How we contain Claude across products
Simon Willison's Weblog59 days agoCommentaryAnthropic explains how process sandboxes, VMs, filesystem boundaries, and egress controls limit what Claude agents can access. Claude.ai uses gVisor; local Claude Code uses Seatbelt on macOS and Bubblewrap on Linux; Cowork runs in a full VM. Simon Willison highlights the documentation quality, notes a previously missed file-exfiltration path, and plans to revisit Anthropic's open-source srt tool.
Running Python ASGI Apps in the Browser via Pyodide + a Service Worker
Simon Willison's Weblog59 days agoTutorialSimon Willison demonstrates an experiment for running Python ASGI apps entirely in the browser using Pyodide and a Service Worker. The approach addresses a Datasette Lite limitation: HTML returned through intercepted navigation did not execute script tags, breaking features and plugins. Claude Opus 4.8, used through Claude Code for web, helped explore the implementation. Basic ASGI and Datasette 1.0a31 demos are available.
I Am Retiring from Tech to Live Offline
Simon Willison's Weblog60 days agoEthicsSimon Willison highlights Chad Whitacre’s decision to leave tech and Open Source, framed not as a forum threat but as concrete action. Whitacre describes wanting to become “AI Amish” or “Internet Amish,” moving toward an offline, analog life closer to 1980 than 1780. A previous post about using Claude Code with Opus 4.5 shows how agentic AI felt intoxicating and unsettling enough to push him away from technological accelerationism.
Quoting Daniel Jalkut
Simon Willison's Weblog60 days agoOpinionSimon Willison quotes Daniel Jalkut’s short comment on the polarized AI debate. Jalkut argues that people against AI are often too against it, while people for AI are often too for it. The post is not a technical update, but a concise opinion pointing to the need for more balanced, less tribal evaluation of AI’s benefits and harms.
GitHub Copilot’s new token-based billing spurs dev backlash★ 72
TechCrunch AI60 days agoBusinessTechCrunch reports that GitHub Copilot will move to token-based billing on June 1, replacing a more predictable flat or request-based model. Some developers say their expected monthly costs could jump dramatically, citing examples from about $29 to nearly $750 or $50 to around $3,000. Others argue the worst cases may reflect heavy vibe-coding usage, while critics say Microsoft encouraged that behavior before changing the economics.
Meta is reportedly developing an AI pendant
TechCrunch AI60 days agoHardwareTechCrunch reports that Meta appears to be making bigger bets on AI-powered hardware, including a reportedly developing AI pendant. The article does not provide confirmed product details, features, pricing, release timing, or model information. The main takeaway is a directional signal that Meta may be exploring more wearable AI hardware form factors.
As browser wars heat up, top Chrome and Safari alternatives in 2026
TechCrunch AI60 days agoCommentaryTechCrunch frames 2026’s browser competition around alternatives to Chrome and Safari. The roundup covers AI-centric browsers like Perplexity Comet, Dia, Opera Neon, OpenAI Atlas, and Aside, alongside privacy-focused options such as Brave, DuckDuckGo, Ladybird, and Vivaldi. It also highlights niche products including Opera Air, SigmaOS, and Zen Browser, showing how browsers are becoming AI assistants, productivity hubs, privacy layers, and wellness-oriented tools.
How one founder’s bet on ‘the old school web’ is paying off
The Verge AI60 days agoBusinessThe Verge profiles Craig Campbell, a former Meta engineer and experienced founder, who chose not to chase AI startup money. After selling his previous e-commerce tool venture in 2022, he instead built a website. The piece frames his decision as a business story about whether the old-school web can still work in the AI and Google Zero era.
Rsync 3.4.3 has hundreds of Claude commits
Hacker News (AI keywords)60 days agoCommentaryThe source is a Hacker News AI-keyword item linking to a Mastodon post titled “Rsync 3.4.3 has hundreds of Claude commits.” No original body text is available, so the only reliable claim is that many commits in Rsync 3.4.3 are described as Claude-related. The exact meaning, review process, quality impact, and author’s stance cannot be confirmed from the title alone.
NVIDIA, Arm and Microsoft Tease a New Era of PC Ahead of GTC Taipei
INSIDE 硬塞 AI60 days agoHardwareNVIDIA, Arm and Microsoft posted coordinated teasers around “A new era of PC,” tied to mysterious coordinates pointing to Taipei. The report frames the move as a pre-COMPUTEX push, with NVIDIA’s rumored N1X Arm chip expected to appear at GTC Taipei. Still, skepticism remains around delays, high pricing, and backlash against overused AI PC messaging.
[AINews] Founders and Forward Deployed Engineers
Latent Space60 days agoCommentaryLatent Space’s AINews notes that it was a quiet AI news day, so the issue highlights new AIE WF focuses. The title points to founders and forward-deployed engineers as the central theme. The available text does not name specific companies, models, tools, launches, papers, or benchmarks, so the takeaway should remain conservative and contextual.
MCP is dead?
Hacker News (AI keywords)60 days agoOpinionQuandri measured MCP tool schemas in its Claude Code setup and found significant context overhead across Linear, Notion, Slack, and Postgres. The post argues MCP can be slower, less reliable, and harder to debug than direct CLI/API usage. It recommends CLI-first workflows and on-demand Skills, while noting MCP still fits services without CLIs, non-developer users, bidirectional communication, and guarded production database access.
Coders are refusing to work without AI — and that could come back to bite them
TechCrunch AI60 days agoCommentaryTechCrunch reports that developers have become so attached to AI coding tools that METR struggled to repeat a no-AI control study. Earlier research found developers felt more productive with AI, while measured task completion could be slower due to debugging, steering, and waiting. The article warns that token usage and code volume are weak productivity proxies if AI-generated code creates more bugs, review work, and long-term maintenance costs.
Show HN: Tiny-vLLM, a C++ and CUDA LLM Inference Engine
Hacker News (AI keywords)61 days agoNew ToolTiny-vLLM is a Show HN project described as a high-performance LLM inference engine implemented in C++ and CUDA. From the provided title alone, the project appears aimed at developers or ML engineers interested in GPU-accelerated local or server-side inference. No further claims about supported models, benchmarks, APIs, licensing, deployment targets, or production readiness are stated in the source.
So you’ve heard these AI terms and nodded along; let’s fix that
TechCrunch AI61 days agoTutorialTechCrunch frames this piece as a glossary for the flood of new AI terms and slang that has followed the rise of AI. It aims to define important words and phrases readers may encounter in coverage, product discussions, or broader industry conversations. Based on the provided text, this is an educational guide rather than a product launch, research paper, or market-moving announcement.
What happens when companies become too AI-pilled?
TechCrunch AI61 days 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.
After Nvidia’s $20B not-aqui-hire, AI chip startup Groq reportedly raising $650M
TechCrunch AI61 days agoHardwareTechCrunch cites Axios reporting that AI chipmaker Groq is seeking $650 million in internal funding. The company is reportedly pivoting from hardware toward AI inference, the stage focused on how models respond to prompts. The report comes after Nvidia’s $20 billion not-aqui-hire, underscoring continued investor attention around AI compute and inference infrastructure.
Cognition’s Scott Wu says AI coding agents shouldn’t replace humans
TechCrunch AI61 days agoCommentaryCognition makes Devin, described by TechCrunch as the first and arguably most successful AI coding agent. Scott Wu says the product is not meant to supplant human programmers. The key takeaway is a positioning statement: AI coding agents are being framed as tools for software work, not as a direct removal of humans from development.
CAPTCHAs can still detect AI agents★ 72
Hacker News (AI keywords)61 days agoPaperRoundtable argues that CAPTCHA image recognition is largely solved, but process-level behavior still separates humans from AI agents. Their CogCAPTCHA30 benchmark combines CAPTCHA with cognitive psychology tasks to test not only outputs, but how answers are produced. Results suggest frontier models like Claude, GPT, and Gemini are not necessarily more humanlike than smaller or cognition-trained models.
Does your CEO have AI psychosis? Aaron Levie thinks most of them do.
TechCrunch AI61 days 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.
Show HN: AISlop, a CLI for catching AI generated code smells
Hacker News (AI keywords)61 days agoNew ToolAISlop appeared on Hacker News as a Show HN project. From the title, it is a command-line tool focused on catching code smells associated with AI-generated code. Without the original article or documentation content, its exact rules, supported languages, accuracy, and workflow integrations cannot be confirmed, but it is relevant to developers using AI coding tools.
Snowflake and AWS Expand AI Infrastructure Bet to $6B Over Five Years★ 72
INSIDE 硬塞 AI61 days agoBusinessSnowflake reported stronger-than-expected results and raised its annual product revenue forecast as enterprise demand grows. The company signed a five-year, $6 billion AI infrastructure agreement with AWS, expanding a previously smaller commitment. It also acquired Natoma to strengthen AI agent governance, positioning itself as a core enterprise AI platform.
Function invocations now billed per unit
Vercel Changelog61 days agoBusinessVercel announced a billing change titled “Function invocations now billed per unit.” Without the full changelog text, the confirmed takeaway is limited to the billing basis for function invocations. Teams using Vercel Functions should review invocation-heavy APIs, background jobs, webhooks, polling, and AI workflows, but should not assume exact pricing or plan impact without checking the official billing details.
Real-Time LLM Inference on Standard GPUs at 3k Tokens/s per Request
Hacker News (AI keywords)61 days agoBenchmarkThe post’s title indicates a performance claim for real-time LLM inference on standard GPUs, reporting 3,000 tokens per second per request. No article body is available, so the underlying model, GPU type, batch size, latency profile, precision, serving stack, and benchmark method are not stated. The item is best treated as an inference-performance benchmark claim rather than a verified deployment guide.
From Map Data to National Security: A Deep Dive into Amap Risks in Taiwan★ 74
INSIDE 硬塞 AI61 days agoRegulationINSIDE examines how China’s Amap has become controversial in Taiwan beyond ordinary mapping or navigation use. The article says its service relies on user data and AI-based inference rather than full official data integrations. That model could send movement traces and behavioral signals back to China, creating risks for hybrid warfare intelligence, influence operations, and Taiwan’s broader governance of map data and digital infrastructure.
Anthropic Releases Claude Opus 4.8 With Integrity Upgrades and Dynamic Workflows
INSIDE 硬塞 AI61 days 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.
Jensen Huang Hosts Taiwan Supply Chain Dinner, Says TSMC Leads Huawei by 10 Years
INSIDE 硬塞 AI61 days agoBusinessNVIDIA CEO Jensen Huang hosted key Taiwanese supply chain partners, with senior leaders from TSMC, Foxconn, and Quanta attending the high-profile dinner. The report frames the event as a signal of Taiwan’s central role in AI hardware, from advanced chips to manufacturing and servers. Huang also said TSMC leads Huawei by 10 years, underscoring the strategic weight of semiconductor capability.
Siri Redesign and Customizable Camera? Biggest iOS 27 Leak Ahead of WWDC 2026
INSIDE 硬塞 AI61 days agoReleaseINSIDE reports that a major iOS 27 leak points to a redesigned Siri experience, potentially arriving as a standalone app rather than only a system voice assistant. The new Siri is said to integrate deeply with Dynamic Island, suggesting a more visible and persistent interaction layer. The headline also mentions camera customization, but the available text does not provide enough detail to confirm how that feature would work.
Protecting against token theft
Vercel Changelog61 days agoTutorialVercel published a post titled “Protecting against token theft,” focused on token security risks and protection. The article body was not provided, so its scope, affected products, attack scenarios, and recommended mitigations cannot be confirmed. Readers should consult the original Vercel page before taking action or attributing specific guidance to the company.
Protecting against inference theft
Vercel Changelog61 days agoCommentaryOnly the title is available, so specific Vercel product changes or implementation steps cannot be confirmed. The topic appears to focus on protecting AI inference resources from unauthorized access, abuse, or cost-draining traffic. For teams deploying AI apps, the practical takeaway is to treat inference endpoints as high-value backend assets requiring access control, monitoring, and abuse prevention.