跳到正文

全部動態

RSS
今日 33 則
11月11日週三
  1. DAVID888 Daily 每日放送🎙️ Podcast58

    [2026-11-11] Fireworks資安淪陷重創Notion、.NET 11重構Runtime Async

    詳細網頁版與參考連結:https://podcast.david888.com/post/2026-11-11 Notion因Fireworks AI漏洞導致部分對話外洩,突顯第三方AI供應鏈風險,專家呼籲應在Proxy層落實DLP防護。同時,社群反思過度依賴AI程式設計將加劇系統脆弱性,促使Addy Osmani推出「agent-skills」以工程紀律約束Agent。此外,.NET 11重構非同步架構;AMD Zen 6c擴大亂序執行;Blender 5.3原生支援高斯潑濺。科技狂飆下,穩健度與治理仍是關鍵。 相關連結: Fireworks AI security breach impacts Notion and their customers What's new in the .NET 11 runtime Exact – a new cross platform UI framework Ultorg: The User Interface for Relational Databases AI usage generates a fragile codebase Show H

  2. Rohan Paul39

    遞迴自我改進(RSI)的安全性仍無法定論,因為目前缺乏清楚的方法判斷它會帶來什麼結果。 風險與安全條件:AI 公司已開始追求 RSI,但對何時才算安全尚無共識;安全工作坊的專家也提出了不同條件。 賽局分析:作者以雙人賽局指出,RSI 競賽只在部分風險程度下構成囚徒困境;風險高時,雙方都暫緩行動可以是穩定結果。

  3. Lex Fridman Podcast🎙️ Podcast58

    #504 – Russia-Ukraine War: Putin, Zelenskyy, Drones and the Future of Warfare

    Preston Stewart is a West Point graduate and Afghanistan veteran who covers the Ukraine war extensively on his YouTube channel, his podcast called War and Conflict, and his newsletter called Defense Tech News. Michael Kofman is a world-class military analyst specializing in the Russian and Ukrainian militaries. He is a senior fellow at the Carnegie Endowment and he hosts The Russia Contingency podcast on the War on the Rocks platform. Thank you for listening ❤ Check out our sponsors: https://lex

  4. Bloomberg Businessweek🎙️ Podcast58

    Anthropic Targets $1.4 Trillion Valuation in Blockbuster IPO

    The people, companies and trends shaping the global economy. Watch Carol and Tim LIVE every day on YouTube: http://bit.ly/3vTiACF Anthropic is seeking a valuation of about $1.4 trillion in its initial public offering, a bid to make history as the first frontier artificial intelligence lab to go public. The Claude chatbot maker and its co-founders are looking to raise as much as $90.4 billion in the IPO and a private placement. The base offering would raise up to $64.4 billion, with shares market

  5. Latent Space66

    為AI 助理設計:Zeno Rocha 談 Resend

    Resend共同創辦人暨 CEO Zeno Rocha 在訪談中談到,如何讓開發者產品更容易被AI 助理推薦、理解與操作。 - 推薦研究:一項 AEO 研究中,Claude Code 在研究者整理出的 102 次電子郵件服務推薦裡,有 64 次首推 Resend;另一項較全面的研究則顯示,Postmark 在多數模型中排名靠前。

  6. Hacker News (official RSS)39

    Claude 帶來的自動化快感與失控感

    Ivan 用 Claude 打造電子郵件、行事曆與檔案自動化系統,接著讓 Claude 協助撰寫、測試並提交程式碼,工作產出大幅提升;但他也逐漸沉迷於持續建置,犧牲睡眠與人際關係,並開始開發工具來提升 Claude 的效能。

  7. Dark Reading (Information Security)52

    Jaya Baloo 談 AI 熱潮中的資安風險與過度擬人化

    AI 越來越像真人,讓人容易把它當成有想法、有道德判斷的物件;企業也可能在還不瞭解風險時,就給 AI 太多資料和許可權。Aisle 的 CISO Jaya Baloo 在訪談中提醒,AI 確實能協助找出漏洞,也可能本身有漏洞,但不必因此陷入末日式恐慌。她建議做好許可權控管與多層防護,讓人審核 AI 提交的程式碼變更,並保留觀察 AI 行為的能力。她也主張先弄清楚系統能做什麼、可能造成什麼傷害,再決定要交給它多少工作。

  8. The Verge🎙️ Podcast58

    It only takes a prick to pop a bubble

    Nvidia CEO Jensen Huang admitted to having a relationship with a woman named Alice Huang (no relation). Was she a business partner, or was she a spy sent by the Chinese government to steal secrets from the most powerful person in AI? The Verge’s Liz Lopatto explains everything we know about the scandal, what a spy might want from Nvidia, and what could happen next for both Huangs. Further reading: ⁠The future of AI hinges on… Jensen Huang’s sex life?⁠ ⁠What could China get from Jensen Huang?⁠ ⁠H

  9. Hacker News (official RSS)71

    推理究竟在最佳化什麼目標?

    作者將推理描述為在 token 空間中搜尋答案路徑,指出單條思維鏈不保證收斂至正確答案;RL 訓練提高的是模型找到成功路徑的整體機率。 訓練機制:具可驗證獎勵的 RL 會提高成功路徑的機率,並把搜尋成本攤銷至模型權重。 推論與風險:訓練時較長的思維鏈可能與成功相關,但強制延長單次推理不一定持續提升準確率;不完美的驗證器也可能讓模型學會獎勵駭客行為。

  10. BiotechTV - News🎙️ Podcast58

    STIFEL 2026: Alkermes' new CEO Blair Jackson describes his priorities for the company - which include a heavy focus on the emerging orexin agonist class

    He describes the profile of the company's orexin and its development path (a key trial in Idiopathic Hypersomnia will be reading out soon), which also includes indications like ADHD, MS, and Parkinson's further on down the line. Plus how the acquisition of Avadel has been a fit, and whether he would consider doing additional deals.

  11. Jordi Visser🎙️ Podcast58

    Bitcoin Is Doing Exactly What It Did Before the Last Bull Run

    In today's episode of Milk Road Crypto, John is joined by crypto trader and technical analyst Tim Warren to break down why Bitcoin's current correction looks remarkably similar to early 2023, when BTC retested its 200-day moving average before launching into a new bull market. Tim explains why he's 90% confident Bitcoin's bear market bottom is already in, the critical support levels he's watching between $73,000 and $68,500, and why extreme fear could be creating one of the biggest buying opport

  12. The Cognitive Revolution🎙️ Podcast58

    The Neuron Basis: Goodfire's Curt Tigges on Agentic Interp, Monosemanticity & Reverse Engineering

    Goodfire technical staff member Curt Tigges discusses research revealing that individual MLP neurons in modern language models are far more interpretable and functionally bounded than previously thought. Using an automated agent to test hypotheses across architectures like Qwen3.6-35B-A3B, Tigges explains how modern gated MLPs act as legible conjunctions and control signals capable of extensive dynamic sparsity without lasting benchmark loss. By showing that many internal representations can be