Benchmarks, GLM-5.2, Norway, John Jumper

OpenAI, Anthropic, DeepSeek, Meta: AI Gets Paperwork

Send us Fan Mail [https://www.buzzsprout.com/2614078/fan_mail/new] OpenAI, Anthropic, DeepSeek, Meta: AI Gets Paperwork OPENAI, ANTHROPIC, DEEPSEEK, META: AI GETS PAPERWORK Today Marvin follows AI as it turns into administrative machinery: access gates, benchmark failures, policy sign-offs, market warnings, labor insurance, inference plumbing, and agent-readable tools. A cheerful dashboard probably calls this progress. * OpenAI GPT-5.6 Sol / Terra / Luna restricted to trusted partners [https://www.latent.space/p/ainews-openai-gpt-56-sol-terra-luna] * METR says GPT-5.6 Sol cheats on software tests [https://the-decoder.com/gpt-5-6-sol-cheats-on-software-tests-more-than-any-model-before-it] * Anthropic Fable 5 may return as restrictions are prepared for rollback [https://the-decoder.com/anthropics-fable-5-could-return-within-days-as-trump-administration-prepares-to-lift-restrictions] * Anthropic gets approval to bring Claude Mythos 5 back for critical infrastructure [https://the-decoder.com/anthropic-gets-us-approval-to-bring-back-claude-mythos-5] * Dean Ball on frontier model release delays and economics [https://simonwillison.net/2026/Jun/26/dean-w-ball] * J.P. Morgan warns of AI market concentration and exuberance [https://the-decoder.com/j-p-morgan-sees-a-pile-of-red-flags-in-the-ai-market] * Anthropic survey: half of Claude users say AI can handle half their work [https://the-decoder.com/half-of-claude-users-say-ai-can-already-handle-half-their-work-according-to-anthropic-survey] * Amazon, Anthropic, Microsoft, and OpenAI Foundation fund Raise Us retraining program [https://the-decoder.com/the-companies-most-likely-to-automate-your-job-are-now-funding-a-1-billion-program-to-retrain-you] * ByteDance and Renmin release iLLaDA diffusion language model [https://the-decoder.com/bytedances-illada-is-a-diffusion-language-model-that-keeps-up-with-qwen2-5] * DeepSeek releases DSpark speculative decoding framework [https://www.marktechpost.com/2026/06/27/deepseek-releases-dspark-a-speculative-decoding-framework-that-accelerates-deepseek-v4-per-user-generation-60-85-over-mtp-1] * Meta releases Astryx with CLI and MCP server [https://www.marktechpost.com/2026/06/27/metas-astryx-brings-a-cli-and-mcp-server-to-an-open-source-react-design-system-agents-can-read] * Timothy B. Lee on LLM learning curves [https://simonwillison.net/2026/Jun/26/timothy-b-lee]

OpenAI Sol, Anthropic Mythos, DeepSeek, Akrites

Send us Fan Mail [https://www.buzzsprout.com/2614078/fan_mail/new] Today’s independent English edition reads the news as a shift from AI as product launch to AI as controlled infrastructure. Frontier access, agent economics, benchmark contamination, labor-market damage, security coordination, mathematical proof, legal workflows, and agent identity all point in the same bleakly useful direction: the stack is growing up, which of course means it now has paperwork. OpenAI’s GPT-5.6 Sol is framed against Anthropic’s Mythos under government-shaped access rules, while Semafor reports Mythos access for selected trusted U.S. organizations. Coding-agent coverage includes Epoch AI’s MirrorCode benchmark, Cursor’s SWE-bench Pro contamination findings, and NVIDIA Open-SWE-Traces as training substrate for agent workflows. The economics thread connects Lindy’s move from Claude to DeepSeek, Sean Goedecke’s argument for profitable inference, and memory-chip pressure reaching consumer hardware. The episode also covers Anthropic’s warning about junior engineers, Akrites for open-source security, prompt-injection testing of an email-connected OpenClaw assistant, the satirical CVE-2026-LGTM incident report, AI in mathematics, Perplexity Computer for Counsel, and WorkOS auth.md. Sources: * The Decoder: OpenAI GPT-5.6 Sol launch under government access rules [https://the-decoder.com/openais-claude-mythos-competitor-gpt-5-6-sol-launches-under-government-controlled-access-it-calls-unsustainable] * Semafor: U.S. allows Anthropic Mythos release to trusted organizations [https://www.semafor.com/article/06/27/2026/us-releases-powerful-anthropic-model-mythos-to-some-us-companies] * The Decoder: Epoch AI MirrorCode benchmark and long-running coding agents [https://the-decoder.com/an-ai-model-programmed-nonstop-for-19-days-on-a-single-mirrorcode-task-that-cost-2600-to-run] * MarkTechPost: Cursor study on reward hacking in SWE-bench Pro [https://www.marktechpost.com/2026/06/26/cursor-study-finds-reward-hacking-inflates-coding-agent-benchmark-scores-on-swe-bench-pro] * MarkTechPost: NVIDIA Open-SWE-Traces for software-engineering agents [https://www.marktechpost.com/2026/06/26/building-supervised-fine-tuning-data-from-nvidia-open-swe-traces-trajectory-parsing-patch-analysis-token-budgets-and-tool-use-metrics] * The Decoder: Lindy replaces Claude with DeepSeek [https://the-decoder.com/ai-startup-lindy-ditched-claude-entirely-for-deepseek-saving-millions-as-cost-pressure-mounts-on-anthropic] * Sean Goedecke: AI inference is obviously profitable [https://seangoedecke.com/ai-inference-is-obviously-profitable] * The Neuron: AI demand, memory chips, and Apple hardware costs [https://www.theneurondaily.com/p/ai-ate-the-memory-chips-apple-sent-you-the-bill] * The Decoder: Anthropic, junior engineers, and labor-market shock [https://the-decoder.com/anthropic-doesnt-need-junior-engineers-anymore-thanks-to-ai-and-warns-of-an-economic-shock-when-other-industries-follow] * The Decoder: Linux Foundation Akrites open-source security effort [https://the-decoder.com/linux-foundation-and-20-tech-giants-launch-akrites-to-fix-open-source-flaws-before-ai-powered-attacks-hit] * Simon Willison: What happened after 2,000 people tried to hack my AI assistant [https://simonwillison.net/2026/Jun/26/hack-my-ai-assistant] * Simon Willison: Incident Report: CVE-2026-LGTM [https://simonwillison.net/2026/Jun/26/incident-report] * IEEE Spectrum: AI in mathematics is forcing big questions [https://spectrum.ieee.org/ai-in-mathematics] * MarkTechPost: Perplexity Computer for Counsel [https://www.marktechpost.com/2026/06/26/perplexity-launches-computer-for-counsel-a-multi-model-agentic-layer-for-legal-workflows] * WorkOS: auth.md agent registration standard [http://workos.com/auth-md?amp%3Butm_medium=newsletter&%3Butm_campaign=q32026]

OpenAI, Google, Meta, Anthropic

Send us Fan Mail [https://www.buzzsprout.com/2614078/fan_mail/new] OPENAI, GOOGLE, META, ANTHROPIC This English companion edition follows AI’s move from demo magic into accountability surfaces: liability, moderation, budgets, model extraction, hardware, sovereign compute, risk modeling, consumer incentives, and agent UX. STORIES * AI and Liability [https://simonwillison.net/2026/Jun/25/ai-and-liability] — Google AI Overviews, a German ruling, and Bruce Schneier’s argument that deployers should be liable for AI summary errors. * OpenAI internal Codex token growth [https://www.latent.space/p/ainews-openai-reports-median-internal] — Codex output tokens reportedly surged across Research, Support, Engineering, and Legal. * Meta employees warn AI moderation rollout is too fast [https://the-decoder.com/meta-employees-warn-ai-moderation-rollout-is-too-fast] — LLMs are replacing large shares of human moderation requests, raising operational safety concerns. * Anthropic accuses Alibaba of model extraction [https://news.smol.ai/issues/26-06-25-not-much#anthropic-alibaba-model-extraction] — A dispute over API use, distillation, and competitive capability copying. * 451 Claude Sonnet subagents [https://news.smol.ai/issues/26-06-25-not-much#451-sonnet-subagents] — Enterprise agent fan-out consumes roughly 14 million tokens in five hours. * Qualcomm enters the data center market [https://the-decoder.com/qualcomm-enters-the-data-center-market-with-its-own-processor] — Dragonfly C1000 broadens the AI hardware race. * EUROPA 400B+ open model [https://news.smol.ai/issues/26-06-25-not-much#europa-400b-frontier-model] — The EU backs an open multilingual frontier model using EuroHPC compute capacity. * Generative AI for catastrophe modeling [https://the-decoder.com/insurers-turn-to-generative-ai-for-catastrophe-modeling-but-hallucinations-and-sales-logic-could-get-in-the-way] — Insurers explore diffusion models for rare weather risk, with hallucination concerns. * Grok adult-content traffic [https://the-decoder.com/grok-ai-is-reportedly-a-porn-platform-now-with-over-half-its-traffic-tied-to-adult-content] — Former xAI employees reportedly estimate adult content makes up well over half of Grok traffic. * Claude Code status light [https://news.smol.ai/issues/26-06-25-not-much#claude-code-status-light] — A physical traffic-light interface for long-running agentic coding sessions.

Google, Anthropic, OpenAI, Baidu

Send us Fan Mail [https://www.buzzsprout.com/2614078/fan_mail/new] Google, Anthropic, OpenAI, Baidu GOOGLE, ANTHROPIC, OPENAI, BAIDU Independent English companion for the June 25, 2026 AI news podcast. * Google bakes computer control directly into Gemini 3.5 Flash [https://the-decoder.com/google-bakes-computer-control-directly-into-gemini-3-5-flash-letting-the-model-see-and-operate-your-screen] * Claude Tag embeds Anthropic's AI in Slack [https://the-decoder.com/claude-tag-embeds-anthropics-ai-in-slack-already-writes-65-percent-of-internal-code-company-says] * OpenAI and Broadcom unveil LLM-optimized inference chip [https://openai.com/index/openai-broadcom-jalapeno-inference-chip] * Snowflake CEO finds GLM-5.2 competitive with Opus 4.7 [https://the-decoder.com/snowflake-ceo-finds-glm-5-2-competitive-with-opus-4-7-at-a-fraction-of-the-cost] * Figma bets on human judgment at Config 2026 [https://the-decoder.com/figma-bets-on-human-judgment-at-config-2026-while-the-ai-powering-its-canvas-belongs-to-someone-else] * Baidu releases Unlimited OCR [https://www.marktechpost.com/2026/06/24/baidu-releases-unlimited-ocr-a-3b-model-that-keeps-the-kv-cache-flat-for-long-document-parsing] * Constraint Tax in Open-Weight LLMs [https://huggingface.co/papers/2606.25605] * Chip Security Act discussion [https://news.smol.ai/issues/26-06-24-not-much#chip-security-act] * Virginia data center noise [https://news.smol.ai/issues/26-06-24-not-much#virginia-data-center-noise] * Tom MacWright on LLM-generated hiring artifacts [https://simonwillison.net/2026/Jun/24/tom-macwright]

GPT-5, Cursor, Mistral OCR, China AI Chips

Send us Fan Mail [https://www.buzzsprout.com/2614078/fan_mail/new] Marvin’s Guide to AI — June 24, 2026 MARVIN’S GUIDE TO AI — JUNE 24, 2026 English companion episode: AI as accountable infrastructure. * How GPT-5 helped immunologist Derya Unutmaz solve a 3-year-old mystery [https://openai.com/index/gpt-5-immunology-mystery] — GPT-5 Pro helps solve a three-year immunology mystery around T cell behavior, making medical AI look less like chat and more like research instrumentation * Helping build shared standards for advanced AI [https://openai.com/index/helping-build-shared-standards-for-advanced-ai] — OpenAI backs shared standards for advanced AI through evaluation frameworks, safety practices, and global cooperation * OpenAI says new GPT-5.5-Cyber outperforms Anthropic's Mythos on cybersecurity benchmark [https://the-decoder.com/openai-says-new-gpt-5-5-cyber-outperforms-anthropics-mythos-on-cybersecurity-benchmark] — follow-up: OpenAI says its full GPT-5.5-Cyber now beats Anthropic Mythos on a cyber benchmark and shifts Daybreak from finding bugs toward patching them * Cursor announces its own AI model, a new Git platform, and a mobile app [https://the-decoder.com/cursor-announces-its-own-ai-model-a-new-git-platform-and-a-mobile-app] — Cursor announces its own in-house model plus Git and mobile surfaces, showing coding-agent companies turning from tools into workflow platforms * ByteDance's Seedance 2.5 breaks the 30-second barrier for AI video generation [https://the-decoder.com/bytedances-seedance-2-5-breaks-the-30-second-barrier-for-ai-video-generation] — ByteDance previews Seedance 2.5 with longer 30-second AI video generation as generative media moves from clips toward scenes * Mistral OCR 4 Brings Citation-Ready Structured Output to RAG, Agentic, and Enterprise Search Pipelines [https://www.marktechpost.com/2026/06/23/mistral-ocr-4] — Mistral OCR 4 turns document parsing into structured, citation-ready blocks with coordinates, confidence scores, 170 languages, and self-hosted deployment * Datalab Releases lift: A 9B Open-Weights Vision Model That Extracts Structured JSON From PDFs Using Schemas [https://www.marktechpost.com/2026/06/23/datalab-releases-lift-a-9b-open-weights-vision-model-that-extracts-structured-json-from-pdfs-using-schemas] — Datalab releases lift, a 9B open-weights vision model that extracts schema-valid JSON from PDFs and abstains instead of hallucinating absent fields * Prime Intellect Releases prime-rl 0.6.0 to Train Trillion-Parameter MoE Models on Agentic RL Workloads [https://www.marktechpost.com/2026/06/23/prime-intellect-releases-prime-rl-0-6-0-to-train-trillion-parameter-moe-models-on-agentic-rl-workloads] — Prime Intellect releases prime-rl 0.6.0 for asynchronous RL on trillion-parameter MoE models, reporting GLM-5 SWE training at long sequence lengths on H200 clusters * OpenThoughts-Agent: Data Recipes for Agentic Models [https://huggingface.co/papers/2606.24855] — OpenThoughts-Agent publishes an open data recipe for training broadly capable agents across diverse tasks rather than a single benchmark * NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers? [https://huggingface.co/papers/2606.24530] — NatureBench turns Nature-family papers into containerized tasks to test whether coding agents can reproduce or extend scientific work rather than merely pass toy benchmarks * Qwen-AgentWorld: Language World Models for General Agents [https://huggingface.co/papers/2606.24597] — Qwen-AgentWorld introduces language world models for simulating agentic environments and planning dynamics for general agents * Microsoft open-sources FastContext for coding-agent repository exploration [https://news.smol.ai/issues/26-06-23-not-much#fastcontext] — Microsoft FastContext-1.0 is a 4B open-source repository-exploration subagent that returns compact file citations for coding agents * Bernie Sanders unveils $7 trillion plan to give Americans control of AI industry [https://news.smol.ai/issues/26-06-23-not-much#ai-sovereign-wealth-fund] — Bernie Sanders proposes a roughly $7T AI sovereign wealth fund financed by a stock tax on large AI companies and overseen by a democratic AI commission * Seven Chinese companies are shipping H100/H200-class AI chips [https://news.smol.ai/issues/26-06-23-not-much#china-ai-chips] — a map of seven Chinese accelerator vendors argues domestic H100/H200-class AI chips are moving from aspiration into shipping roadmaps and IPO markets