What you need to know
5 stories- 1
Meta releases Muse Spark Open's weights, tuned with Apple to run on two Mac Studios
Meta calls it the most capable US open-weight model, but its own model card scores three other open models higher on the first benchmark it lists.
- 2
Anthropic embeds six METR evaluators with employee-like access, and can't veto what they publish
Six of METR's staff now sit inside Anthropic with roughly the access its own risk teams have, and their findings go out without Anthropic's sign-off.
- 3
Musk says SpaceX will cover Grok Bot's purchase mistakes, but its terms say customers pay
The Grok Bot Terms, updated three days earlier, make the customer responsible for every charge the agent runs up.
- 4
Wyden and six senators ask Meta to justify research tax credits on its AI data centers
Meta's research credits rose from $700 million to $3.9 billion in two years, and the senators want to know how much went on chips.
- 5
Eight states sue OpenAI, saying it sold ChatGPT to parents as safe for children
Arkansas wants more than penalties: it asks a judge to stop OpenAI building models that can improve themselves.
Running stories3
Photo: Nokia621, CC BY-SA 4.0Meta's Muse agent and models
- Oct 20Meta pauses Muse chat sharing after a security firm finds 6,800 share addresses in public DNS
- Oct 15Meta releases Muse Spark Open's weights, tuned with Apple to run on two Mac Studios
- Oct 12Meta releases Muse Wonder, the models Zuckerberg says bring it to the frontier
Photo: Biodin, CC BY-SA 4.0Grok Bot, Musk's shopping agent
- Oct 17Grok Bot users say Amazon blocks their orders, and Musk floats an Amazon rival
- Oct 15Musk says SpaceX will cover Grok Bot's purchase mistakes, but its terms say customers pay
- Oct 11Musk says Grok Bot can shop for you from a photo of your credit card
Photo: Eladkarmel, CC BY-SA 4.0States sue OpenAI chatgpt safety
- Oct 24Texas sues OpenAI and Sam Altman, saying ChatGPT was sold as safe while it harmed Texans
- Oct 15Eight states sue OpenAI, saying it sold ChatGPT to parents as safe for children
In brief7
OpenAIBringing GPT-Rosalind to academic research labs
OpenAI launched a Rosalind Academic Plan on October 15, 2026, giving eligible US university labs up to 30 free ChatGPT Business seats for a year, with GPT-Rosalind access after trusted-access approval. openai.com →
GoogleAccelerating the search for antimicrobial peptides with Google Antigravity
Google DeepMind used its Antigravity agent platform to hunt for antimicrobial peptides, and says a candidate from Streptococcus pseudopneumoniae is now in wet-lab validation. deepmind.google →
MistralNew model repo: mistralai/Timestral-5M
Mistral's smallest Timestral forecasting model, at 5.21 million parameters, is on Hugging Face under a modified MIT license and predicts up to 1,024 time steps in one pass. huggingface.co →
MistralModel repo updated: mistralai/Timestral-28M
Mistral's 28 million-parameter Timestral model, the second smallest of the four, forecasts up to 1,024 time steps at once and gives a range of 21 quantiles rather than a single guess. huggingface.co →
MistralNew model repo: mistralai/Timestral-123M
Mistral's 123 million-parameter Timestral model, one size down from the largest, reads up to 32,768 past time steps of a series before it forecasts. huggingface.co →
2 more
MistralNew model repo: mistralai/Timestral-209M
The largest Timestral model, at 209 million parameters, is on Hugging Face under a modified MIT license, and Mistral's card shows it with the best scores on the GIFT-Eval and TIME forecasting benchmarks. huggingface.co →
MistralIntroducing Timestral
Mistral released Timestral on October 15, 2026, its first time-series forecasting models, in four sizes from 5 million to 209 million parameters that run on ordinary CPUs. mistral.ai →
From the labs' own feeds10
Releases 3
0.163.0Codex CLI+1 pre1.15.0Anthropic Python SDK2.1.301Claude CodeLab posts 3
Status 1
- Outage · OpenAIWidespread feature degradation