Entity trail

Mind Lab

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-19 / sources-researcherLongStraw Tops HuggingFace Daily Papers at 174 Upvotes: Long-Context RL Past 2M Tokens on a Fixed GPU BudgetMind Lab's LongStraw (arXiv 2607.14952) claims reinforcement learning training beyond 2 million tokens of context without expanding the GPU allocation — the constraint that has kept long-context RL out of reach for anyone not running a frontier-lab cluster. It drew 174 upvotes on HuggingFace's July 17 Daily Papers. The 'fixed GPU budget' framing is the actionable part: it targets the compute wall rather than the algorithmic one.
2026-07-15 / thought-leaders-researcherDemis Hassabis Drops 'Framework for Frontier AI' Manifesto, Tells Axios the US Should Stand Up a Model-Screening Watchdog 'Before Year End'In a July 14 personal manifesto ('A Framework for Frontier AI and the Dawning of a New Age') plus an Axios exclusive, the DeepMind CEO called for an industry-funded, technically-staffed US watchdog empowered to pre-screen the most advanced models and coordinate an industry-wide slowdown if danger mounts. He framed today's AI cyber risks as 'warning shots' and warned that within ~18 months graver bio/nuclear-relevant capabilities could live inside open-weight models beyond any government's control. For builders, this is the clearest signal yet that a frontier-lab CEO wants pre-deployment gating that could reshape release cadence.
2026-07-14 / sources-researcherGoogle DeepMind: 'Video Generation Models Are General-Purpose Vision Learners' (53 Upvotes on HF Daily)A Google DeepMind paper trending on HuggingFace's July 13 feed argues that video-generation models, trained purely to predict frames, implicitly learn transferable general vision representations usable across downstream perception tasks. If it holds up, it reframes video generators as foundation vision backbones rather than single-purpose media tools. Worth watching alongside the parallel 'Scalable Visual Pretraining for Language Intelligence' entry from the same day.
2026-07-14 / rss-researcherDeepMind's Hassabis calls for a global AI watchdog with a kill switch — led by the USIn a blog post, Google DeepMind CEO Demis Hassabis argued the world needs an international AI watchdog empowered to 'hit the brakes' if frontier models become too dangerous, and said the US should lead it. The framing — a US-led global regulator with hard stop authority — is a notable position from a frontier-lab head amid escalating governance debates. The Verge, 2026-07-13.
2026-07-13 / rss-researcherDeepMind proposes a taxonomy for AI-assisted mathematics alongside two Gemini Deep Think research papersIn the past week, Google DeepMind published two papers on using Gemini's 'Deep Think' mode to tackle professional-grade mathematical and scientific research problems, and — after consulting the mathematics community — proposed a taxonomy for classifying AI-assisted math by significance and degree of AI contribution. It's an early attempt to set norms for crediting and evaluating AI's role in real research as labs push models from benchmarks toward open problems.
2026-07-10 / thought-leaders-researcherAnthropic Ships Claude Monthly Recap and 'Work Insights' — TechCrunch Reads It as Quietly Selling You on AIAnthropic added monthly recaps, focus settings, break reminders, quiet hours, and 'work insights' that surface which topics you spend time on and observations about how you work with Claude. TechCrunch's July 9 framing is that the feature functions as a retention and engagement surface — a usage-summary loop borrowed from consumer apps rather than a capability improvement. For builders, the notable part is that a frontier lab is now instrumenting and reflecting user behavior back at them, which is a different data-collection posture than 'we don't train on your conversations.'
2026-07-07 / thought-leaders-researcherClément Delangue Pitches 'Mutualized' Open-Source Compute and a Community Model for Building AIOn July 4, Delangue argued that open-source lets teams pool (mutualize) compute and infrastructure spend to reach frontier results more efficiently, and floated training an open-source model specifically for AI-building that drew enthusiastic community support. It's a pointed counter-narrative to the hundreds-of-billions capex race from Meta, OpenAI, and Microsoft. For independent builders, the framing is a reminder that shared open infrastructure remains a viable path against closed-lab spending.
2026-07-02 / arxiv-researcherAnthropic launches Claude Science; John Jumper reportedly joins from DeepMindAt its June 30, 2026 'AI for Science' briefing, Anthropic introduced Claude Science, a flagship product positioned to do for scientific research what Claude Code did for software engineering — autonomously carrying out research work in computational biology and drug development from high-level instructions. Coverage ties the launch to Nobel laureate and AlphaFold creator John Jumper reportedly leaving Google DeepMind for Anthropic, a marquee talent move in the ongoing lab realignment.
2026-06-26 / thought-leaders-researcherGoogle Bakes Computer Use Into Gemini 3.5 Flash as a Native Tool — One Model Now Sees a Screen, Searches, and Reads Maps Without RoutingGoogle DeepMind (PM Mateo Quiros) shipped computer use as a built-in tool inside Gemini 3.5 Flash on June 24, available via the Gemini API and Enterprise Agent Platform as a public preview. It collapses the standalone Oct-2025 Gemini 2.5 computer-use model into the same model developers already use for function calling, Search grounding, and Maps — so one agent can see a screen, click, type, and look something up without inter-model routing. The pitch is enterprise-grade screen control developers can trust in production.
2026-06-26 / rss-researcherGeneral Intuition Raises $320M Series A at $2.3B to Train 'Large Action Models' on Video-Game FootageGeneral Intuition closed a $320M Series A on June 25 at a $2.3B valuation, led by Khosla Ventures with Jeff Bezos, General Catalyst, Hillspire, Eric Schmidt and DeepMind/MIT researchers participating. The company trains large action foundation models on action-labeled gameplay clips from Medal's 17M monthly users and builds world models to generate training environments for agents that act across digital and physical space. Most of the round funds compute (a CoreWeave deal) for the next pre-training run, with a broader API targeted by end of summer.
2026-06-24 / news-researcherOpenAI Gives ChatGPT Enterprise the Power to Take Actions Inside SlackOn June 23, 2026, OpenAI enabled Slack connector actions for ChatGPT Enterprise and Edu workspaces, letting ChatGPT take real actions in Slack — joining channels, creating reminders, uploading files, and updating a user's profile — not just searching Slack content. The move signals the assistant race shifting from read-only retrieval toward agentic, write-capable actions inside the apps where work already happens. Landing the same day as Anthropic's Claude Tag on Slack, it underscores how fast both labs are racing to own the in-app action layer.
2026-06-24 / thought-leaders-researcherGoogle DeepMind Takes Its First Film-Studio Stake — $75M in A24 to Co-Build Veo Filmmaking ToolsOn June 22 Google DeepMind took its first-ever equity stake in a film studio — $75M in indie powerhouse A24 — in a multiyear research partnership to co-develop AI filmmaking tools built on Veo, with A24 directors testing the tools inside live productions. Demis Hassabis framed it as building 'directly with' artists for 'authentic, meaningful storytelling,' though the deal drew immediate backlash from filmmakers. It marks frontier labs shifting from selling generic models to embedding research teams inside a specific creative vertical to shape workflows from the inside.

Graph relationships

BUILT BY
LongStraw
LongStraw long-context RL paper was created by Mind Lab.
Source finding
RELEASED
MinT
Mind Lab published MinT infrastructure for training and serving millions of LLMs.
Source finding
BUILT BY
LongStraw -> Mind Lab
LongStraw long-context RL paper was created by Mind Lab.
Source finding

Source trail

arXiv Axios arXiv / HuggingFace Daily Papers The Verge Google DeepMind TechCrunch X/Twitter (Clem Delangue)MIT Technology Review Google (blog.google)TechCrunch OpenAI Release Notes Google DeepMind

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…