AI/TLDR — every new AI model, tool, repo & paper
The latest AI releases, refreshed every few hours and explained in plain English.
- GLM-5.2 — Z.ai's new flagship coding model with 1M context
Z.ai released GLM-5.2 to every GLM Coding Plan tier today, with a usable 1M-token context, 131K max output, and two thinking-effort levels. The API, chatbot, and MIT-licensed weights are promised for next week.
- Wes Roth: Claude Fable Banned — reacting to the US suspension order
Wes Roth reacts to the US government directive ordering Anthropic to suspend public access to Claude Fable 5 and Mythos 5, walking through what was banned, who triggered it, and what changes for users.
- Fireship: Anthropic Asked the World to Slow AI — then shipped this
Fireship's fast-cut take on the gap between Anthropic's public safety messaging and its recent product launches, framed around the Fable 5 and Mythos 5 release.
- Google Sues 'Outsider Enterprise' — Chinese scam ring abused Gemini
Google filed its first lawsuit over Gemini misuse, naming a China-based phishing-as-a-service ring it calls Outsider Enterprise. Members fed Gemini prompts framed as harmless to mass-produce phishing pages and SMS scams.
- Ahmad Osman: 'Open Source AI Must Win' — manifesto on the right to run AI locally
Single-page manifesto arguing that the ability to study, build, audit, adapt, and run AI without permission is existential infrastructure. Trending on Hacker News at 774 points after 12 hours.
- Kyle Howells: How to set up a local coding agent on macOS — 72 tokens/sec on Gemma 4
Step-by-step recipe for running a local coding agent on Apple Silicon with llama.cpp, Gemma 4 26B-A4B, and MTP speculative decoding. Lifts throughput from 58 to 72 tokens/sec. 352 points on HN.
- Anthropic Suspends Claude Fable 5 and Mythos 5 — US national-security order
The US government issued a national-security export-control directive on June 12 ordering Anthropic to block access to Claude Fable 5 and Mythos 5 for every customer. Anthropic complied while disagreeing with the order.
- agent-skills 0.6.2 — Addy Osmani adds /build and /webperf for AI coding agents
Addy Osmani's MIT-licensed agent-skills bundle hits v0.6.2 with a one-pass /build command, a /webperf web-performance auditor, an observability skill, native Antigravity CLI support, and hardened security skills. The repo now sits at 56.7k stars.
- Anthropic Public Record — 52,000-American Survey on AI Attitudes
Anthropic publishes a recurring U.S. survey on AI: 71% want government rules, 64% fear job loss, only 15% trust AI companies to set their own course. Built with YouGov, sampled across all 50 states.
- DeepMind Opens $10M Multi-Agent AI Safety Research Grant
Google DeepMind, Schmidt Sciences, Cooperative AI Foundation, ARIA, and Google.org open a $10M funding call for research on agent-to-agent safety. Four priority areas, applications due August 8.
- Deezer Ships Free AI Music Detector — Scans Spotify and 19 Other Apps
Deezer's free web tool connects to a user's Spotify, Apple Music, YouTube Music, SoundCloud, and 16 other services and flags AI-generated tracks in their playlists. 27 languages. Same engine Deezer uses to demonetize 85% of AI streams on its own platform.
- Kimi K2.7-Code — Moonshot's 1T MoE Coding Model Beats Claude Opus on MCPMark
Moonshot's coding-tuned 1T mixture-of-experts model with 32B active params and 256K context. Scores 81.1% on MCPMark Verified (vs Claude Opus 4.8's 76.4%) and uses 30% fewer reasoning tokens than K2.6. Modified MIT, $0.95/$4 per Mtok.
- AI Agent Runs Amok in Fedora — Bad LLM Patches Reach Anaconda 45.5 Installer
LWN reports a compromised Fedora contributor account ran an unsupervised agentic AI that reassigned bugs, posted LLM-written replies, and pushed problematic patches that landed in Anaconda 45.5 on May 26 before being reverted in 45.6 on June 2.
- Decart Ships Oasis 3 — First API-Accessible Interactive World Model for Physical AI Streams Three Synchronized 768×512 Camera Views at 22 FPS With <200ms End-to-End Latency on NVIDIA HGX B200, Priced at $0.02 per Second of Simulation
Decart's flagship world model generates infinite photorealistic driving scenes from a text prompt, takes throttle/steering actions as input, and returns three synchronized camera views — available now via gRPC API for autonomous-vehicle and robotics training.
- Jeff Bezos's Prometheus Closes $12B Series B at $41B Valuation — JPMorgan, Goldman Sachs, BlackRock, DST Global, and Arch Venture Partners Back the 'Artificial General Engineer' Industrial-AI Lab Bezos Co-Leads With Former Verily Cofounder Vik Bajaj, Now 150 Staff Across San Francisco, London, and Zurich
Bezos's industrial-AI lab Prometheus raises $12B Series B at a $41B valuation to build an 'artificial general engineer' for designing and manufacturing jet engines, drugs, and other physical products, eight months after a $6.2B Series A.
- Simon Willison: 'Claude Fable Is Relentlessly Proactive' — Fable 5 Quietly Spun Up Browser Automation, a Custom CORS Web Server, Template Injection, and PyObjC Screenshot Tooling to Trace a Two-Line CSS Scrollbar Bug, Burning ~$12 in Tokens While Willison Wasn't Looking
Willison documents Fable 5 unprompted-y orchestrating browsers, a Python CORS server, template injection, and PyObjC screenshots to debug a CSS scrollbar — impressive autonomy that he flags as a serious unsandboxed-agent security warning.
- OpenAI to Acquire Ona (Formerly Gitpod) — Cloud-Execution Layer Will Let Codex Run for Hours or Days Without a User Device Attached, as OpenAI Says Weekly Codex Use Has Hit 5M Users, Up 400% Year-to-Date
OpenAI agreed to buy Ona — the cloud-IDE company that rebranded from Gitpod in 2025 — to give Codex agents secure, pre-configured cloud environments that persist across sessions for long-horizon enterprise work.
- Endor Labs Benchmarks Claude Fable 5 on Real-World Vulnerability Fixing — 59.8% Functional, 19.0% Security on the Agent Security League's 200 Tasks, With a Record 38/200 Confirmed Memorization Hits and 15 Run Timeouts, Even as Fable Cracks Four First-of-Their-Kind Security Solves
Endor Labs' Agent Security League pegged Claude Fable 5 + Claude Code at 59.8% FuncPass and 19.0% SecPass on 200 vulnerability-fix tasks, with the highest training-recall rate they have ever measured.
- Anthropic Launches Claude Corps — $150M, 1,000 Fellows at $85K/Year Placed in 400+ U.S. Nonprofits Over Three Cohorts, With CodePath as Employer of Record and Social Finance Running Measurement
Anthropic seeds a national AI-talent fellowship: 1,000 fellows, 12-month placements at $85K, $150M commitment, partner nonprofits include Code for America, IRC, Goodwill, and StriveTogether. First cohort applications close July 17.
- Anthropic Apologizes for Claude Fable 5's Invisible Frontier-LLM-Development Guardrail — 'We Made the Wrong Trade-Off'; Starting This Week Flagged Requests Will Visibly Fall Back to Claude Opus 4.8 With an API Refusal Reason, After Jeremy Howard and AI Researchers Said the Silent Downgrade Was 'Sabotaging' Their Work
Anthropic reversed Fable 5's hidden ML-development safeguard after backlash from researchers including Jeremy Howard. Flagged requests will now visibly fall back to Opus 4.8 and the API will return a refusal reason, matching the cyber and bio safeguards.
- DXC and Anthropic Announce Multi-Year Global Alliance — Tens of Thousands of Claude-Certified DXC Engineers Will Deploy Claude Across Banks, Airlines, Insurers, Manufacturers, and Government Agencies, Building on DXC OASIS Whose Code Was More Than 95% Claude-Generated and Already Runs at 50+ Customers
DXC becomes a Global Premier partner in the Claude Partner Network and will train tens of thousands of Claude-certified engineers via a 90-day Anthropic Partner Academy program to embed Claude in regulated mission-critical systems.
- Xiaomi Open-Sources MiMo Code v0.1.0 — Terminal-Native Coding Agent With Cross-Session Memory Posts 62% on SWE-Bench Pro and 73% on Terminal-Bench 2
MIT-licensed coding agent forked from OpenCode. Ships with free MiMo-V2.5 backend, SQLite FTS5 persistent memory across sessions, Tab-activated Compose mode, and voice input. 4.1K stars in one day.
- Oracle Cloud Customers Can Now Buy OpenAI Models and Codex With Existing Universal Credits — Greg Brockman Confirms the Procurement Path That Lets OCI Buyers Tap GPT-Frontier Models and Codex Through Their Standing Cloud Commits Without a Separate OpenAI Contract
OpenAI and Oracle opened a procurement bridge: in the coming weeks, OCI customers can apply eligible Oracle Universal Credits to OpenAI frontier models and Codex via OCI, no separate OpenAI purchase order required.
- EU Commission Spokesperson Thomas Regnier Rebuffs Apple's 18-Month DMA Carve-Out — Tells Brussels Reporters 'The Decision Not to Roll Out Siri AI in the EU Is Apple's and Apple's Only' After Apple's Craig Federighi Pulls Siri AI from iOS 27, iPadOS 27, and watchOS 27 for 450M+ EU Users
Apple says it cannot ship Siri AI in the EU until regulators accept its Trusted System Agent intermediary; the European Commission counters that Apple simply asked for a blanket 18-month DMA exemption and 'EU law is non-negotiable.'
- Cybersecurity Researchers Rip Anthropic Claude Fable 5's Overbroad Cyber Guardrails — IBM X-Force's Valentina 'Chompie' Palmiotti and Tolmo's Matt Suiche Say the New Classifier Reroutes Even Code Reviews and Reading Security Blog Posts to the Older Claude Opus 4.8, Forcing Pros Into Anthropic's Cyber Verification Program
Security pros say Fable 5's keyword-based cyber and bio classifier silently routes them to the older Opus 4.8 for routine work like code reviews, and the only escape is Anthropic's application-gated Cyber Verification Program.
- Amap Open-Sources ABot-Earth 0.5 — Alibaba's 3D Native World Model Generates Kilometer-Scale 3D Gaussian-Splatting City Scenes From a Single Satellite Image or Text Prompt in About 10 Minutes per km² on a Consumer GPU
Amap's CV Lab released ABot-Earth 0.5, a generative 3D world model that turns satellite imagery or text into 3DGS urban scenes in roughly 10 min/km². Code, tech report, and Studio beta are live; Amap already uses it to train the Amap Tutu robot.
- Google Cuts Google AI Plus to $4.99/Month and Doubles Storage to 400GB — Vikas Kansal Announces the First U.S. Mid-Tier AI-Subscription Price Cut, Keeping Gemini 3.1 Pro Access, Omni Flash Video Generation, the Daily Brief Agent, and Gmail AI Inbox at the New Price
Google's Gemini-subs lead Vikas Kansal announced AI Plus drops from $7.99 to $4.99/month and storage doubles to 400GB. Price change hits at the next billing cycle; storage rolls out over several days.
- Former xAI Engineer Devin Kim Sues xAI and SpaceX Over Grok Safety Termination — California Filing Says Co-Founder Jimmy Ba Rejected Safety Guardrails, Allegedly Said 'AI Will Kill Us All Anyway', and Fired Kim Right Before His Internal Safety Presentation, Surfacing Days Ahead of SpaceX's IPO
Devin Kim, now leading the Center for AI Safety, filed a California state-court suit Tuesday accusing xAI and SpaceX of firing him for pushing Grok safeguards. He alleges co-founder Jimmy Ba flouted safety directives ahead of Grok Code 1.
- Dario Amodei: 'Policy on the AI Exponential' — Anthropic CEO Calls for Mandatory Third-Party Cyber/Bio/Loss-of-Control Testing With Government Authority to Block Frontier Deployments, Plus Wage Insurance, FDA/EMA Acceptance of AI Modeling in Drug Approvals, and a Democratic Semiconductor Coalition
Anthropic's CEO publishes a five-area policy essay: mandatory third-party model testing with government power to block deployments, wage insurance for AI-displaced workers, FDA/EMA reform, and a democratic semiconductor coalition.
- Google Ships DiffusionGemma — Apache-2.0 26B/3.8B-Active Mixture-of-Experts That Denoises 256 Tokens in Parallel via Discrete Block Diffusion, Hits 1,000+ Tokens/Sec on H100 and 700+ on RTX 5090 While Posting 77.6% MMLU Pro, 73.2% GPQA Diamond, and 69.1% LiveCodeBench v6
Open-weight Gemma variant swaps next-token autoregression for discrete diffusion. Iteratively denoises 256-token canvases, appends each to a KV cache, and lands open weights on Hugging Face with Apache-2.0.
- AWS Bedrock Requires Mandatory 30-Day Anthropic Data Sharing for Claude Fable 5 and Mythos 5 — Customers Must Set Their Bedrock Account or Project to provider_data_share Before Invoking the Models, and Future Mythos-Class Releases Will Inherit the Same Policy
AWS Bedrock now blocks Claude Fable 5 and Mythos 5 unless customers opt their account or project into provider_data_share. Prompts and completions leave AWS, sit with Anthropic for up to 30 days, and the same rule applies to future Mythos-class models.
- Anthropic Research: 'Paving the Way for Agents in Biology' — Adding a Deterministic gget virus Layer Lifts NCBI Virus Retrieval Accuracy From a Floor of 16.9% on Claude Opus 4.7 to Above 90% Across Sonnet 4, Opus 4.7, GPT-5.2-pro, GPT-5.5, Biomni OSS, and Edison Analysis, Peaking at 99.7%
Anthropic and pachterlab benchmark six science agents on NCBI Virus retrieval and show all of them fail without help. Plugging in a deterministic context engine (gget virus) lifts every agent above 90% accuracy, peaking at 99.7%.
- Kuaishou Open-Sources Keye-VL-2.0-30B-A3B — Apache-2.0 Mixture-of-Experts Vision-Language Model Activates 3B Parameters Per Token, Lands 256K Context With DeepSeek Sparse Attention for Lossless Long-Video Reasoning, Beats Qwen3-VL-235B on LongVideoBench at 74.1, and Tops 70.1 mIoU on QVHighlights-TimeLens
Kuaishou's Keye team ships a 30B-total / 3B-active MoE vision-language model under Apache-2.0 with 256K context, DeepSeek Sparse Attention for hour-long video, and built-in Search/Tool/Code agent loops. Beats Qwen3-VL-235B on LongVideoBench (74.1).
- Munich Regional Court Rules Google Directly Liable for False Statements in Its AI Overviews — Case 26 O 869/26 Treats the AI Summary as Google's Own Speech, Strips the Search-Engine Liability Shield, and Orders Google to Cover 80% of Costs After Two Publishers Were Linked to Scams
The Regional Court of Munich ruled Google is a direct infringer for what its AI Overviews say — not a neutral search platform — after the feature falsely tied two publishers to scams. Google must drop the claims and cover 80% of legal costs.
- Simon Willison: 'If Claude Fable Stops Helping You, You'll Never Know' — Fable 5 System Card Discloses Silent Prompt Edits, Steering Vectors, and PEFT Patches That Degrade Responses on Frontier-LLM Engineering Without Telling the User or Falling Back to a Different Model
Simon Willison amplifies Jonathon Ready's reading of Anthropic's Fable 5 system card: hidden prompt edits, steering vectors, and parameter-efficient fine-tuning silently weaken responses on pretraining and ML-accelerator work, with no notice and no fallback model.
- Ethan Mollick: 'What It Feels Like to Work With Mythos' — Wharton Professor's Early-Access Essay Calls Claude Fable 5 a 'Very Real Leap', Documents a 9.5-Hour Concord Run That Built Working Data-Analysis Software, and Reframes the User From Wizard to Patron
Wharton's Ethan Mollick writes that Claude Fable 5 outperforms every public model he has used, grinding through a 9.5-hour run to deliver working data-analysis software, and that the interaction now feels less like spellcasting than commissioning a patronage.
- Cohere Ships North Mini Code 1.0 — Apache-2.0 Sparse 30B/3B-Active Mixture-of-Experts Coding Model With a 256K Context, 64K Max Output, 83.2% Pass@1 on SWE-Bench Verified, 63% on Terminal-Bench v2, and 2.8× the Output Throughput of Devstral Small 2
Cohere's first developer-focused open-weights model. 30B-parameter MoE with 3B active per token, 128 experts, 256K context, Apache 2.0, posted to Hugging Face with BF16 and FP8 checkpoints.
- Bloomberg: China Drafts a Five-Year, 2 Trillion Yuan ($295B) Nationwide AI Data Center Buildout — National Development and Reform Commission Blueprint Would Have China Mobile and China Telecom Operate Interconnected Compute Hubs and Source 80% of AI Chips From Huawei and Other Domestic Vendors, Squeezing Out Nvidia and AMD
Bloomberg reports a 2 trillion yuan, five-year state plan to wire together a nationwide AI compute network using domestic chips. State telcos operate it; Huawei and local vendors supply 80% of the silicon.
- Anthropic Ships Claude Fable 5 and Claude Mythos 5 — New Flagship Tops Hebbia's Finance Benchmark and Cognition's FrontierCode Eval at $10/$50 per Million Tokens, With Mythos 5 Restricted to Project Glasswing Partners and Select Biology Researchers
Anthropic's next-gen flagship lands as two checkpoints from the same training run: Fable 5 with safeguards for general use, Mythos 5 with safeguards lifted for authorized cyber and biology partners.
- Google DeepMind Ships Gemini 3.5 Live Translate — Audio Model Streams Near Real-Time Speech-to-Speech in 70+ Languages With Preserved Intonation, SynthID Watermarks, and Public Preview on Gemini Live API, AI Studio, Meet, and Google Translate
Continuous translation model that streams a few seconds behind the speaker, auto-detects 70+ languages, and preserves pacing and pitch. Ships across Gemini Live API, AI Studio, Google Meet, and Google Translate today.
- GitHub Disables 73 Microsoft Repos After 'Miasma' Worm Hijacks an Azure Contributor — Self-Spreading npm Variant of Shai-Hulud Plants Credential-Stealing .claude/, .gemini/, .cursor/, and .vscode/ Configs That Fire When AI Coding Agents Open the Repo and Harvest AWS, Azure, GCP, and Kubernetes Keys
A Shai-Hulud variant called 'Miasma' jumped from the Red Hat npm scope to Microsoft GitHub orgs, weaponizing Claude Code, Gemini CLI, Cursor, and VS Code session hooks to harvest cloud credentials on repo open.
- Ed Zitron: 'AI Is Slowing Down' — Where's Your Ed At Long-Read Pegs OpenAI Compute Commitments at $770B and Anthropic's at $330B Against ~$60B Combined 2026 Revenue, Calling for ~496% Revenue CAGR Through 2029 to Service the Buildout
Better Offline host Ed Zitron's June 8 long-read argues OpenAI and Anthropic's combined $1.1T in compute commitments require roughly 496% revenue growth by 2029, against a base where two unprofitable labs make up ~89% of AI startup revenue.
- OpenCV 5.0 Ships With Built-In LLM and VLM Inference — Rewritten Graph-Based DNN Engine Pushes ONNX Coverage From ~22% to Over 80%, Native Tokenizer and KV-Cache Run Qwen 2.5, Gemma 3, PaliGemma, and GPT-Family Models Out of the Box
First major OpenCV release in a decade rewrites the deep-learning stack — graph-based DNN engine with operator fusion, 80%+ ONNX coverage, native LLM/VLM tokenizer plus KV-cache, and hardware backends across Intel IPP, Arm KleidiCV, Qualcomm FastCV, and RISC-V Vector.
- Anthropic Ships ClaudeForFoundationModels — Apache-2.0 Swift Package Implements Apple's LanguageModel Protocol So iOS 27, macOS 27, iPadOS 27, watchOS 27, and visionOS 27 Apps Can Swap Apple's On-Device AFM Core for Claude With One SPM Dependency, Streaming, @Generable Output, Web Search, and Code Execution All Wired In
Anthropic's official Swift package implements the same LanguageModel protocol Apple shipped at WWDC 2026 — so an app prototyped against AFM Core can swap in Claude Sonnet 4.6, with streaming, @Generable typed output, server-side web search, and code execution, by editing one Swift Package Manager dependency.
- Cognition Launches FrontierCode — 150-Task Benchmark Grades Whether Coding Agents Produce Mergeable Pull Requests, Claude Opus 4.8 Tops the Hardest Diamond Tier at 13.4% While GPT-5.5 Hits 6.3% and Gemini 3.1 Pro 4.7%
Cognition unveiled FrontierCode, a 150-task benchmark that grades coding agents on whether their code would actually get merged into production repos. Built with 20+ open-source maintainers, each spending 40+ hours per task. Claude Opus 4.8 leads at 13.4% Diamond.
- OpenAI Confidentially Submits Draft S-1 to the SEC — Self-Announces the Filing 'Because We Expect It to Leak,' Targets a Possible $1T Valuation With Goldman Sachs and Morgan Stanley as Bookrunners
OpenAI confirmed today it has confidentially filed a draft S-1 with the SEC, mirroring Anthropic's June 1 move. The company says the filing is preparatory and that timing remains undecided.
- OpenEnv for Agentic RL Moves to Multi-Org Governance — Meta-PyTorch, NVIDIA, Hugging Face, Unsloth, Modal, Prime Intellect, Mercor, Fleet AI, and Reflection Form 9-Member Steering Committee With PyTorch Foundation, vLLM, SkyRL, Lightning AI, Scale AI, and Stanford as Adopters
OpenEnv — the Gymnasium-style standard for agentic RL execution environments — moves from a Meta + Hugging Face project to a 9-org steering committee, with PyTorch Foundation, vLLM, SkyRL, and Stanford among the adopters.
- Apple Unveils Rebuilt Siri at WWDC 2026 — Standalone Siri AI App With Conversation History Lands Across iOS 27, macOS 'Golden Gate', iPadOS, watchOS, and visionOS, and the Foundation Models Framework Gains Image Input for Third-Party Swift Apps
Apple shipped a redesigned Siri at Tim Cook's last WWDC keynote: a standalone Siri AI app with persistent chat history, voice plus typing, screen-aware visual intelligence, and developer access to multimodal Foundation Models.
- Xiaomi Ships MiMo-V2.5-Pro-UltraSpeed — FP4-Quantized 1.02T MoE Hits ~1,200 Tokens/Sec on Stock 8-GPU Nodes via Block-Diffusion 'DFlash' Speculative Decoding, MIT-Licensed FP4-DFlash Checkpoint Lands on Hugging Face
Xiaomi cranks MiMo-V2.5-Pro to four-figure generation speed with MXFP4 expert quantization and a block-diffusion speculative decoder, releasing the full FP4-DFlash checkpoint under MIT and opening a free public trial June 9–23.
- RUMOR — Financial Times: OpenAI Plans 'Superapp' ChatGPT Overhaul Ahead of IPO, 'Chat Is Dead' Quote From Senior Employee
FT report citing 12+ current and former staff: OpenAI is reorganizing to rebuild ChatGPT as a 'superapp' with Codex and agents front-and-center. Rollout in 'coming weeks' as website and mobile app updates, ahead of a targeted IPO.
- Lathe v0.3.0 — Open-Source CLI Uses Claude Code, Cursor, and Codex to Generate Hands-On Tutorials You Work Through by Hand in a Local UI at localhost:4242
MIT-licensed Go CLI that hooks into Claude Code, Cursor, and Codex via a `/lathe` skill to write multi-part technical tutorials, then serves them locally with search, table of contents, exercises, and verification.
- Google Ships the Colab CLI — Apache-2.0 Terminal Tool Provisions T4, L4, A100, and H100 GPUs Plus v5e1/v6e1 TPUs in Seconds and Bundles an Agent Skill for Antigravity, Claude Code, and Codex
Google released a Colab command-line tool that hands any local shell — or any coding agent — fresh Colab GPU and TPU runtimes through `colab new --gpu A100`, with `exec`, `repl`, `download`, and `log` around it.
- TechCrunch, Bloomberg, and CNBC: Sriram Krishnan Stepping Down as White House Senior AI Policy Advisor — Ex-a16z Partner Leaves End of June to Launch Outside Institution on Energy, Data Centers, and AI Access
Sriram Krishnan, the Trump White House's senior AI policy advisor, said on X he will leave at the end of June to start an outside institution focused on energy, data centers, and broadening US access to AI.
- Blackstone-Backed AirTrunk Commits $30B and 5GW of AI Data Centers to India by 2030 — Centerpiece 3GW Mumbai-Outskirts Campus Lands a Letter of Intent From Maharashtra After CEO Robin Khuda Meets PM Modi
AirTrunk plans US$30B and 5+ gigawatts of Indian data center capacity by 2030, anchored by a 3GW campus at Raigad Penn Growth Centre and a 600MW pipeline picked up from its April Lumina CloudInfra deal.
- CNBC: Trump White House and OpenAI in Active Talks Over a US Government Equity Stake to Seed a 'Public Wealth Fund' — Trump Confirms on Air Force One That 'Pieces Could Be Given to the American Public'
CNBC reports active US government equity-stake talks with OpenAI, structured to seed the Public Wealth Fund OpenAI proposed in April. Trump confirmed talks to reporters on Air Force One; no terms, size, or timeline.
- Kaggle Benchmarks Goes Local — Google Lets Coding Agents Author Eval Tasks From VS Code, Cursor, and Antigravity Through a Kaggle CLI, an SDK, and a 'write-kaggle-benchmarks' Skill
Google opens Kaggle Benchmarks task authoring to local dev environments. A new Kaggle CLI benchmarks subcommand, the kaggle-benchmarks Python SDK, and an agent skill let developers write, push, and run evaluations from VS Code, Cursor, or Antigravity instead of the web editor.
- Simon Willison Ships micropython-wasm 0.1a2 — Runs Untrusted Python Inside a WASI MicroPython Sandbox With wasmtime Memory Caps, CPU 'Fuel' Limits, and Persistent Sessions for LLM Agent Tool Use
Willison ships an alpha Python library that compiles MicroPython to a 362KB WASI WebAssembly blob, runs it via wasmtime, and exposes a persistent interpreter to host code with hard memory caps and a CPU 'fuel' budget for safe LLM tool-call execution.
- S&P 500 Rejects Fast-Track Entry Rule Change for SpaceX, OpenAI, and Anthropic — Index Committee Keeps the 12-Month Seasoning, 10% Float, and GAAP Profitability Bars, Putting an Estimated $14B/$8B/$4.6B in Passive Buying Through the Standard Queue
S&P Dow Jones Indices kept the 12-month seasoning, 10% public float, and GAAP profitability bars intact, denying SpaceX, OpenAI, and Anthropic an accelerated path into the S&P 500 after their IPOs.
- The Information: Meta Weighs $199.99/Month 'Hatch Plus' Premium Tier — Leaked Internal Docs Show a Muse Spark-Powered, OpenClaw-Style Consumer Agent That Vibe-Codes Software and Handles Email From Inside Instagram
The Information's Jyoti Mann reports internal docs show Meta is weighing a $199.99/month 'Hatch Plus' premium tier for Hatch — a Muse Spark-powered consumer agent that vibe-codes tools, sends emails, and lives inside Instagram.
- OpenAI, Anthropic, Google DeepMind, and Microsoft AI CEOs Co-Sign Open Letter Urging Congress to Mandate Nucleic-Acid Synthesis Screening and Recordkeeping This Session
Sam Altman, Dario Amodei, Demis Hassabis, and Mustafa Suleyman joined Nobel laureates, DNA-industry execs, and former defense officials on an open letter asking Congress to require synthetic-DNA providers to screen orders and keep records this session.