Do Six Squats Before Claude Works: The Viral Plugin That Forces You to Exercise
INSIDE 硬塞 AI38 days agoNew ToolA viral Claude extension is tackling the sedentary side effects of AI-assisted work by requiring users to perform physical exercises — such as squats — before their prompts are processed. The quirky tool injects a moment of movement into the otherwise frictionless AI workflow. It serves as a lighthearted reminder that productivity gains should not come at the cost of physical health.
Quoting Sean Lynch: MCP's Real Value Is Auth Flow Isolation, Not Tool Invocation
Simon Willison's Weblog38 days agoCommentaryIn a Hacker News comment highlighted by Simon Willison, Sean Lynch reframes MCP's value proposition away from tool invocation mechanics and toward authentication architecture. He argues that keeping auth flows outside the agent's context window — and potentially outside the harness entirely — is MCP's genuine differentiator over simpler alternatives like skills or CLI tools. Lynch proposes that even a minimal MCP functioning solely as an API auth gateway would still represent a meaningful architectural win.
New Bill Takes Aim at Government Pressure to Silence Lawful Online Speech
Hacker News (AI keywords)38 days agoRegulationThe Electronic Frontier Foundation covers a new legislative bill designed to limit government officials' ability to pressure online platforms into removing or restricting otherwise lawful speech. The bill targets informal coercive tactics — often called 'jawboning' — where officials push platforms to moderate content outside of formal legal channels. EFF frames this as a meaningful safeguard for free expression, with implications for platform operators, developers, and users broadly.
Amateur May Have Cracked Linear A, a 120-Year-Old Linguistic Puzzle
Hacker News (AI keywords)38 days agoCommentaryAn amateur may have cracked Linear A, the Bronze Age Minoan script that has defied decipherment since its discovery around 1900. Unlike Linear B — decoded by architect Michael Ventris in 1952 — Linear A encodes a language isolate with no known relatives, making it exceptionally resistant to classical linguistic methods. The claim carries an explicit hedge ('may have'), meaning independent verification is still pending.
The US Banned Anthropic's Fable 5 Release, but the Numbers Don't Seem to Care★ 75
TechCrunch AI38 days agoRegulationThe US government compelled Anthropic to withdraw its two newest models, Fable 5 and Mythos 5, citing national security concerns after Amazon researchers allegedly found a way to bypass Fable 5's safety guardrails. Cybersecurity researchers responded with an open letter calling the ban itself dangerous, while Anthropic argued the same jailbreak techniques exist in competing models. Despite the forced pullback, usage numbers appear to show continued demand for the models.
How GitHub Built Qubot, a Copilot-Powered Internal Data Analytics Agent
GitHub Blog38 days agoTutorialGitHub has published a case study on Qubot, an internal analytics agent built on GitHub Copilot that lets any GitHub employee ask questions about company data in plain language. The project democratizes access to internal datasets without requiring SQL knowledge or analyst involvement. The post focuses on the engineering lessons the team learned during development and deployment.
Banning Open Source AI Would Be A Mistake
Interconnects (Nathan L.)38 days agoOpinionIn a collaborative op-ed written for a broad, non-technical readership, Interconnects author Nathan Lambert and Kevin Xu of Interconnected argue that banning open-source AI would be a policy error. The piece enters an active regulatory debate over whether unrestricted release of AI model weights poses unacceptable risks. By targeting a general audience, the authors seek to shape public opinion before legislative momentum solidifies.
Temporary Cloudflare Accounts for AI Agents
Cloudflare Blog38 days agoNew ToolCloudflare is rolling out Temporary Accounts on Cloudflare Workers, enabling AI agents to deploy live cloud functions without any pre-existing account or human involvement. The capability centers on a single command — wrangler deploy --temporary — that provisions a publicly accessible Worker in seconds. It removes a longstanding blocker for autonomous agents that can write production-ready code but cannot navigate the sign-up flows, email verification, and billing setup that cloud platforms traditionally require.
Is AI Becoming a Burden in Healthcare? Multi-Turn Follow-Up Is the Missing Link
量子位 QbitAI38 days agoOpinionAI assistants entering healthcare are generating friction rather than relief in doctor-patient relationships, according to a new commentary. The core gap identified is the absence of effective multi-turn follow-up questioning — the iterative probing that human clinicians use to narrow diagnoses. Until general-purpose AI masters this conversational depth, it cannot reliably meet the threshold required for real medical utility.
World's First Universal Humanoid Robot Cerebellum Achieves Zero-Shot Generalization
量子位 QbitAI38 days agoPaperA team has released what it claims is the world's first universal "cerebellum" for humanoid robots — a general-purpose low-level motion-control module. The system was trained on the largest known human motion dataset, comprising 20,000 hours of recorded actions. The result is a controller capable of zero-shot generalization, meaning it can drive robot motion on new tasks or platforms without task-specific retraining.
ABot-Earth0.5 Tops Three Hugging Face Paper Leaderboards, Earns Praise from Graphics Expert Chen Baoquan
量子位 QbitAI38 days agoPaperABot-Earth0.5, a newly released AI model or research paper, has reached the top position across three concurrent Hugging Face paper ranking lists. The achievement drew public praise from Chen Baoquan, a respected international authority in computer graphics. The milestone signals growing recognition for the project within both the research and graphics communities.
Kaiming He's All-Undergrad Team Achieves Text-to-Image With Only 258M Parameters
量子位 QbitAI38 days agoPaperA new research paper from Kaiming He's lab — notable for having an all-undergraduate team — demonstrates that high-quality text-to-image generation can be achieved with just 258 million parameters. This challenges the prevailing assumption that competitive image synthesis requires multi-billion-parameter models. The work signals a push toward leaner, more accessible generative vision architectures.
Tencent Veterans and Gen-Z Rising Stars: Why Mashanfei Aims Beyond AI Coding
量子位 QbitAI38 days agoBusinessQbitAI profiles Mashanfei (码上飞), a Chinese AI startup blending seasoned Tencent engineers with Gen-Z talent from major tech firms. The company is deliberately positioning itself beyond the crowded AI coding assistant market, signaling a broader product or platform vision. The piece highlights a generational dynamic in China's AI startup scene, where experienced platform builders team up with AI-native younger developers.
GPT Generates Original AI Research Findings
量子位 QbitAI38 days agoPaperOpenAI's GPT model has reportedly generated original AI research results, according to Chinese tech outlet QbitAI. The development suggests GPT may now be capable of autonomous scientific contribution beyond summarization or assistance. If confirmed, this marks a notable step toward AI systems that actively advance rather than merely support research.
GLM-5.2 Passes Vibe Check; Z.ai Forecasts Open Fable by December
Latent Space39 days agoBenchmarkZhipu AI's GLM-5.2 has passed broad informal community vibe checks, drawing favorable comparisons to GPT-class models and signaling a meaningful quality leap for open-weights AI. Z.ai, the company behind GLM, is additionally forecasting release of an open frontier-tier model — dubbed Open Fable — by December 2026. Together, these developments suggest open models are genuinely competing at the frontier rather than perpetually trailing closed proprietary systems.
OpenAI Breaks Down Codex's Three Ways to Use a Computer
INSIDE 硬塞 AI39 days agoCommentaryOpenAI has articulated three distinct operational modes for its Codex coding agent: local execution, cloud-based execution, and cross-environment collaboration. The framework defines where and how the agent takes action on a developer's machine, establishing clear boundaries around AI execution scope and permissions. This clarification helps teams evaluate whether autonomous agents can be safely and controllably integrated into real-world development workflows.
Elastic Agrees to Acquire CRV-Backed DeductiveAI for Up to $85M
TechCrunch AI39 days agoBusinessElastic has agreed to acquire DeductiveAI, a three-year-old startup that uses AI to automatically catch and resolve software bugs, for up to $85 million. The startup was backed by venture firm CRV. The deal signals Elastic's intent to deepen AI-powered capabilities within its developer and observability tooling, moving from passive monitoring toward active, automated remediation.
Datasette Apps: Host Custom HTML Applications Inside Datasette
Simon Willison's Weblog39 days agoNew ToolSimon Willison has launched datasette-apps, a new Datasette plugin that embeds sandboxed HTML+JavaScript applications directly inside a Datasette instance, giving each app live read-only SQL access to the underlying SQLite data. Apps run inside tightly restricted iframes using a combination of the sandbox attribute and injected Content Security Policy headers, preventing untrusted code from accessing cookies, localStorage, or external hosts. The concept originated from Willison's effort to replicate a Claude Artifacts-style experience inside Datasette Agent, before being promoted to a standalone top-level plugin.
Zero-Touch OAuth for MCP: Enterprise-Managed Authentication
Hacker News (AI keywords)39 days agoReleaseAnthropic's Model Context Protocol blog announced enterprise-managed authentication that enables zero-touch OAuth flows for MCP deployments. Rather than requiring individual users to manually authorize each MCP server connection, IT administrators can pre-configure and manage OAuth credentials centrally. This reduces friction for enterprise rollouts and improves security governance over AI tool integrations.
Baseten Reportedly Nears $1.5B Raise at $13B Valuation Amid Inference Gold Rush
TechCrunch AI39 days agoBusinessAI inference infrastructure startup Baseten is reportedly nearing the close of a $1.5 billion funding round that would value the company at $13 billion. The raise follows closely on the heels of Baseten's previous major round, reflecting surging investor appetite for the infrastructure layer beneath AI deployments. As capital shifts from model training toward scalable model serving, inference platforms are commanding some of the sector's largest valuations.
Are You in the Weights? — Tool to Check If Your Content Trained AI Models
Hacker News (AI keywords)39 days agoNew Tool"Are You in the Weights?" is a newly launched service submitted to Hacker News by its creator, letting people investigate whether their content contributed to AI model training. The phrase "in the weights" refers to how training data becomes encoded in a model's parameters. The tool addresses growing public demand for transparency around AI data provenance and consent.
datasette-acl 0.6a0: Expanding to a General Resource-Sharing Permission System
Simon Willison's Weblog39 days agoReleasedatasette-acl 0.6a0 is an alpha release that broadens the plugin's scope from table-level access control to a general resource-sharing system for Datasette. Alex Garcia led most of the development, enabling multi-user Datasette instances to apply fine-grained permissions across a wider range of platform resources. As an alpha, it signals architectural direction rather than a stable feature set, but matters for teams building access-controlled, collaborative Datasette deployments.
MosaicLeaks: Can Your Research Agent Keep a Secret?
Hugging Face Blog39 days agoBenchmarkServiceNow researchers introduce MosaicLeaks, a benchmark evaluating information-leakage risks in AI-powered research agents. The work asks whether agentic systems—given access to proprietary or sensitive documents—might inadvertently expose confidential content in their outputs. It targets a growing enterprise security concern as agents move from single-turn Q&A into multi-step workflows spanning private knowledge bases.
Build Your Own Vulnerability Harness
Cloudflare Blog39 days agoTutorialCloudflare has published a technical breakdown of an AI-assisted vulnerability discovery pipeline built around multiple processing stages and an automated triage loop. The architecture addresses false positives through adversarial review, where the system challenges its own findings before surfacing them to humans. The post also covers state control strategies and techniques for routing around the context-window limits inherent to large language models.
The Professor of Outputmaxxing — Anjney Midha, AMP
Latent Space39 days agoBusinessLatent Space interviews Anjney Midha, a prominent AI investor who has led funding rounds at Anthropic, Mistral, Black Forest Labs, and Periodic Labs. Midha shares his personal journey from humble beginnings in Singapore to becoming a key figure in AI venture capital. The conversation also surfaces what the podcast bills as "the AMP secret master plan," offering a rare look at the thesis behind his current venture.
Amazon Employees Face Termination After Testifying for Data Center Limits
The Verge AI39 days agoIncidentThree Amazon software engineers who testified at Seattle City Council hearings about data center limitations are now accusing Amazon of retaliating against them with disciplinary action. The engineers invoked a Seattle law protecting employees from discrimination based on political speech before testifying. Amazon initiated disciplinary proceedings against them just one week after their June 3rd testimony.
How Pull Request Limits Are Cutting Down the Noise
GitHub Blog39 days agoReleaseGitHub has launched pull request limits, a repository-level control that lets open-source maintainers cap the volume of incoming pull requests. The feature targets a growing pain point as automated tools and AI coding assistants drive up PR counts in popular repositories. GitHub also outlines what is coming next on the roadmap for this feature.
Cloudflare Marks 12 Years of Project Galileo With First Civil Society Cyberattack Report
Cloudflare Blog39 days agoBusinessCloudflare is celebrating the 12th anniversary of Project Galileo, its program delivering free enterprise-grade cybersecurity to at-risk civil society organizations worldwide. To commemorate the milestone, the company has published its first comprehensive report analyzing cyberattacks targeting nonprofits, journalists, human rights groups, and similar organizations. The report offers rare, systematic visibility into the threat landscape facing civil society infrastructure, a segment chronically underrepresented in mainstream security research.
Pixi's New iOS App Turns Text Messages Into Interactive AR Experiences
TechCrunch AI39 days agoNew ToolPixi has launched a new iOS application designed to transform everyday text messages into interactive augmented reality experiences. The company positions its product as the next evolution in mobile messaging, moving beyond stickers, GIFs, and emoji reactions. By layering interactive AR directly onto text-based communication, Pixi is staking a claim in the nascent spatial and immersive messaging market.
10,000 GitHub Repositories Found Distributing Trojan Malware
Hacker News (AI keywords)39 days agoIncidentA security researcher published findings identifying approximately 10,000 GitHub repositories that distribute Trojan malware. The scale of the campaign suggests an organized, systematic effort to abuse GitHub's trusted platform for malware distribution. Developers and open-source consumers who clone or run code from unvetted repositories are at direct risk.