今日辑要 · 2026-10-18（前沿 12 条）

今日从公开源汇得 12 条与人工智能、智能体系统、链上机制与数字经济相关的条目，按方向归拢如下。所有条目均保留来源与原文链接，只作线索与机制观察，不代表任何结论。

AI 全球趋势

• Ask HN: Non-deterministic leet code small LLM challenge site?

Hacker News · 23:52

[查看原文](https://news.ycombinator.com/item?id=50132166)

• Why Do Chinese LLMs Switch to Chinese in Complex Interactions? (2024)

Hacker News · 23:43

[查看原文](https://pub.towardsai.net/why-do-chinese-llms-switch-to-chinese-in-complex-interactions-d18daac872b8)

• Are You Sure Your LLM Is Sure?

Hacker News · 21:39

[查看原文](https://arxiv.org/abs/2610.22139)

• Prior-Data Gravity and Framework Reabsorption: A Behavioral Theory of LLMs

Hacker News · 21:35

[查看原文](https://zenodo.org/records/23431917)

• Show HN: Vramkit – lossless VRAM compression for LLM inference, on a MacBook

Hacker News · 21:32

[查看原文](https://github.com/SRKRZ23/vramkit)

• Ask HN: How much are developers trusting LLMs to write prod code?

Hacker News · 20:53

[查看原文](https://news.ycombinator.com/item?id=50130686)

Agent 与智能体

• Websites are closing their doors to AI agents

Hacker News · 23:47

[查看原文](https://www.theverge.com/ai-artificial-intelligence/1013195/ai-agents-blocked-web)

• FBI awarding UFC exclusive agreement to train agents

Hacker News · 23:32

[查看原文](https://www.reuters.com/sports/report-fbi-awarding-ufc-exclusive-agreement-train-agents--flm-2026-10-10/)

• Proactive AI for Superintelligent Agents

Hacker News · 22:44

[查看原文](https://research.google/pubs/proactive-ai-for-superintelligent-agents/)

• morluto/rea — Reverse engineer anything with agents, from app behavior down to native binaries.

GitHub 新星 · 15:57

[查看原文](https://github.com/morluto/rea)

• anomalyco/opencode — The open source coding agent.

GitHub 新星 · 15:54

[查看原文](https://github.com/anomalyco/opencode)

观察

把条目按方向摊开看，能看出两类不同的进展速度：一类是能力侧（模型、工具使用、评测方法），更新快、噪声也大；一类是机制侧（治理、共识、支付与监管），更新慢，但一旦成型影响更长久。

对做系统的人来说，前者要跟得勤，后者要看得远：跟得勤是为了不落伍，看得远是为了不白忙。

---

本文由本站编辑台自动汇编，条目版权归原作者与来源站点所有；译文与摘要仅供速览，请以原文为准。