Microsoft offers devs a better way to control AI agent behavior
TechCrunch AI1h agoReleaseMicrosoft is offering a specification for controlling AI agent behavior through portable policy files. Developer, compliance, and security teams can define their own policies for agents to follow. The approach focuses on making organizational rules easier to express and carry across agent deployments, although the provided source excerpt does not describe implementation details or supported environments.
Microsoft’s Project Solara Is an OS for AI Agent Gadgets
The Verge AI2h agoReleaseMicrosoft announced Project Solara at Build 2026, describing it as a platform built for agent-driven experiences. The OS is based on Android rather than Windows, signaling a focus on new device formats beyond traditional PCs. Microsoft demonstrated two concept devices: a desk-oriented concept and a badge-style gadget. The available excerpt does not specify launch timing or technical details.
Beyond LLMs: Why Scalable Enterprise AI Adoption Depends on Agent Logic
Hugging Face Blog1d agoOpinionThe article appears to argue that enterprises need more than LLM capabilities to adopt AI at scale. Its title shifts attention toward agent logic and how AI systems execute tasks in practice. Because the source text was not provided, the specific architecture, evidence, examples, and recommendations cannot be verified.
Qualcomm Unveils Dragonfly Data Center Brand for the Agentic AI Era
INSIDE 硬塞 AI1d agoHardwareAt Computex 2026, Qualcomm described AI agents as a major driver of cross-device hardware upgrades. The company unveiled Dragonfly, a new data center brand focused on inference computing. The announcement outlines a broader strategy spanning endpoint devices and cloud infrastructure, although the source does not provide specifications, performance figures, or deployment timelines.
Jensen Huang: Now Is a Great Time to Start a Software Company
INSIDE 硬塞 AI1d agoCommentaryJensen Huang argues that AI does not spell the end of software companies. Instead, he says this is an excellent time to start one. He also dismisses claims that AI will reduce job opportunities as nonsense. Based on the provided excerpt, the core message is optimistic: AI may create new software opportunities rather than simply eliminate existing businesses and jobs.
Jensen Huang: NVIDIA and Microsoft Aim to Turn PCs into AI Agent Hubs
INSIDE 硬塞 AI1d agoHardwareJensen Huang compared the PC's future to the smartphone's evolution: people still call it a phone, although calling is no longer its primary use. He predicts that PCs will look fundamentally different in ten years, moving beyond today's click-and-type interaction model. The original headline frames this vision as an NVIDIA and Microsoft effort to turn PCs into AI agent hubs.
What happens when companies become too AI-pilled?
TechCrunch AI4d agoCommentaryTechCrunch discusses the danger of companies becoming overly convinced that AI can replace human roles. Box founder Aaron Levie argues that the people making those decisions often understand the jobs least, calling it a form of “AI psychosis.” The piece cites ClickUp cutting 22% of its workforce for AI agents and notes that 2026 tech layoffs are already nearly matching all of 2025.
Does your CEO have AI psychosis? Aaron Levie thinks most of them do.
TechCrunch AI4d agoCommentaryBox founder Aaron Levie calls some executive thinking around AI replacement “AI psychosis.” He argues that the people deciding AI can replace workers are often the least likely to understand what those jobs truly involve. The article frames this against ClickUp cutting 22% of staff for AI agents and 2026 tech layoffs nearly matching all of 2025.
Anthropic Releases Claude Opus 4.8 With Integrity Upgrades and Dynamic Workflows
INSIDE 硬塞 AI4d agoReleaseAnthropic released Claude Opus 4.8 as a rapid iteration focused on stronger integrity and reliability for high-risk tasks. The company also previewed Dynamic Workflows, a feature designed to coordinate multiple agents on large-scale jobs such as code migration. The article mentions Mythos entering a countdown toward unblocking, but does not provide detailed availability or product specifics.
The internet is being rebuilt for machines
TechCrunch AI4d agoBusinessAs AI agents move from experiments into production, internet traffic patterns are expected to shift. AWS, Cloudflare, and others are redesigning cloud infrastructure for a future where machine-generated traffic may dominate over human users. The article frames this as an infrastructure-level change, not a single model or product launch.
The Age of Async Agents — Cognition's Walden Yan & OpenInspect's Cole Murray
Latent Space5d agoCommentaryLatent Space interviews Cognition's Walden Yan and OpenInspect's Cole Murray on the rise of async coding agents. The discussion centers on Devin-related workflows, including 80% Devin commits, spec-to-PR development, full VMs, agent memory, and PMs shipping code. The key theme is not a model release, but a shift toward agents that can work asynchronously inside more complete software delivery loops.
Sesame, the conversational AI startup from Oculus founders, launches iOS app
TechCrunch AI5d agoNew ToolSesame, a conversational AI startup from Oculus founders, has launched a new iOS app for the public. The app brings its AI agents to users with a focus on more natural back-and-forth interactions. Based on the available summary, the product is positioned less like a traditional chatbot and more like talking to a person.
ITBench-AA: Frontier Models Score Below 50% on Enterprise IT Tasks★ 72
Hugging Face Blog6d agoBenchmarkArtificial Analysis and IBM present ITBench-AA, described in the title as the first benchmark for agentic enterprise IT tasks. The headline result is that frontier models score below 50%, suggesting current systems still struggle with enterprise-grade agent workflows. The original article text is unavailable here, so task design, evaluated models, scoring methodology, and rankings cannot be confirmed.
Robinhood Will Let Your AI Agent Trade Stocks and Make or Lose Money
The Verge AI6d agoBusinessRobinhood says traders can create a separate account for an AI agent and fund it with a chosen amount of money. The agent will then be able to buy and sell stocks across the market. The move pushes AI agents beyond advice or research into direct financial action, with real gains and losses possible.
Robinhood now lets your AI agents trade stocks
TechCrunch AI6d agoBusinessRobinhood will allow users to create a separate account with a pre-loaded balance that an AI agent can use to trade stocks. The limited description suggests a structure where agent activity is separated from the user’s main funds. The article does not specify supported agents, risk controls, launch timing, confirmation flows, or eligible assets.
Google Cloud on Moving Enterprise AI Agents from PoC to Production
INSIDE 硬塞 AI6d agoBusinessINSIDE frames enterprise AI through a sharp ROI gap: a 2025 MIT survey said 95% of companies had not seen returns despite massive AI spending. It also cites Gartner’s forecast that Fortune 500 companies may average 150,000 agents by 2028. The article focuses on Google Cloud’s view of how enterprises should prepare for AI agents and allocate IT budgets for real deployment.
Some ideas for what comes next, May 2026
Interconnects (Nathan L.)7d agoCommentaryNathan Lambert argues that 2026 AI progress is becoming higher-stakes, with model capabilities, work patterns, economics, and real-world risks all escalating. He says open models still lack a true Claude Code and Opus 4.5-style agent moment, and Gemini has no clear competitor to Claude Code or Codex yet. The essay also tracks Mythos, American open-model momentum, frontier-lab competition, and mounting intervention from governments and other power structures.
Import AI 458: Reckoning with the future; and a singularity story★ 74
Import AI (Jack Clark)7d agoCommentaryThis Import AI issue is a long essay and fiction piece about living through rapid AI progress. Clark uses personal experience and Anthropic’s internal use of Claude to show work shifting toward delegation, verification, observability, and agent management. He then offers speculative 2026-2028 predictions around biology, autonomous companies, robotics, recursive self-improvement, and a positive singularity story focused on healthcare.
Vercel Sandbox firewall now supports request proxying and filtering
Vercel Changelog22d agoReleaseVercel 宣布其 Sandbox 防火牆現在支援「請求代理」與「過濾」功能。這項更新讓開發者在 Vercel 安全沙盒環境中執行程式碼時，能更精細地控制與監控網路流量。這對於需要執行第三方或 AI 生成程式碼、並防範惡意網路請求（如 SSRF 或資料外洩）的應用程式來說，是一項關鍵的安全提升。
[AINews] The Other vs The Utility
Latent Space28d agoCommentary在 AI 發展的十字路口，業界正對其定位展開深思。一派主張 AI 應如 Clippy 般作為無形、高效的「實用工具」（The Utility），專注於完成任務；另一派則主張 AI 應作為「他者」（The Other），具備獨特的性格與主體性。這場爭論不僅關乎產品設計，更深植於人類如何與非人類智慧共存的哲學思考。
Advanced egress firewall filtering for Vercel Sandbox
Vercel Changelog111d agoReleaseVercel 針對其 Sandbox 環境推出「進階輸出防火牆過濾」功能。此更新讓開發者能精細限制沙盒內程式碼可存取的外部網域、IP 與連接埠，有效防止資料外洩與 SSRF 攻擊。這對於需要安全執行 AI 生成程式碼或第三方腳本的開發團隊來說，是一項關鍵的安全升級。
Vercel Sandboxes are now generally available★ 75
Vercel Changelog123d agoReleaseVercel 宣布 Vercel Sandboxes 正式啟用（GA）。這項功能提供了一個安全、隔離且快速啟動的運行環境，特別適合用於 AI 代理（AI Agents）的程式碼執行、動態原型設計以及不受信任程式碼的運行。開發者可以藉此輕鬆為 AI 應用建構類似 Code Interpreter 的功能，提升應用的互動性與安全性。
Run untrusted code with Vercel Sandbox, now generally available★ 80
Vercel Changelog123d agoReleaseVercel 宣布 Vercel Sandbox 正式全面開放（GA）。此功能專為現代 AI 應用設計，提供安全、隔離且低延遲的執行環境，讓開發者能安全地執行用戶提交或 AI 實時生成的程式碼（如 Code Interpreter 功能）。它與 Vercel 生態系深度整合，免去開發者自行維護複雜沙盒基礎設施的煩惱。
How Sensay went from zero to product in six weeks★ 70
Vercel Changelog126d agoBusinessAI 數位分身新創公司 Sensay 分享了他們如何在六週內將產品推向市場。透過 Vercel 生態系，他們利用 v0 快速生成 UI 組件，並藉由 Vercel AI SDK 輕鬆整合多個大語言模型，實現流暢的 AI 串流對話。Vercel 的 Serverless 部署更讓團隊無需操心基礎設施，專注於產品迭代與用戶體驗。
Filesystem snapshots supported on Vercel Sandboxes★ 70
Vercel Changelog131d agoReleaseVercel 宣布其 Sandboxes（沙盒環境）正式支援檔案系統快照（Filesystem Snapshots）功能。開發者現在可以捕捉沙盒在特定時間點的完整檔案系統狀態，並以此快照快速初始化新的沙盒實例。這項更新能顯著降低冷啟動時間，並為 AI Agent 執行、程式碼測試與動態開發環境提供更靈活的狀態管理。
Import AI 440: Red queen AI; AI regulating AI; o-ring automation ★ 75
Import AI (Jack Clark)141d agoOpinion知名 AI 政策專家 Jack Clark 在最新一期電子報中提出三個核心觀點：首先是「紅皇后 AI」，指出 AI 的攻防與演化正陷入不斷奔跑才能維持原狀的競爭；其次是「AI 監管 AI」，隨著 AI 產出速度超越人類極限，未來必須依賴 AI 進行自動化合規與監管；最後是「O型環自動化」，探討在高度自動化的工作流中，最脆弱的單一環節將決定整個系統的成敗。
Giving your AI a Job Interview★ 80
One Useful Thing (Mollick)202d agoOpinion隨著 AI 提供的決策與建議在工作中變得越來越重要，傳統的簡單測試已不足以評估其極限。華頓商學院教授 Ethan Mollick 指出，我們需要透過結構化的「工作面試」流程，包含情境問答、極限測試與邏輯追問，來評估 AI 在特定任務中的真實實力、潛在偏見與幻覺機率，從而決定如何安全地與其協作。
Introducing x402-mcp: Open protocol payments for MCP tools★ 75
Vercel Changelog263d agoReleaseVercel 宣布推出「x402-mcp」，這是一個針對模型上下文協議（MCP）工具的開放式支付協議。該協議旨在解決 AI Agent 與工具互動時的付費與授權問題，靈感源自 HTTP 402（Payment Required）狀態碼。透過 x402-mcp，開發者可以更輕鬆地為其 MCP 工具整合計費與支付機制，促進 AI 工具生態系的商業化發展。
<script type="text/llms.txt">★ 75
Vercel Changelog286d agoOpinionVercel 發表一項新提案，建議在 HTML 中引入 `<script type="text/llms.txt">` 標籤。此舉延伸了近期流行的 `llms.txt` 規範，讓開發者能將針對 AI 的指令直接寫在特定網頁內。這不僅能讓 AI 爬蟲或 Coding Agent（如 Cursor）更精準地理解網頁內容與 API，還能避免維護單一巨大檔案的麻煩，且完全不影響一般用戶的瀏覽體驗。
The three types of AI bot traffic and how to handle them★ 75
Vercel Changelog293d agoTutorialVercel 官方部落格分析了當前網站面臨的三種主要 AI 機器人（Bot）流量：用於模型訓練的「訓練爬蟲」、用於即時生成回答的「搜尋引擎」，以及代表用戶執行任務的「AI 代理人」。這三者對網站的價值與頻寬消耗各不相同。文章指導開發者如何利用 robots.txt、Vercel 防火牆（Firewall）與 Edge Middleware，針對不同類型的 AI 流量進行精準的允許、限制或阻擋，以在保護智慧財產權與獲取搜尋流量之間取得平衡。