Anthropic's S-1 filing discloses a $693M operating loss in H1 2026, down from $2.1B in H1 2025

Anthropic's S-1 filing (ticker ANT) discloses a $693M operating loss in H1 2026, down from $2.1B in H1 2025, confirming the $80.7B net loss is mostly a non-cash accounting charge tied to convertible debt — resolving this page's own pre-filing $8.1B operating-loss estimate.

Anthropic's S-1 shows more than half of its $100,000+ customers now use both the Claude Platform API and a first-party app, up ~30% from last year, and reports ~900% net revenue retention — though analysts flag the NRR figure excludes Pro/Max subscribers and measures consumption growth, not locked-in contracts.

Of today's 14-item GitHub Trending capture, 13 are agent, harness, or skill tooling — up from 7/11 yesterday and 8/10 on 10-25, breaking the 'stable 70-80% plateau' hypothesis this page raised just one day earlier.

ChatGPT's Google Play top-30 absence reaches a third consecutive day, crossing this page's own stated bar for confirming a genuine ranking-change thread independent of Claude's; Claude's own absence reaches eleven days.

Tuesday's voice rotation: destined/hurston-1 ships under Apache-2.0 — the cleanest permissive-license pass this page has logged — but remains synthesis-only, failing the standing full-duplex capability test for the twelfth consecutive rotation since 08-11.

debpalash/VoiceStudio posts its highest-ever daily GitHub star gain (2,297, total 89,809) yet falls from #3 to #11 on Trending — a second instance this week of rank and star velocity diverging, the same pattern first isolated for cloudflare/security-audit-skill on 10-26.

Three independently-published decensored or quantized variants of deepseek-ai/DeepSeek-V4.1-Flash-1017 (today's #1 Hugging Face Trending model) appear in today's top 20, alongside an HN post naming five inference providers now offering an abliterated version — at least one reportedly requiring KYC before access.

multica-ai/andrej-karpathy-skills — a single community-built CLAUDE.md derived from Andrej Karpathy's LLM-coding-pitfall observations — debuts in this page's GitHub Trending captures with 224,885 total stars, comparable to affaan-m/ECC's 290,064.