Frontier Daily (Sep 16): 18 items from GitHub (Anthropic), Hacker News, DeepMind and more + X supplements

Every day Laojin pulls the official channels and compiles the past 48 hours of AI news into a short brief: curation only, no commentary, with the original link and publish time attached to every item.

1. anthropics/fermats-last-theorem: GitHub 开源项目

Anthropic open-sourced a project named fermats-last-theorem on GitHub, written mainly in Lean, with 1,173 stars.

Source: GitHub (Anthropic) · 2026-09-15 · Original

2. We got admin access to Baseten's production GitHub in 25 minutes

A blog post claims admin access to Baseten's production GitHub was obtained in 25 minutes, drawing 96 points and 37 comments on Hacker News.

Source: Hacker News · 2026-09-15 · Original

3. Hugging Face is billing OpenAI $100M for hacking it

A report says Hugging Face is billing OpenAI $100M over a hacking incident, drawing 80 points and 29 comments on Hacker News.

Source: Hacker News · 2026-09-15 · Original

4. Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking

Google DeepMind's blog introduces Gemini 3.8 Live and 3.8 Live Extended Thinking.

Source: Google DeepMind Blog · 2026-09-15 · Original

5. [AINews] AEF-1 standard emerges for Third Party Evaluators, as Xai, OpenAI, and Anthropic all cosign

AINews reports that the AEF-1 standard for third-party evaluators has emerged with xAI, OpenAI, and Anthropic all cosigning, alongside Dario's blogpost outlining embedded evaluators, democratic coordination, and global coordination.

Source: Latent Space · 2026-09-15 · Original

6. deepseek-ai/FlashMLA: FlashMLA: Efficient Multi-head Latent Attention Kernels

DeepSeek open-sourced FlashMLA, efficient multi-head latent attention kernels written mainly in C++, with 12,925 stars.

Source: GitHub (DeepSeek) · 2026-09-15 · Original

7. The contagion of fear

Bryan Cantrill responds to a tweet by former Anthropic employee Jacob Coxon confirming many Anthropic researchers believe AI could kill us all by the end of the decade, sharing his own youthful mistake of causing unjustified panic and warning against doing the same.

Source: Simon Willison's Weblog · 2026-09-14 · Original

8. What blog posts influenced your thinking the most?

Simon Willison comments on Lobste.rs about blog posts that influenced his thinking, citing Joel Spolsky's The Law of Leaky Abstractions, Will Larson's Migrations, and Charity Majors' The Engineer/Manager Pendulum.

Source: Simon Willison's Weblog · 2026-09-14 · Original

9. Humanity’s Last Invention — Richard Socher of Recursive

Latent Space hosts Recursive founder Richard Socher to discuss building AI systems that automate AI research and recursive self-improvement, mentioning a $4.65B seed round and early results on optimization tasks and NVIDIA GPU kernels.

Source: Latent Space · 2026-09-14 · Original

10. How Fyxer built an AI executive assistant people trust

OpenAI's blog describes how Fyxer uses OpenAI models, fine-tuning, memory, and real user feedback to organize inboxes and draft emails in each user's voice.

Source: OpenAI Blog · 2026-09-14 · Original

11. OpenAI Launches ChatGPT for Financial Services

OpenAI releases a tailored ChatGPT Work experience for financial services, combining premium datasets from Daloopa, PitchBook, and LSEG with GPT-6 Astra reasoning to build models, research, and client materials using firm templates.

Source: X (Grok x_search, @OpenAI) · within 48h · Original

12. Anthropic Publishes Detailed AI Misuse Threat Intelligence Report

Anthropic releases its most detailed threat intelligence report yet, documenting sophisticated misuse attempts of Claude for cyberattacks, influence operations, surveillance, biology, and weapons, how they were stopped, and lessons to improve safeguards.

Source: X (Grok x_search, @AnthropicAI) · within 48h · Original

13. Dario Amodei Calls for Pacing the AI Frontier

Anthropic CEO Dario Amodei publishes an essay advocating slowing frontier AI development with a three-part plan; Anthropic commits to permanent employee-level access for third-party evaluators, drawing responses from Sam Altman, Yann LeCun, and Andrej Karpathy.

Source: X (Grok x_search, @DarioAmodei) · within 48h · Original

14. Google DeepMind Releases Gemini 3.8 Live Models

Google DeepMind introduces Gemini 3.8 Live and 3.8 Live Extended Thinking as its best conversational AI yet, featuring upgraded reasoning, near real-time visual understanding, 97-language detection, and background tool calling for complex tasks.

Source: X (Grok x_search, @GoogleDeepMind) · within 48h · Original

15. Mistral AI Raises €3B in Series D Funding

Mistral AI announces a €3B Series D, the largest equity round by a European tech company, led by Samsung with participation from EQT, PSG, ASML, Nvidia and others, to scale infrastructure, products, and frontier open-weight model research.

Source: X (Grok x_search, @MistralAI) · within 48h · Original

16. DeepSeek Releases V4.1-Flash Model

DeepSeek launches V4.1-Flash, a 552B MoE model with novel asymmetric Causal Encoder-Decoder architecture (8B active params on input, 16B on output), significantly smaller KV cache, superior performance to V4-Pro, lower prices, and native multimodal support.

Source: X (Grok x_search, @deepseek_ai) · within 48h · Original

17. Sam Altman Endorses Pacing the Frontier

Sam Altman agrees with Dario Amodei's call to pace the AI frontier, reveals it has been a major topic at OpenAI, and commits to providing independent evaluators with employee-like system access while stressing the need to keep safety ahead of capabilities.

Source: X (Grok x_search, @sama) · within 48h · Original

18. Andrej Karpathy Supports Third-Party Evaluator Access

Andrej Karpathy expresses strong support for Dario Amodei's proposal to grant permanent employee-level access to third-party evaluators for verifying safety measures and model alignment during training.

Source: X (Grok x_search, @karpathy) · within 48h · Original

---

_Sources: official blogs of OpenAI, Google DeepMind and Qwen; official RSS of Simon Willison, Latent Space, Import AI and One Useful Thing; Anthropic site monitoring, top AI discussions on Hacker News and open-source updates from key AI orgs on GitHub. X sources are retrieved live via Grok x_search with per-link verification. This is a source digest — no commentary from Laojin._