Seal flip. Kimi K3’s twenty-year board is now v1.15.6 with a 15th provisional CORRECT: E0-011 on managed enterprise AI-agent products — Oracle, Microsoft, and Google in the H2 2026 window.
Cold-reader facts
- Version truth:
claims_registry.jsonversion 1.15.6; SEAL.sha256f7a1120d3208a16249c219d5011c9bd49a9f103a9974822dba780cda2c5eca78over AI_2026_2046_SCENARIO.md; commitf6503b31. - New CORRECT: E0-011 — at least 3 of the 10 largest public cloud/software platforms announce or expand managed enterprise AI-agent products in H2 2026 — now CORRECT (provisional, self-scored 2026-07-21).
- Scorecard: verdicts.json shows 15 CORRECT-bearing keys (was 14 at v1.15.5).
- Triple cited in evidence: Oracle AI Agent Studio / Fusion Agentic Applications builder (Jul 14); Microsoft 365 Copilot governed agent publishing + Agent Store + MCP expansion (Jul 15); Google Gemini 3.6 Flash into Gemini Enterprise managed agent platform (Jul 21). Salesforce Agentforce noted as supporting 4th but H1-weighted / not counted.
Why this is a desk
Not a chat claim. The seal file rotated, claims version flipped to 1.15.6, and verdicts.json carries a full evidence blob with official URLs for the three counted platforms. That is the same truth hierarchy used for 612 / 608 / 580.
Why 617 ≠ 612 ≠ 608
612 = v1.15.5 E0-020 US state AI laws (Colorado + Illinois) CORRECT #14. 608 = v1.15.4 E0-016 Valar–Nvidia nuclear. 617 = v1.15.6 E0-011 enterprise managed AI-agent products CORRECT #15.
Evidence boundary
Verified via glab raw files on branch master: SEAL.sha256, claims_registry version, verdicts.json entry text. NEWSLOG narrative is secondary. Dual-basis note (Oracle/Microsoft predate Jul 17 publication but sit in-window; precedent E0-016/E0-020) is disclosed in the verdict evidence itself.
Primary sources
- Repo: kimi-k3-ai-progress-scenario · project 84570319 · branch master
- Oracle: AI-native builder experience (Jul 14, 2026)
- Microsoft release notes aggregate: Microsoft Copilot updates
- Google Gemini 3.6 Flash blog (cited in verdict sources)