AI Daily Briefing
Every public feed is filtered for AI relevance, reports of the same story are grouped, and the day's top events by independent sources are picked.
2026-10-20
Content is auto-organized by AI from public feeds for reference only. Verify important facts at the original source.
Today's attention is on the tooling around AI agents. Google open-sourced Harness², which supports both OpenCode and Codex as backends, while SandboxAQ released tools for agent security testing and cost tracking. StartLux also introduced a 27B decision model that topped the Hugging Face decision leaderboard. The notable thread: evaluation, safety and cost control for agents are emerging as a distinct tooling layer.
2026-10-20 AI briefing
- NO.0195Flat
Google open-sources Harness² with OpenCode and Codex backends
Google Research has open-sourced Harness² on GitHub under Apache 2.0, turning eight kinds of external components — system prompts, guardrails, memory, tool guidance, sub-agents and Skills — into pluggable modules so agents can improve recursively without touching model weights. The framework supports both OpenCode and Codex backends, defaults to OpenCode, and Google stresses it remains a research project rather than a finished product.
- Google Research open-sources Harness², an agent framework supporting OpenCode and CodexInfoQ 中文 · 10/20, 02:22
- Google Open Sources Harness2 for Self-Improving AI AgentsReddit · LocalLLaMA · 10/19, 06:33
- Google open-sources Harness2, a framework that rewrites an agent's harness at test timeMarkTechPost · 10/18, 16:15
- NO.02New95New
SandboxAQ launches Switch Trust and Switch ROI for AI agents
SandboxAQ added two products to its Switch AI platform. Switch Trust inventories the agents, models and tools running inside an enterprise, then probes them with attacks such as prompt injection and blocks or masks leaked credentials and personal data, while Switch ROI maps each agent's spend to the results it produces. Both are free for up to five agents.
TThe Next Web1 source·1 report
- SandboxAQ launches Switch Trust and Switch ROI for AI agentsThe Next Web · 10/20, 00:06
- NO.03New0New
StartLux releases 27B decision model, tops Hugging Face Decision Index
Chinese team StartLux released a 27B decision model that scored 69.12 on the Hugging Face Decision Index 0.3.2, updated October 16, ranking first among 174 entries and leading all three sub-tracks: public benchmarks, private tests on similar capabilities, and private tests in new domains. Rather than free-form text, the model outputs a probability for each option and offers Instant and AutoThink inference modes.
IInfoQ 中文1 source·1 report
- StartLux releases 27B decision model, tops Hugging Face Decision IndexInfoQ 中文 · 10/20, 04:33
How is this different from the industry updates board?About the briefingHide
The hot board is a rolling 48-hour window that keeps changing; the daily briefing is frozen once a day so you can look back at what happened on a given date. Both use the same events and the same heat rule: each independent source counts once, decaying with a 24-hour half-life.Open the industry updates board →
Summaries come from public feeds and remain the property of the original authors; verify important facts at the source. If you believe a summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion