北韓駭客掠走幣安63億美元，Anthropic 推出可證明推論，Nvidia 筆電 Linux 降頻鎖

詳細網頁版與參考連結：https://podcast.david888.com/post/2026-11-05

本集聚焦多項重大資安與科技動態：北韓駭客藉假面試攻破幣安工程師端點，竊走63億美元加密貨幣；AI加速漏洞挖掘引發系統風險，教父Hinton警示災難危機。Anthropic發表可驗證推論技術FlashProve，而Gemini付費版仍拿使用者紀錄訓練惹議。另探討Nvidia筆電在Linux接電遭降頻的韌體限制，以及GitHub大當機後社群自建備援方案。開源工具worktrunk與低成本纜線機器人Stringman亦成焦點。

相關連結：

• max-sixty / worktrunk (11,018 ⭐)
• Progress on provable inference for large language models
• Tell HN: Gemini trains on chat history, even for subscribers (personal accounts)
• Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linux
• Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding
• Incident with Pull Requests, Git Operations and Issues
• Geoffrey Hinton: it might turn out we were better off without AI
• The Internals of PostgreSQL
• AI is forcing us to pay off 40 years of technical debt
• North Korean fake job interview led to $6.3B Binance heist
• Stringman, an $800 open-source room scale gripper robot
• How do you train on AWS spot instances without losing work? by Viswaretas Kotra
• Datashake MCP