Sources
- AI at Meta / @AIatMeta
- Amazon Web Services / @awscloud
- Anthropic / @AnthropicAI
- Cursor / @cursor_ai
- Google / @Google
- Google Cloud Tech / @GoogleCloudTech
- Google DeepMind / @GoogleDeepMind
- Grok / @grok
- Hugging Face / @huggingface
- Microsoft / @Microsoft
- OpenAI / @OpenAI
- OpenClaw🦞 / @openclaw
- Sequoia Capital / @sequoia
- Tesla / @Tesla
- Twitter / @a16z
- Waymo / @Waymo
- xAI / @xai
- Y Combinator / @ycombinator
Company@X — 2026-06-13#
Signal of the Day#
Anthropic was forced to abruptly disable its frontier Fable 5 and Mythos 5 models for all customers globally to comply with a US government export control directive. The national security order explicitly suspends access to these models by any foreign national—including Anthropic’s own employees—marking a severe escalation in geopolitical restrictions on artificial intelligence development.
Key Announcements#
[Anthropic] · Source In response to a sudden US government export control directive targeting foreign nationals, Anthropic has disabled customer access to both Fable 5 and Mythos 5. The company stated they believe the order is a misunderstanding and are working to restore service, noting that access to older Claude models remains unaffected.
[Rio de Janeiro City Government] · Source In a highly unusual open-source release from a municipal IT department, the city of Rio de Janeiro post-trained and launched Rio 3.5 Open 397B, based on the Qwen 7/2 architecture. The model introduces “SwiReasoning,” a novel framework guided by entropy-based confidence signals that dynamically switches between token-heavy chain-of-thought and efficient latent-space reasoning.
[Datacurve] · Source YC-backed Datacurve launched DeepSWE, establishing a new defining benchmark for software engineering agents to replace the heavily gamed SWE-Bench Pro. Because DeepSWE writes tasks from scratch rather than scraping GitHub issues—preventing models from recovering fixes from commit history—the new Coding Agent Index dramatically reordered, debuting Claude Fable 5 as the new top model over Codex with GPT-5.5.
[Aster] · Source Autonomous research startup Aster launched its platform designed to run thousands of AI agents in parallel, aiming to deliver 1000x speedups for open-ended research. The company demonstrated its capability by setting a new world record on the ProteinGym benchmark in just 30 minutes.
[AWS] · Source AWS detailed its partnership with Formula 1 to deploy Track Pulse, a serverless architecture solution that processes raw racing telemetry into actionable broadcast data. The system automatically detects and highlights critical race storylines, such as sudden pace changes and weather shifts, to provide real-time insights for F1’s production team.
Also Noted#
- [Google Cloud] (Source): Google launched Gemini Enterprise Agent Ready (GEAR) resources, outlining five critical architectural components necessary for building modern AI agents beyond simple state graphs.
- [Alibaba] (Source): Industry commentators suggest Alibaba’s Qwen3.7 model is fading from frontier relevance due to the company’s increasingly proprietary stance on its latest weights.