra fuses two dual-die M5 Max chips into its first quad-die design, with up to 80 GPU cores, 512GB of unified memory and 1.2TB/s of bandwidth. Every performance figure so far is Apple s own.
Read More
Z.ai says the anonymous Ox Alpha model was GLM-5.3-Flash, now on Hugging Face under an MIT licence, and that it served the whole stealth week on Chinese AI chips.
Read More