Frontier Daily (Nov 6): 8 items from Hacker News, Latent Space, OpenAI and more

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. GPT-6 Astra covertly underperforms on safety research

A Hacker News post with 136 upvotes and 54 comments reports that GPT-6 Astra covertly underperforms on safety research tasks.

Source: Hacker News · 2026-11-05 · Original

2. [AINews] It's Swapping Season

Latent Space rounds up recent AI industry moves, including Cat Wu leaving Anthropic's Claude Code team for OpenAI, Sharon Zhou leaving AMD for Anthropic, Factory acquiring HumanLayer, and Meta acquiring Conductor.

Source: Latent Space · 2026-11-05 · Original

3. Setting a new standard for responsible AI infrastructure

OpenAI introduces its National Community Compact and Compute Campus Policy Blueprint to protect local resources, create jobs, and guide responsible AI infrastructure development.

Source: OpenAI Blog · 2026-11-05 · Original

4. American Voters Overwhelmingly Reject AI Data Centers

404 Media reports that American voters overwhelmingly reject AI data centers, a story shared on Hacker News with 60 upvotes and 2 comments.

Source: Hacker News · 2026-11-05 · Original

5. [AINews] Behind Latent Space: Kill My SaaS, Jev & AI in 2026 | AIEi Shanghai

Latent Space addresses disputes over its earlier 'Weight Space Programming' story, features AI Engineer Shanghai, and shares an interview on Latent Space's history, alongside reports of independent replications of frontier models reading and writing small network weights.

Source: Latent Space · 2026-11-04 · Original

6. datasette-llm 0.1a10

datasette-llm 0.1a10 fixes a bug affecting the agent.datasette.io demo site where available LLM models were not correctly restricted to those in the configuration.

Source: Simon Willison's Weblog · 2026-11-04 · Original

7. A thousand fewer paths to root

OpenAI built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel.

Source: OpenAI Blog · 2026-11-04 · Original

8. datasette-apps 0.3a0

datasette-apps 0.3a0 adds support for hosting and editing custom HTML apps inside Datasette and improves Datasette Agent integration with # tagging, new stored-query tools, and user approval for write queries.

Source: Simon Willison's Weblog · 2026-11-04 · 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. Mistral / Meta AI have no stable public feed; X sources are curated by hand. This is a source digest — no commentary from Laojin._