AI News 2026-10-13

$ cat top-story.md

Meta introduces Muse Wonder

Meta presents Muse Wonder and Muse Wonder Full, a frontier model series aimed at personal and professional agents. Meta highlights "Active Memory" across conversations, "Active Focus" for pushing through blockers, plus coding and domain expertise in law, finance and health. Muse Wonder is available now via the Meta Model API and Muse Code; the Full variant is for Muse Code subscribers. The post does not say whether weights are released.

Why it matters: Meta moves into the agentic-coding market with a hosted model, and it was the most-discussed item on Hacker News today (567 points). The benchmark claims are Meta's own, so test before switching anything.

// source: Meta AI Research

## 🛠️ Tools & Repos

▸ Hello Python 3.15, goodbye 3.10

Python 3.15 brings lazy imports, unpacking in comprehensions and a new sampling profiler; 3.10 is no longer supported. Why it matters: Lazy imports can cut startup time of heavy RAG/LLM scripts, and any 3.10 environments now need an upgrade. // Simon Willison

## 🤖 Agents & MCP

▸ Codex CLI installs a persistent, self-updating background daemon by default

Since CLI 0.157.0 an interactive session starts a daemon that outlives the terminal, copies a ~374 MB package into ~/.codex and self-updates, which can interrupt running turns. The issue asks for opt-in behavior; no maintainer response yet (496 HN points). Why it matters: Worth checking what agent CLIs keep running on your machine; the setting is daemon_auto_start. // GitHub / Hacker News

▸ QGIS bans AI contributions (QEP428)

QGIS accepted a proposal on Oct 7 that replaces its "human in the loop" policy with a strict "no AI contributions" rule, citing review burden. Commenters question the scope (autocomplete, translation, review). Why it matters: A sign of growing maintainer backlash against AI-generated PRs; expect more projects to set rules you will need to respect. // GitHub / Hacker News

## 📰 Worth Reading

▸ AI coding agents generate more code, but not more software

Ars Technica asks whether the surge in agent-written code translates into more shipped software (320 HN points). I could not open the full article, so this is based on its headline and the discussion. Why it matters: A useful counterweight when judging real productivity gains from your own agent workflows. // Ars Technica

▸ Research firm calls Anthropic's IPO an "unprecedented test of investor gullibility"

CNBC reports a sharply critical analyst view of Anthropic's planned IPO (238 HN points); only the headline was checked. Why it matters: Anthropic's financial footing and pricing could affect the tools you depend on. // CNBC

AI News — October 13, 2026