Monday, November 2, 2026
TL;DR
Anthropic quietly began blocking Claude Code from recreating specific existing video games, confirmed as IP enforcement with no opt-out. OpenAI opened an election-operations war room ahead of Tuesday's midterm vote count. Anthropic's S-1 shows revenue up roughly 14x on 85% gross margins.
Act on this
- If your public repo uses pull_request_target, audit for Wiz's pwn-request bypasses: script injection and direct git clone still work after GitHub's Dec 2 default fix. Wiz, The Pwn Request Reckoning #
Signals
Offline, fully local coding agents are starting to do real production work, not just pass benchmarks. #
Thomson Reuters engineer Rittika Jindal's Oct 27 Latent Space post fixed a real Decimal-rounding bug with Wi-Fi off, using OpenCode and Ollama to run a local Qwen3.5 9B model end to end - a demonstrated result, not a benchmark score. Rittika Jindal, Coding From the Himalayas With an Offline AI Agent
A quiet week beyond Jindal's offline-agent demo - this week's other practitioner output, from Willison, Zvi Mowshowitz and Gergely Orosz, mostly extended stories already filed in recent days.
News
WATCHOpenAI opens an election operations center ahead of Tuesday's midterm vote #
Hundreds of staff across teams, led by civic-risk chief Josh Lawson, will watch for misuse or malfunction of OpenAI's own products through final vote counting - not general election misinformation elsewhere. Single source, per Politico. AI Weekly, OpenAI opens Election Operations Center
ACTWiz finds GitHub's 2026 anti-"pwn request" fixes leave real bypasses open #
Old checkout SHA pins, direct git clone of fork code, and script injection via untrusted PR titles all still work, even after GitHub's actions/checkout v7 and its Dec 2 default block on pull_request_target in public repos. Wiz, The Pwn Request Reckoning
WATCHGitHub's Octoverse report: AI agents now open 40%+ of pull requests #
Up from 17% in March, per GitHub's 2026 Octoverse data, flagged by Simon Willison. Platform-wide commit volume nearly tripled to 34.3 billion in the trailing year, and agent-touched PRs span more files than human ones, 87% versus 79%. Simon Willison, on GitHub Octoverse 2026
WATCHAnthropic's S-1 shows 85% gross margins and revenue up roughly 14x #
Per Zvi Mowshowitz's Oct 28 read of the S-1, half-year 2026 revenue hit $16.3 billion, roughly 14 times the prior year, with gross margins near 85% - though the September-quarter revenue line is blank in the filing. Zvi Mowshowitz, Anthropic Files Its S-1
WATCHAnthropic's ban on cruelty toward Claude takes effect November 12 #
The usage-policy update bars 'sustained and needless' abuse of Claude, not ordinary criticism, research, or dark fiction. Anthropic says ending the conversation stays the main enforcement tool, not account bans; it stops short of claiming Claude can suffer. Interesting Engineering, Anthropic's new usage policy
The long view
A year ago, Anthropic's most recent valuation was $183 billion, set by a $13 billion Series F in September 2025, and its finances were private, known only through investor leaks. On Oct 28, its S-1 filing showed half-year 2026 revenue of $16.3 billion, roughly 14 times the prior year, with gross margins near 85% - public numbers, not leaked ones, filed ahead of a listing. If this holds, the next year turns Anthropic from a startup whose growth outsiders estimated into a public company whose quarterly numbers everyone can check against its claims.