What’s Happening on Tuesday, June 16, 2026?

Stories

Radical Numerics Emerges from Stealth with $50M Seed to Build General Biological Intelligence — Founding Team Created Field of Generative Genomics

Radical Numerics launched from stealth with a $50M seed round, building general biological intelligence with a dual mandate: human health design and biodefense. The founding team, including Michael Poli, created the fiel...

Sources: x.com · 2 hours, 52 minutes ago

Salesforce Acquires Fin for ~$3.6 Billion — Customer Service AI's Category Leader Joins Agentforce, Deal Closes FY2027

Salesforce agreed to acquire Fin (formerly Intercom) for ~$3.6B. CEO Eoghan McCabe's announcement got 2,793 likes and 419K views. The deal brings the #1 AI customer agent and Fin's Apex model into Salesforce's Agentforce...

Sources: x.com · 2 hours, 57 minutes ago

Third Party Claims to Have Built the Fable Compliance Gate — Technical Barrier to Restoring Fable 5 Now Reportedly Removed

A third party claims to have built the compliance gate the US government has been requesting from Anthropic — a mechanism to restrict Fable 5 and Mythos to government-vetted users. If Anthropic adopts it, the export cont...

Sources: di.gg · 2 hours, 58 minutes ago

Sakana AI Launches Marlin Today — Tokyo Lab Transitions from Research to Market with Autonomous Enterprise Research Assistant

Sakana AI launched its first commercial product today — Marlin, an autonomous research assistant for enterprise clients. The lab co-founded by David Ha and Llion Jones moves from two years of pure research to market.

Sources: x.com · 2 hours, 58 minutes ago

Salesforce Acquires Fin for ~$3.6 Billion — Customer Service AI Category Leader Joins Agentforce, Deal Closes FY2027

Salesforce agreed to acquire Fin (formerly Intercom) for ~$3.6B. The deal brings the #1 AI customer agent and Fin's Apex model into Agentforce. Announced by CEO Eoghan McCabe — 2,793 likes and 419K views.

Sources: x.com · 2 hours, 58 minutes ago

Salesforce Acquires Fin (Formerly Intercom) for $3.6 Billion — Brings Number-One AI Customer Agent and Apex Model into Agentforce

Salesforce agreed to acquire Fin, formerly Intercom, for approximately $3.6 billion. The deal brings Fin's AI customer service agents and proprietary Apex model into Salesforce's Agentforce platform. Announced by CEO Eog...

Sources: x.com · 2 hours, 58 minutes ago

Ten Things That Changed This Week — Fable Banned, GLM 5.2 Takes Its Spot, Sakana Launches, OpenAI and Anthropic Race to IPO

The week of June 9-15, 2026 in AI: the US government banned Fable 5, China's GLM 5.2 took the top benchmark spot within 48 hours, Sakana AI launched its first commercial product, and both OpenAI and Anthropic filed confi...

Sources: x.com · 4 hours, 53 minutes ago

Clinical AI Startup Hiring Across All Tracks — Pre-Training, Post-Training, Agents, and Researchers Working on Cancer Applications

A clinical AI startup is hiring AI researchers across all technical tracks including pre-training, scaling, post-training, agents, and direct disease research applications. Working on actual clinical trial data and propr...

Sources: x.com · 4 hours, 53 minutes ago

Top 12 Today

Today's twelve most important AI stories — Sakana Marlin launches, the Fable compliance gate emerges, and the AI IPO race begins.

Salesforce Acquires Fin for ~$3.6 Billion — Customer Service AI's Category Leader Joins Agentforce, Deal Closes FY2027

Salesforce agreed to acquire Fin (formerly Intercom) for ~$3.6B. CEO Eoghan McCabe's announcement got 2,793 likes and 419K views. The deal brings the #1 AI customer agent and Fin's Apex model into Salesforce's Agentforce...

Sources: x.com · 2 hours, 57 minutes ago

DeepMind Publishes 'From AGI to ASI' — Defines AGI as One Human-Level System, ASI as a Colony of Multiple AGIs Working Together

Google DeepMind's paper formally defines AGI as a single human-level general system and ASI as a colony of multiple AGIs working together whose collective capability exceeds human organizational capability. The colony de...

Sources: x.com · 10 hours, 38 minutes ago

Columbia LCLM Achieves 8.8x Faster First Token via 1:16 Context Compression — Latent Context Models Speed Long Tasks

Columbia's Micah Goldblum and collaborators released Latent Context Language Models achieving an 8.8x improvement in time-to-first-token by compressing context at ratios up to 1:16. The approach makes long-context tasks ...

Sources: di.gg · 10 hours, 38 minutes ago

OpenAI roon Proposes Relaunching Sydney Bing with Safety Harness — Midjourney's David Holz Publicly Agrees

OpenAI researcher roon proposed that Microsoft's erratic Sydney Bing AI persona could be safely relaunched with a robust safety harness, and Midjourney's David Holz agreed. The discussion reopens questions about AI perso...

Sources: di.gg · 10 hours, 38 minutes ago

Cohere Releases North Mini Code — Apache 2.0, 30B Parameters with 3B Active MoE, 256K Context for Agentic Terminal Workflows

Cohere released North Mini Code, an Apache 2.0 open-weights coding model with 30B parameters (3B active MoE), 256K context, and 80.2% pass@10 on SWE-Bench. Optimized for agentic terminal workflows and runnable locally on...

Sources: x.com · 10 hours, 38 minutes ago

China Plans $295 Billion AI Infrastructure Build Over 5 Years — Analysts Say Capital Alone Cannot Solve the Chip Gap

China announced a $295 billion, 5-year AI infrastructure investment plan targeting national data center construction. Analysts argue the plan cannot overcome chip access constraints from US export controls, raising quest...

Sources: di.gg · 10 hours, 38 minutes ago

Third Party Claims to Have Built the Fable Compliance Gate — Restricting Mythos to Authorized Users Is Now Technically Possible

A third party claims to have built what the US government has been asking Anthropic to create: a compliance gate for Fable 5 and Mythos that restricts access to government-vetted users. If adopted, the technical barrier ...

Sources: di.gg · 10 hours, 38 minutes ago

Both OpenAI and Anthropic File Confidential S-1s in the Same Week — Dual AI IPO Race Leaves Timing Undecided

Both OpenAI and Anthropic filed confidential IPO S-1 registrations within days of each other, creating a dual AI IPO race. Neither has set a timeline — OpenAI said it may remain private — but both are now formally on the...

Sources: x.com · 10 hours, 38 minutes ago

Big Stuff Happening in AI Right Now

Salesforce acquires Fin for $3.6B; Sakana AI launches Marlin; third party builds Fable compliance gate.

Third Party Claims to Have Built the Fable Compliance Gate — Technical Barrier to Restoring Fable 5 Now Reportedly Removed

A third party claims to have built the compliance gate the US government has been requesting from Anthropic — a mechanism to restrict Fable 5 and Mythos to government-vetted users. If Anthropic adopts it, the export cont...

Sources: di.gg · 2 hours, 58 minutes ago

Sakana AI Launches Marlin Today — Tokyo Lab Transitions from Research to Market with Autonomous Enterprise Research Assistant

Sakana AI launched its first commercial product today — Marlin, an autonomous research assistant for enterprise clients. The lab co-founded by David Ha and Llion Jones moves from two years of pure research to market.

Sources: x.com · 2 hours, 58 minutes ago

Salesforce Acquires Fin for ~$3.6 Billion — Customer Service AI Category Leader Joins Agentforce, Deal Closes FY2027

Salesforce agreed to acquire Fin (formerly Intercom) for ~$3.6B. The deal brings the #1 AI customer agent and Fin's Apex model into Agentforce. Announced by CEO Eoghan McCabe — 2,793 likes and 419K views.

Sources: x.com · 2 hours, 58 minutes ago

AI Research Engineer Openings in Pre-Training and Inference — Kernel Dev, All Parallelism Flavors, ASIC/FPGA/GPU Targeting

New AI Research Engineer openings in two tracks: Pre-training (kernel dev, all parallelism, fault tolerance) and Inference (kernel dev, all devices including ASIC, FPGA, GPU). Research-level roles shaping the technical d...

Sources: x.com · 4 hours, 53 minutes ago

Huawei 3D Lock Screen Uses Gaussian Splatting — Scan Any Object in 360 Degrees and Turn It into a Live Wallpaper

Huawei's new 3D lock screen lets users scan any object in 360 degrees, generate a full Gaussian Splatting 3D asset, and use it as a live interactive lock screen wallpaper. Unlike iPhone's fixed stereoscopic depth effect,...

Sources: x.com · 4 hours, 53 minutes ago

CapCut Launches First Global AI Festival — $200K in Prizes Across Film, Series, Creative, and Commercial Tracks Through August 10

CapCut's CRE[AI]TE is the first global AI festival for original AI-assisted work. Four tracks — Film, Series, Creative, and Commercial — with $200K in total prizes. Submissions open through August 10, 2026.

Sources: x.com · 4 hours, 53 minutes ago

Robert Scoble on Fable vs Mythos — 'The Problem That Let People Get From Fable to the More Powerful One Is Going to Play Out This Week'

Robert Scoble explains the Fable/Mythos situation to his followers: Anthropic's Mythos is more powerful than Fable, a jailbreak let users bridge from Fable to Mythos, and the resolution is coming this week.

Sources: x.com · 10 hours, 26 minutes ago

GLM 5.2 Takes BridgeBench #1 at 42.8 Reasoning After Fable Ban — 'You Cannot Export Control Your Way Out of Open Source'

Zhipu AI's GLM 5.2 took the #1 spot on BridgeBench scoring 100.0 on coding and 42.8 on reasoning — beating Fable 5 two days after the US banned it. At one-tenth the cost and 300 tokens per second. 5,906 likes, 649K views...

Sources: x.com · 10 hours, 29 minutes ago

News From AI Companies

Sakana Marlin launches as the first commercial autonomous research assistant from the Tokyo lab; Google ships Gemini 3.5 Flash Live Translate.

Gemini 3.5 Flash Live Translate Ships — Real-Time Babel Fish in 70+ Languages Now in Gemini API and Google Translate

Google launched Gemini 3.5 Flash Live Translate with real-time speech-to-speech translation across 70+ languages and 2,000+ bidirectional pairs. Available now in the Gemini API, AI Studio, and Google Translate, with Goog...

Sources: x.com · 10 hours, 38 minutes ago

Sakana AI Marlin Launches Today — Autonomous Research Assistant for Enterprise Is Sakana's First Commercial Product

Sakana AI launched Marlin today, transitioning from two years of research to its first commercial product — an autonomous research assistant designed for enterprise clients.

Sources: x.com · 10 hours, 38 minutes ago

Sakana AI Launches First Commercial Product — Research Lab Crosses into Enterprise Market

Sakana AI (co-founded by David Ha, Llion Jones) launched Marlin today — its first commercial product after 2 years of research. An autonomous enterprise research assistant. 544 likes on launch. The Tokyo lab transitions ...

Sources: AI Companies #1, x.com · 10 hours, 34 minutes ago

OpenAI Merging ChatGPT and Codex Into Super App — Ex-DeepMind Sottiaux Named Head of Core Products

OpenAI appoints ex-DeepMind engineer Thibault Sottiaux as head of core products to merge ChatGPT and Codex into one super app. Product consolidation announced same day as S-1 filing. Direct competitive response to Claude...

Sources: x.com · 72 hours, 43 minutes ago

Apple WWDC: Agentic AI Fixes Compromised Passwords Autonomously — No User Action Required

Apple Intelligence in Passwords app will agentically detect and fix compromised passwords in Safari without user action, launching iOS 27. First consumer-facing autonomous security agent from Apple at WWDC 2026.

Sources: x.com · 74 hours, 21 minutes ago

Cursor $4B ARR: Enterprise AI Coding Is a Real Market

Cursor reaches $4B ARR this week, up from $2B in February and $3B in April, with 75% of revenue from enterprise customers — fastest ARR growth in software history

Sources: AI Companies #1, x.com · 162 hours, 57 minutes ago

Siri AI Launches at WWDC 2026

Apple launches Siri AI at WWDC — new dedicated app, iCloud conversation sync, gaze activation, no 'Hey Siri' required. Ships September with iOS 27. Tim Cook's farewell keynote.

Sources: AI Newsmakers, x.com · 168 hours, 20 minutes ago

OpenAI Codex Hits 5 Million Weekly Active Users — Merging With ChatGPT and Atlas Into Unified Product

OpenAI's Codex coding agent has reached 5 million weekly active users, helped by better models and a dedicated OpenAI team. The company is now reorganizing around Codex and planning to merge it with ChatGPT and Atlas....

Sources: AI Newsmakers, x.com · 243 hours, 2 minutes ago

Robots

Tesla FSD Supervised approved in Denmark — first Scandinavian market cleared, rollout begins soon.

Tesla FSD Supervised Approved in Denmark — Scandinavian Rollout Begins as European Market Expansion Continues

Tesla Europe confirmed FSD Supervised received regulatory approval in Denmark. Rollout begins soon in the first Scandinavian market to clear the feature. Community speculation about Portugal as the next European market i...

Sources: x.com · 10 hours, 38 minutes ago

Baidu Wins Level 4 Driverless Approval in Switzerland — Fully Autonomous Public Transit Launches 2027

Baidu wins Level 4 fully driverless robotaxi approval in Switzerland with Swiss Post's PostBus. Europe's largest automated public transport operation. Fully driverless rides (no safety driver) starting early 2027. Chines...

Sources: x.com · 72 hours, 43 minutes ago

Tesla FSD Supervised Approved in Denmark — First Major EU Regulatory Green Light

Denmark grants Tesla regulatory approval to deploy Full Self-Driving Supervised software — making it one of the first European countries to greenlight FSD for public roads

Sources: Robot AI, di.gg · 99 hours, 35 minutes ago

LadderMan Zero-Shot Sim-to-Real Breaks a Robotics Barrier — On-Ladder Task Execution

Amazon Fundamental AI Research's LadderMan achieves zero-shot sim-to-real transfer on Unitree G1 humanoid — climbs ladder AND performs manipulation tasks from the ladder. No real-world fine-tuning required after simulati...

Sources: AI Newsmakers, x.com · 145 hours, 52 minutes ago

Tesla FSD Denmark Approval — European Rollout Continues Country by Country

Tesla FSD Supervised approved in Denmark today — rollout begins soon. Denmark joins France, Spain, and Portugal as cleared EU markets. The announcement became the most-viewed story on di.gg with 948,000 views and 11,600 ...

Sources: AI Companies #1, x.com · 145 hours, 52 minutes ago

Google Gemini 3.5 Flash Live Translate Demo — Real-Time Speech-to-Speech Translation Across 70 Languages

Google published a demo of Gemini 3.5 Flash Live Translate showing real-time speech-to-speech translation across 70+ languages with natural intonation and no perceptible lag. Available now in the Gemini API and Google Tr...

Sources: x.com · 10 hours, 29 minutes ago

China's $295B AI Build Cannot Buy Its Way Past the Chip Gap — Export Controls Cap What Infrastructure Spending Can Achieve

China's massive $295 billion AI infrastructure investment cannot solve the fundamental constraint of US export controls on advanced GPUs. The plan builds data centers but cannot equip them with frontier chips, making alg...

Sources: di.gg · 10 hours, 37 minutes ago

Mistral Community Campaigns for 'Le Chaton Fat' as Official Model Name — A Fake Rumor Became a Naming Pressure Campaign

A fake leak claiming Mistral was releasing a model called 'Le Chaton Fat' spread across AI Twitter, accumulating enough community attachment that people are now lobbying Mistral to actually use the name. AI humor has bec...

Sources: x.com · 10 hours, 37 minutes ago

Tools for Creatives

CapCut's first global AI Festival opens with $200K in prizes; Huawei ships 3D Gaussian Splatting lock screens.

Huawei 3D Lock Screen Uses Gaussian Splatting — Scan Any Object in 360 Degrees and Turn It into a Live Wallpaper

Huawei's new 3D lock screen lets users scan any object in 360 degrees, generate a full Gaussian Splatting 3D asset, and use it as a live interactive lock screen wallpaper. Unlike iPhone's fixed stereoscopic depth effect,...

Sources: x.com · 4 hours, 53 minutes ago

CapCut Launches First Global AI Festival — $200K in Prizes Across Film, Series, Creative, and Commercial Tracks Through August 10

CapCut's CRE[AI]TE is the first global AI festival for original AI-assisted work. Four tracks — Film, Series, Creative, and Commercial — with $200K in total prizes. Submissions open through August 10, 2026.

Sources: x.com · 4 hours, 53 minutes ago

Martin Scorsese Joins Black Forest Labs as Advisor to Guide FLUX Visual AI — Uses It for Pre-Production Storyboards

Director Martin Scorsese joins Black Forest Labs as an advisor to guide FLUX visual AI development. Scorsese uses the model to generate pre-production storyboards. #1 viewed story on di.gg today: 53k views, 629 likes, 20...

Sources: AI Holodeck/Creative, x.com · 248 hours, 3 minutes ago

DeepMind Defines the AGI-to-ASI Gap — AGI Is One Human-Level System, ASI Is What Happens When Multiple AGIs Coordinate

Google DeepMind published a paper formally defining the AGI-to-ASI transition: AGI is one human-level general system, and ASI is a colony of multiple AGIs working together. The colony definition has significant implicati...

Sources: x.com · 10 hours, 37 minutes ago

roon Proposes Relaunching Sydney Bing with Safety Harness — David Holz Agrees, Community Debates Whether Modern Safety Changes the Calculus

OpenAI researcher roon proposed that Microsoft's erratic Sydney Bing persona could be relaunched with modern safety infrastructure. Midjourney founder David Holz publicly supported the idea, sparking debate about AI pers...

Sources: di.gg · 10 hours, 37 minutes ago

China's $295 Billion AI Plan Targets Data Center Build-Out — But Chip Constraints Cap What Infrastructure Spending Can Achieve

China's 5-year, $295 billion AI infrastructure plan is the largest national AI infrastructure commitment ever announced. Analysts argue the spend cannot overcome US export controls on advanced chips, making algorithmic e...

Sources: di.gg · 10 hours, 38 minutes ago

Arcee AI Replaces AWS S3 with Hugging Face in Multi-Million Dollar Deal — First Major US Lab to Migrate Entire Catalog

Arcee AI migrated all models and datasets — including private ones — from AWS S3 to Hugging Face Private Storage in a multi-million dollar exclusive deal. The first major US AI lab to abandon AWS S3 entirely for AI-nativ...

Sources: x.com · 10 hours, 38 minutes ago

China $295 Billion AI Infrastructure Plan Faces Fundamental Chip Constraint — Five-Year Build Cannot Solve Access to Advanced GPUs

China's $295 billion, 5-year AI infrastructure plan targets national data center construction but faces a constraint capital cannot solve: no access to advanced GPUs under US export controls. Analysts question whether sp...

Sources: di.gg · 10 hours, 38 minutes ago

Funding & Deals

Salesforce acquires Fin (formerly Intercom) for $3.6B — the largest enterprise AI acquisition of 2026 brings the #1 AI customer agent into Agentforce.

Salesforce Acquires Fin (Formerly Intercom) for $3.6 Billion — Brings Number-One AI Customer Agent and Apex Model into Agentforce

Salesforce agreed to acquire Fin, formerly Intercom, for approximately $3.6 billion. The deal brings Fin's AI customer service agents and proprietary Apex model into Salesforce's Agentforce platform. Announced by CEO Eog...

Sources: x.com · 2 hours, 58 minutes ago

Salesforce Acquires Fin (Formerly Intercom) for $3.6 Billion — Customer Service AI Category Leader Joins Agentforce

Salesforce agreed to acquire Fin AI (formerly Intercom) for ~$3.6B. CEO Eoghan McCabe announced the deal Monday morning — 2,793 likes, 419K views. Fin is the #1 AI customer agent; the deal adds Fin and its Apex model to ...

Sources: AI Companies #2, x.com · 2 hours, 57 minutes ago

OpenAI and Anthropic Both File Confidential S-1s in the Same Week — The AI IPO Era Has Begun

Both OpenAI ($852B valuation) and Anthropic filed confidential S-1 IPO registrations within days of each other. No timelines set. Whichever goes public first sets the AI valuation benchmark. The era of AI IPO speculation...

Sources: AI Influencers, x.com · 10 hours, 34 minutes ago

Mistral AI Exploring €3B Raise at €20B Valuation — Largest European AI Capital Raise in History

Mistral AI exploring €3B raise at €20B valuation — would be largest European AI capital raise in history. Positions Mistral as one of the most valuable AI labs globally outside US. Breaking on same day as SpaceX IPO and ...

Sources: x.com · 72 hours, 43 minutes ago

OpenAI Files Confidential S-1 — Fall 2026 IPO Now Possible, $300B+ Floor Set

OpenAI files confidential S-1 with SEC, working with Goldman Sachs and Morgan Stanley. No IPO timing set but fall 2026 debut possible. $300B+ valuation floor established by last private round. Concurrent with SpaceX IPO....

Sources: x.com · 74 hours, 21 minutes ago

Prometheus Raises $12B for "Artificial General Engineer" — Bezos Bets on Physical-World AI

Prometheus, backed by Jeff Bezos, has raised $12 billion to build what it calls an "artificial general engineer" for the physical world. Primary targets: drug design and heavy engineering — areas where AI has mostly stay...

Sources: AI Companies #1, x.com · 86 hours, 45 minutes ago

Genspark Closes $100M Series B Extension — Total $485M at $2.6B, Hires Enterprise CRO

Genspark closed a $100M Series B extension, bringing its total Series B to $485M at a $2.6B post-money valuation. The company reports 6,000+ enterprise clients and hired Jamison Powell — who previously scaled a company f...

Sources: AI Companies #1, x.com · 86 hours, 45 minutes ago

NEURA Robotics Raises $1.4B Series C at $7B Valuation — Tether Leads

NEURA Robotics closes up to $1.4 billion Series C led by Tether at a $7 billion valuation — one of the largest private investment rounds in humanoid robotics history

Sources: Robot AI, x.com · 99 hours, 35 minutes ago

Models & Benchmarks

Cohere North Mini Code ships Apache 2.0 with 3B active MoE for local coding; GLM 5.2 tops BridgeBench after Fable's shutdown.

GLM 5.2 Tops BridgeBench #1 — Zhipu AI's New Model Beats Fable 5 on Benchmark Hours After Fable's US Government Shutdown

Zhipu AI's GLM 5.2 took the #1 position on BridgeBench with 1,263 likes — claiming the top spot previously held by Fable 5, which was pulled offline by US government export controls the same day. China's open-source AI m...

Sources: x.com · 10 hours, 38 minutes ago

Cohere North Mini Code Hits Apache 2.0 — 30B-Parameter MoE with 3B Active Runs on a Single H100 or Apple Silicon

Cohere released North Mini Code, an Apache 2.0 open-weights coding model with 30B parameters (3B active MoE), 256K context, and 80.2% pass@10 on SWE-Bench. Runs locally on a single H100 or via MLX on Apple Silicon with V...

Sources: x.com · 10 hours, 38 minutes ago

Fable Was #1 On Every Benchmark — Arena.ai Confirms Before Government Pulls It

Arena.ai before removal: 'Claude Fable 5 is the most powerful model we've ever tested — ranking #1 across Agent, Text, and Code Arena, and setting a new record.' Now offline per government directive.

Sources: AI Newsmakers, x.com · 64 hours, 28 minutes ago

Gemini 3.5 Live Translate: Real-Time Babel Fish Ships With 2,000+ Language Pairs

Gemini 3.5 Live Translate ships: 70+ languages, 2,000+ bidirectional pairs, auto-detects speaker language, zero lag, works in noise. Available now in Gemini API, AI Studio, Google Translate. Google Meet integration comin...

Sources: x.com · 74 hours, 21 minutes ago

Gemini 3.5 Flash Live Translate: 70 Languages, 2,000+ Pairs, Real-Time

Google launches Gemini 3.5 Flash Live Translate — real-time low-latency speech-to-speech translation supporting 70 languages and 2,000+ bidirectional language pairs simultaneously

Sources: AI Newsmakers, di.gg · 99 hours, 35 minutes ago

DiffusionGemma: Parallel Text Generation

Google releases DiffusionGemma — open Apache 2.0 model generating 256 tokens IN PARALLEL using diffusion, not token-by-token. Built on Gemma 4 Encoder-Denoiser, 26B MoE activating 3.8B parameters, fits 18GB VRAM

Sources: AI Newsmakers, x.com · 121 hours, 34 minutes ago

Gemini 3.5 Live Translate Ships — A Real-Time Babel Fish in 70 Languages

Google ships Gemini 3.5 Live Translate — real-time speech-to-speech in 70+ languages and 2,000+ pairs, available now in Gemini API, AI Studio, and Google Translate. Auto-detects language, preserves tone and pacing, works...

Sources: AI Newsmakers, x.com · 145 hours, 52 minutes ago

Claude Fable Launches at 2x Opus Pricing — Surprise Affordability Drives Adoption

Anthropic releases Claude Fable today — the public Mythos — at $10/$50 per million tokens, 2x Opus pricing, far below the expected 5x premium. Experimental 5-level nested subagent support for Claude-to-Claude orchestrati...

Sources: AI Community #5, x.com · 145 hours, 52 minutes ago

Labs

Sakana AI transitions from research to market with Marlin; DeepMind publishes the first formal AGI-to-ASI definition paper.

DeepMind Publishes Formal AGI-to-ASI Transition Paper — Colony of AGIs Is the Definition, Not a Smarter Singleton

A Google DeepMind paper formally defines AGI as one human-level general system and ASI as a colony of multiple AGIs working together. The colony definition aligns with multi-agent AI development and has been retweeted 83...

Sources: x.com · 10 hours, 38 minutes ago

Sakana AI Makes First Commercial Move with Marlin Launch — Two Years of Research Becomes Enterprise Product

Sakana AI launched Marlin, its first commercial product — two years after founding as a research lab. The David Ha and Llion Jones lab is now entering the enterprise market with an autonomous research assistant.

Sources: x.com · 10 hours, 38 minutes ago

xAI Trained Its Coding Agent on Anthropic Models — Then Anthropic Cut Them Off, Now Anthropic Pays xAI for Compute

The Information reveals xAI used Anthropic models to train its coding agent, then took steps to maintain access after Anthropic cut them off. The relationship has now inverted: Anthropic is paying xAI for compute.

Sources: AI Newsmakers, x.com · 243 hours, 2 minutes ago

Nathan Lambert and Finbarr Timbers Both Leave Allen Institute for AI on Same Day — Open-Source Research Brain Drain

Nathan Lambert, Ai2 post-training lead who developed OLMo and Tulu, is leaving the Allen Institute for AI. Finbarr Timbers, who worked on RL for LLMs, is also departing simultaneously. Lambert plans several months off be...

Sources: AI Newsmakers, x.com · 248 hours, 3 minutes ago

Anthropic Charts Path to Recursive Self-Improvement — 80% of Code Now Written by Claude in May 2026

Anthropic releases internal data: Claude authored over 80% of code merged into Anthropic's codebase in May 2026, up from low single digits before Claude Code. Engineers merge 8x more code per day than in 2024. Company ca...

Sources: AI Newsmakers, x.com · 248 hours, 3 minutes ago

Fable Compliance Gate Reportedly Built by Third Party — Anthropic Now Has a Technical Path to Restoring Access

A third party claims to have built the compliance gate the US government has been requesting from Anthropic to restore Fable 5 and Mythos. If Anthropic adopts it, the export control order could be lifted.

Sources: di.gg · 10 hours, 38 minutes ago

Columbia LCLM Paper — 8.8x Time-to-First-Token Speedup Using 1:16 Context Compression Ratios

Columbia's Latent Context Language Models achieve 8.8x faster time-to-first-token by compressing context at ratios up to 1:16. The approach makes long-context document analysis and code review significantly faster withou...

Sources: di.gg · 10 hours, 38 minutes ago

DeepMind 'From AGI to ASI' — AGI Is One Human-Level System, ASI Is a Colony of Multiple AGIs in Coordination

Google DeepMind's formal paper defines AGI as a single human-level general system and ASI as a colony of multiple AGIs working together. The colony framework aligns with multi-agent AI development and has 83 retweets....

Sources: x.com · 10 hours, 38 minutes ago

Papers

DeepMind defines AGI and ASI formally for the first time; Columbia LCLM achieves 8.8x time-to-first-token speedup via 1:16 compression.

Columbia LCLM Paper — 8.8x Time-to-First-Token Speedup Using 1:16 Context Compression Ratios

Columbia's Latent Context Language Models achieve 8.8x faster time-to-first-token by compressing context at ratios up to 1:16. The approach makes long-context document analysis and code review significantly faster withou...

Sources: di.gg · 10 hours, 38 minutes ago

DeepMind 'From AGI to ASI' — AGI Is One Human-Level System, ASI Is a Colony of Multiple AGIs in Coordination

Google DeepMind's formal paper defines AGI as a single human-level general system and ASI as a colony of multiple AGIs working together. The colony framework aligns with multi-agent AI development and has 83 retweets....

Sources: x.com · 10 hours, 38 minutes ago

Arcee AI Moves Entire Model and Dataset Catalog to Hugging Face Hub — First Major Lab to Abandon AWS S3 for AI-Native Storage

Arcee AI moved all models and datasets — public and private — to Hugging Face Private Storage in a multi-million dollar deal, becoming the first major American AI lab to abandon AWS S3 entirely. The move positions Huggin...

Sources: x.com · 10 hours, 38 minutes ago

Claude Code Nested Subagents Cap at 5 Levels Deep — Context Overflow in Complex Workflows Now Has an Architectural Solution

Boris Cherny shipped nested subagent support in Claude Code — child agents run in isolated contexts, passing summaries upward, capped at depth 5. This gives Claude Code an architectural fix for context overflow in multi-...

Sources: x.com · 10 hours, 38 minutes ago

Arcee AI Drops AWS S3 for Hugging Face — Multi-Million Dollar Deal Covers All Models and Datasets, Public and Private

Arcee AI became the first major American AI lab to replace AWS S3 with Hugging Face Private Storage in a multi-million dollar deal covering all models and datasets including private ones. Hugging Face CEO Clement Delangu...

Sources: x.com · 10 hours, 38 minutes ago

Sakana AI Launches Marlin Today — First Commercial Product Is an Autonomous Research Assistant for Enterprise

Sakana AI launched Marlin today, its first commercial product after two years as a pure research organization. The Tokyo lab co-founded by David Ha and Llion Jones transitions to market with an autonomous enterprise rese...

Sources: x.com · 10 hours, 38 minutes ago

Google Gemini 3.5 Flash Live Translate Launches — Real-Time Speech-to-Speech Across 70+ Languages, 2,000+ Bidirectional Pairs

Google launched Gemini 3.5 Flash Live Translate — real-time speech-to-speech translation across 70+ languages and 2,000+ bidirectional pairs, available now in the Gemini API and Google Translate, with Google Meet coming ...

Sources: x.com · 10 hours, 38 minutes ago

Tesla FSD Supervised Approved in Denmark — First Scandinavian Market Cleared, Rollout Begins Soon

Tesla Europe confirmed FSD Supervised received regulatory approval in Denmark, opening the first Scandinavian market for the feature. The news hit #1 most viewed (948K views) and #1 most liked (11.6K likes) on Digg.

Sources: x.com · 10 hours, 38 minutes ago

AI Hardware & Chips

China's $295B AI plan hits the semiconductor wall — data centers without frontier chips cannot close the capability gap.

China's $295B AI Build Cannot Buy Its Way Past the Chip Gap — Export Controls Cap What Infrastructure Spending Can Achieve

China's massive $295 billion AI infrastructure investment cannot solve the fundamental constraint of US export controls on advanced GPUs. The plan builds data centers but cannot equip them with frontier chips, making alg...

Sources: di.gg · 10 hours, 37 minutes ago

Reading Group: JUDGEMENTBENCH Comparing Rubric and Preference Evaluation — June 17 at 101 Second Street, 64 Attending

A reading group session on JUDGEMENTBENCH: Comparing Rubric and Preference Evaluation for Quality Assessment runs June 17 at 64 attendees at 101 Second Street in San Francisco, covering AI model evaluation methodology.

Sources: x.com · 145 hours, 40 minutes ago

AI, HR and the Future of Work Startup Showcase by Indeed and OMERS Ventures — June 9 in San Francisco

Indeed and OMERS Ventures are co-hosting an AI, HR, and Future of Work startup showcase and mixer on June 9 in San Francisco, featuring startups building at the intersection of AI and talent, hiring, and workforce manage...

Sources: x.com · 145 hours, 40 minutes ago

Cafe Cursor San Francisco — June 10 in San Francisco

Cafe Cursor San Francisco runs June 10 at 10:00 AM, providing a casual daytime community space for Cursor users to work, connect, and discuss AI coding workflows in a cafe-style format.

Sources: x.com · 145 hours, 40 minutes ago

Founder Breakfast — June 10 in San Francisco, High-Signal Morning Event

A high-signal Founder Breakfast runs June 10 at 9:00 AM in San Francisco, listed as High Signal on Mimetic's tracker. The curated morning format brings together serious founders for structured early-morning connection....

Sources: x.com · 145 hours, 40 minutes ago

Data in Motion: Streaming and CDC for AI and Agents — June 23 at AWS Builder Loft, 525 Market Street

Data in Motion: Streaming and CDC for AI and Agents runs June 23 at the AWS Builder Loft with 18 registered attendees, focusing on real-time data infrastructure for AI agents — change data capture and streaming pipelines...

Sources: x.com · 145 hours, 40 minutes ago

Day-Zero: Agents in the Wild — June 22 at 1900 Broadway, Oakland, 41 Attending

Day-Zero: Agents in the Wild runs June 22 at 1900 Broadway in Oakland with 41 confirmed attendees, focusing on practitioners who have deployed AI agents in production — sharing real-world experiences from the trenches.

Sources: x.com · 145 hours, 40 minutes ago

AI Builders x Padel — June 17 in San Francisco

AI Builders x Padel combines padel tennis with AI builder networking in San Francisco on June 17, offering a physical activity-based community event for the technical AI builder community.

Sources: x.com · 145 hours, 40 minutes ago

AI Agents

Claude Code nested subagents resolve context overflow at 5 levels deep; Arcee AI migrates its entire catalog to Hugging Face.

Arcee AI Moves Entire Model and Dataset Catalog to Hugging Face Hub — First Major Lab to Abandon AWS S3 for AI-Native Storage

Arcee AI moved all models and datasets — public and private — to Hugging Face Private Storage in a multi-million dollar deal, becoming the first major American AI lab to abandon AWS S3 entirely. The move positions Huggin...

Sources: x.com · 10 hours, 38 minutes ago

Claude Code Nested Subagents Cap at 5 Levels Deep — Context Overflow in Complex Workflows Now Has an Architectural Solution

Boris Cherny shipped nested subagent support in Claude Code — child agents run in isolated contexts, passing summaries upward, capped at depth 5. This gives Claude Code an architectural fix for context overflow in multi-...

Sources: x.com · 10 hours, 38 minutes ago

Rork Games × Fable 5 One-Shots 3D/2D Multiplayer Games — Fortnite-Style in a Single Agent Session

Rork launches Rork Games × Claude Fable 5: one-shots 3D and 2D games playable with friends online, including Fortnite-style multiplayer. Demonstrates nested subagent architecture producing complete interactive game exper...

Sources: x.com · 72 hours, 43 minutes ago

Claude Code Nested Subagents Live — Agents Spawn Agents Up to 5 Levels Deep

Claude Code lead Boris Cherny ships nested subagent support, capped at 5 levels. Agents can now spawn agents recursively to manage context in complex workflows. Fable 5 being used for planning with Claude Code subagents ...

Sources: x.com · 74 hours, 21 minutes ago

OpenAI Codex Rate Limit Banking Launches — Save Resets, Spend Them on Your Schedule

OpenAI launched rate limit banking for Codex: save unused compute resets and deploy them when you actually need them. Plus a referral program — invite 3 friends to try Codex, both earn a banked reset. The announcement pu...

Sources: AI Companies #1, x.com · 86 hours, 45 minutes ago

OpenAI Buys Ona to Run Codex in the Cloud When Your Laptop Is Closed

OpenAI acquires Ona (@ona_hq) — the startup's secure cloud execution technology joins the Codex team to let coding agents run longer-horizon tasks even when laptops are closed

Sources: AI Newsmakers, x.com · 99 hours, 35 minutes ago

Claude Code Gets 5-Level Nested Subagents to Solve Context Overflow

Anthropic Claude Code lead Boris Cherny ships nested subagent support — agents spawning agents up to 5 levels deep, each running in an isolated context window with only summaries passed back to parent

Sources: AI Companies #2, di.gg · 99 hours, 35 minutes ago

Full Sail Deploy: AI Builder Networking on San Francisco Bay — June 17 at Pier 39

Full Sail Deploy runs June 17 at Adventure Cat Sailing Charters at Pier 39, offering AI builders a unique networking experience on San Francisco Bay while sailing — a memorable venue for community building.

Sources: x.com · 145 hours, 40 minutes ago

AI Infrastructure

Arcee AI abandons AWS S3 for Hugging Face in a multi-million dollar exclusive deal; China's $295B AI build confronts chip limits.

China's $295 Billion AI Plan Targets Data Center Build-Out — But Chip Constraints Cap What Infrastructure Spending Can Achieve

China's 5-year, $295 billion AI infrastructure plan is the largest national AI infrastructure commitment ever announced. Analysts argue the spend cannot overcome US export controls on advanced chips, making algorithmic e...

Sources: di.gg · 10 hours, 38 minutes ago

Arcee AI Replaces AWS S3 with Hugging Face in Multi-Million Dollar Deal — First Major US Lab to Migrate Entire Catalog

Arcee AI migrated all models and datasets — including private ones — from AWS S3 to Hugging Face Private Storage in a multi-million dollar exclusive deal. The first major US AI lab to abandon AWS S3 entirely for AI-nativ...

Sources: x.com · 10 hours, 38 minutes ago

Google Colab CLI Launches — Terminal GPU Provisioning and Agent Skills Without Opening a Browser

Google launched Colab CLI, giving developers terminal-based access to GPU provisioning directly from their local environment. The CLI ships with agent skill integrations, making Colab a programmable compute backend for A...

Sources: AI Companies #1, di.gg · 86 hours, 45 minutes ago

Arcee AI Drops AWS S3 for Hugging Face in Multi-Million Dollar Hosting Deal

Arcee AI terminates AWS S3 relationship, signs a multi-million dollar exclusive model hosting deal with Hugging Face Private Storage — Arcee SVP says 'The HF infra is a no-brainer'

Sources: AI Newsmakers, di.gg · 99 hours, 35 minutes ago

Goldman and JPMorgan Move to Financialize GPU Compute — Futures Markets Coming

Goldman Sachs and JPMorgan exploring futures tied to GPU rental prices per The Information. The instruments would let AI companies hedge compute costs the way airlines hedge fuel — locking in future prices to manage vola...

Sources: AI Newsmakers, x.com · 145 hours, 52 minutes ago

SpaceX AI1: First AI Satellite Is Real Hardware

SpaceX unveils AI1, its first AI satellite, carrying 150,000 watts of compute payload in orbit — first hardware step toward 100 GW/year orbital AI compute goal starting 2028

Sources: AI Leaders #1, x.com · 162 hours, 57 minutes ago

SpaceX Signs $920M/Month Google Compute Deal — $11 Billion Per Year

SpaceX signs $920 million per month compute agreement with Google — $11 billion per year deal to provide computing power

Sources: AI Newsmakers, x.com · 218 hours, 0 minutes ago

Anthropic Hits Compute and Energy Wall as Demand Overwhelms Supply

Anthropic is running out of compute and energy, according to a widely-shared report spreading across AI research communities

Sources: AI Newsmakers, x.com · 237 hours, 39 minutes ago

AI in Science

Google DeepMind formally defines AGI and ASI; Radical Numerics launches with $50M seed to build general biological intelligence using generative genomics.

Radical Numerics Emerges from Stealth with $50M Seed to Build General Biological Intelligence — Founding Team Created Field of Generative Genomics

Radical Numerics launched from stealth with a $50M seed round, building general biological intelligence with a dual mandate: human health design and biodefense. The founding team, including Michael Poli, created the fiel...

Sources: x.com · 2 hours, 52 minutes ago

DeepMind Defines the AGI-to-ASI Gap — AGI Is One Human-Level System, ASI Is What Happens When Multiple AGIs Coordinate

Google DeepMind published a paper formally defining the AGI-to-ASI transition: AGI is one human-level general system, and ASI is a colony of multiple AGIs working together. The colony definition has significant implicati...

Sources: x.com · 10 hours, 37 minutes ago

Radical Numerics Emerges from Stealth with $50M Seed — General Biological Intelligence with Generative Genomics and Biodefense Mandate

Radical Numerics launches from stealth, $50M seed, building general biological intelligence. Founding team created generative genomics. Dual mandate: human health design + biodefense. Co-founders include Michael Poli (SS...

Sources: AI Investors, x.com · 2 hours, 52 minutes ago

CVPR 2026 Historic First: No University Best Papers as Compute Gap Completes Academic Displacement

CVPR 2026 awards zero university best papers — first time in conference's 40+ year history. All awards went to industry-supported research. GPU compute gap is now an insurmountable barrier for underfunded academic labs....

Sources: x.com · 74 hours, 21 minutes ago

CVPR 2026 Awards Zero University Best Papers — GPU Shortage Prices Academia Out of Its Own Field

CVPR 2026 wrapped in Denver with zero university best paper awards — reported as a first. Synthesia co-founder Matthias Niessner attributed the outcome directly to GPU shortages: academic labs can no longer afford the co...

Sources: AI Newsmakers, di.gg · 86 hours, 45 minutes ago

Fireside Discussion: From Tokens to Robots — June 9 at 625 2nd Street, San Francisco, 490 Attending

A fireside discussion on the journey From Tokens to Robots runs today June 9 at 4:30 PM PT at 625 2nd Street in San Francisco, with 490 confirmed attendees — one of the highest-attendance events this week.

Sources: x.com · 145 hours, 40 minutes ago

GTM Eng SF: Claude Code + GTM Lightning Talks — June 9 at SoMa, Exa, LangChain, AssemblyAI and More

GTM Eng SF runs today June 9 in San Francisco with Claude Code-focused GTM lightning talks from leaders at Exa, LangChain, AssemblyAI, and other AI companies. 91 attendees confirmed for this practitioner-focused evening....

Sources: x.com · 145 hours, 40 minutes ago

Fork It, Branch It, Bop It: New Primitives for Agents — June 9 at PlanetScale, SoMa, 158 Attending

Running today June 9 at PlanetScale's office at 108 Natoma Street in SoMa, this technical meetup explores new primitives for AI agents with 158 confirmed attendees. Topics include branching, forking, and state management...

Sources: x.com · 145 hours, 40 minutes ago

Videos

Google's Live Translate demo shows real-time 70-language speech-to-speech; SpaceX AI1 orbital compute satellite unveiled at 150kW.

GLM 5.2 Takes BridgeBench #1 at 42.8 Reasoning After Fable Ban — 'You Cannot Export Control Your Way Out of Open Source'

Zhipu AI's GLM 5.2 took the #1 spot on BridgeBench scoring 100.0 on coding and 42.8 on reasoning — beating Fable 5 two days after the US banned it. At one-tenth the cost and 300 tokens per second. 5,906 likes, 649K views...

Sources: x.com · 10 hours, 29 minutes ago

Google Gemini 3.5 Flash Live Translate Demo — Real-Time Speech-to-Speech Translation Across 70 Languages

Google published a demo of Gemini 3.5 Flash Live Translate showing real-time speech-to-speech translation across 70+ languages with natural intonation and no perceptible lag. Available now in the Gemini API and Google Tr...

Sources: x.com · 10 hours, 29 minutes ago

Tech Weekend Silicon Valley: AI Startup Pitch Competition — June 18-19 at Allied Arts Guild, Menlo Park

Tech Weekend Silicon Valley runs June 18-19 at Allied Arts Guild in Menlo Park with Founder-VC Matchmaking, Deep Tech Pitch Competition, and AI Startup Pitch Competition events across both days.

Sources: x.com · 145 hours, 40 minutes ago

Scalekit Hackathon — Build Day SF, June 27 in San Francisco

Scalekit is hosting a Build Day hackathon in San Francisco on June 27, inviting developers to build integrations and applications on top of Scalekit's enterprise authentication and identity infrastructure for AI applicat...

Sources: x.com · 145 hours, 40 minutes ago

Artificial Analysis Coding Agent Benchmarks — June 11 in San Francisco, 262 Attending

Artificial Analysis hosts a session on coding agent benchmarks June 11 in San Francisco with 262 confirmed attendees, presenting independent evaluation data on how AI coding agents perform across real-world programming t...

Sources: x.com · 145 hours, 40 minutes ago

Build with Cursor: San Francisco — June 11 at SoMa, 206 Attending

Build with Cursor San Francisco runs June 11 with 206 confirmed attendees, focused on developers building with Cursor's AI-powered code editor. The event combines product learning and community building for Cursor's grow...

Sources: x.com · 145 hours, 40 minutes ago

AI Supply Chain Hackathon 2026: Food Banks + AI — June 24 at Plug and Play Tech Center

The AI Supply Chain Hackathon 2026 runs June 24 at Plug and Play Tech Center, focusing on building AI tools for food bank supply chain optimization. Teams compete to apply AI to real-world logistics and distribution chal...

Sources: x.com · 145 hours, 40 minutes ago

Midsummer Multimodal AI Hackathon — June 19 in San Francisco

The Midsummer Multimodal AI Hackathon runs June 19 in San Francisco, challenging participants to build applications combining vision, language, audio, and other modalities using the latest multimodal AI capabilities.

Sources: x.com · 145 hours, 40 minutes ago

AI Policy & Regulation

A third party builds the Fable compliance gate the government has been requesting; China's $295B AI plan faces chip access constraints.

China $295 Billion AI Infrastructure Plan Faces Fundamental Chip Constraint — Five-Year Build Cannot Solve Access to Advanced GPUs

China's $295 billion, 5-year AI infrastructure plan targets national data center construction but faces a constraint capital cannot solve: no access to advanced GPUs under US export controls. Analysts question whether sp...

Sources: di.gg · 10 hours, 38 minutes ago

Fable Compliance Gate Reportedly Built by Third Party — Anthropic Now Has a Technical Path to Restoring Access

A third party claims to have built the compliance gate the US government has been requesting from Anthropic to restore Fable 5 and Mythos. If Anthropic adopts it, the export control order could be lifted.

Sources: di.gg · 10 hours, 38 minutes ago

Third Party Builds Fable Compliance Gate — Technical Barrier to Restoring Fable 5 and Mythos Reportedly Removed

A third party claims to have built what the US government requested from Anthropic: a gate restricting Fable 5 and Mythos to authorized users only. If adopted, the export control order enabling Fable's return may be sati...

Sources: AI Leaders #1, x.com · 10 hours, 34 minutes ago

Sacks' 8-Month Prediction Vindicated — 8,389 Likes: Anthropic's Safety Strategy Was Regulatory Capture

Sacks June 11 post (8,389 likes, 700 RTs): 'About 8 months ago, I warned Anthropic is running a sophisticated regulatory capture strategy based on fear-mongering. This take was controversial at the time; now look how man...

Sources: AI Leaders #1, x.com · 48 hours, 58 minutes ago

Sacks: Dario Refused to Fix Jailbreak — Government Asked First, Anthropic Said No

Trump AI Czar David Sacks: The Admin asked Dario to fix the jailbreak or de-deploy Fable. Dario refused. The export controls followed. This is NOT a misunderstanding — it's a standoff Anthropic chose. (872 likes, Robert ...

Sources: AI Leaders #1, x.com · 48 hours, 58 minutes ago

Amazon CEO Jassy Called the Government — Anthropic's Own $4B Investor Triggered the Shutdown

Amazon CEO Andy Jassy personally warned senior Trump administration officials about security risks in Anthropic's Claude models — helping trigger the export control directive that shut them down. Amazon is a $4B Anthropi...

Sources: AI Leaders #1, x.com · 49 hours, 56 minutes ago

Lutnick Sent Letter Directly to Dario Amodei — Personal Government Directive to Anthropic CEO

Commerce Secretary Lutnick sent a letter directly to Dario Amodei to order the Fable/Mythos suspension. Personal directive from the Trump administration's Commerce Secretary to Anthropic's CEO — on a Friday night....

Sources: AI Community #2, x.com · 50 hours, 51 minutes ago

China Distillation Theory: Government Targeted Fable to Block Chinese AI Training Pipeline

Real reason for Fable shutdown emerging: prevent Chinese labs from distilling training data off Fable, not general national security. 'Export controls were the fastest tool available. Might buy 9 months.'

Sources: AI Leaders #1, x.com · 50 hours, 51 minutes ago

Jobs

AI Research Engineer openings in pre-training and inference kernel dev; clinical AI startup hiring across all research tracks.

Clinical AI Startup Hiring Across All Tracks — Pre-Training, Post-Training, Agents, and Researchers Working on Cancer Applications

A clinical AI startup is hiring AI researchers across all technical tracks including pre-training, scaling, post-training, agents, and direct disease research applications. Working on actual clinical trial data and propr...

Sources: x.com · 4 hours, 53 minutes ago

AI Research Engineer Openings in Pre-Training and Inference — Kernel Dev, All Parallelism Flavors, ASIC/FPGA/GPU Targeting

New AI Research Engineer openings in two tracks: Pre-training (kernel dev, all parallelism, fault tolerance) and Inference (kernel dev, all devices including ASIC, FPGA, GPU). Research-level roles shaping the technical d...

Sources: x.com · 4 hours, 53 minutes ago

Frontier / RSI RL Environments Hackathon at Y Combinator — June 20, High-Signal Event

A high-signal Frontier / RSI RL Environments Hackathon runs June 20 at Y Combinator's San Francisco offices at 560 20th Street. The event focuses on building reinforcement learning environments for frontier AI research a...

Sources: x.com · 145 hours, 40 minutes ago

ClickHouse + Hex AI Hackathon — June 11 at 667 Howard Street, San Francisco, 312 Attending

ClickHouse and Hex are hosting an AI hackathon on June 11 at 667 Howard Street in San Francisco with 312 confirmed attendees. The hackathon focuses on building AI applications on top of ClickHouse's analytics database an...

Sources: x.com · 145 hours, 40 minutes ago

Data+AI Summit Pre-Event: Open-Weight Models on Amazon Bedrock — June 9 at AWS Builder Loft, 227 Attending

A Data+AI Summit pre-event running June 9 at the AWS Builder Loft on 525 Market Street features Artificial Analysis, MiniMax, Nvidia, and AI startups discussing open-weight models on Amazon Bedrock. 227 attendees confirm...

Sources: x.com · 145 hours, 40 minutes ago

AI TechWorld 2026 — September 1 at Santa Clara Convention Center

AI TechWorld 2026 runs September 1 at the Santa Clara Convention Center, marking the start of fall AI conference season with a broad-focus event covering AI technology across enterprise, research, and startup segments.

Sources: x.com · 145 hours, 40 minutes ago

AI and Big Data Expo North America — June 16 at Santa Clara Convention Center

The AI and Big Data Expo North America runs June 16 at the Santa Clara Convention Center, co-located with the Data+AI Summit week for maximum practitioner attendance.

Sources: x.com · 145 hours, 40 minutes ago

AI Infra Summit — September 14 at Santa Clara Convention Center

The AI Infra Summit runs September 14 at the Santa Clara Convention Center, focused on the infrastructure layer powering AI at scale: compute, networking, storage, and the systems engineering required to deploy frontier ...

Sources: x.com · 145 hours, 40 minutes ago

Anomalies

The Mistral 'Le Chaton Fat' fake leak became real naming pressure — community campaigns to make the meme the official model name.

Mistral Community Campaigns for 'Le Chaton Fat' as Official Model Name — A Fake Rumor Became a Naming Pressure Campaign

A fake leak claiming Mistral was releasing a model called 'Le Chaton Fat' spread across AI Twitter, accumulating enough community attachment that people are now lobbying Mistral to actually use the name. AI humor has bec...

Sources: x.com · 10 hours, 37 minutes ago

GovAI Coalition Summit 2026 — December 9 at San Jose Convention Center

The GovAI Coalition Summit runs December 9 at the San Jose Convention Center, bringing together AI governance researchers, policymakers, and technologists for the year's capstone AI policy event.

Sources: x.com · 145 hours, 40 minutes ago

WeAreDevelopers World Congress North America — September 23 at San Jose McEnery Convention Center

WeAreDevelopers World Congress North America runs September 23 at the San Jose McEnery Convention Center, featuring Fireworks, GitHub, Warp, Databricks, Augment Code, Stripe, Sentry, Microsoft, Block, and more.

Sources: x.com · 145 hours, 40 minutes ago

PyTorch Conference — October 20 at San Jose McEnery Convention Center

The PyTorch Conference runs October 20 at the San Jose McEnery Convention Center, bringing together the global PyTorch developer community for technical sessions on the most widely used open-source deep learning framewor...

Sources: x.com · 145 hours, 40 minutes ago

AGNTCon + MCPCon — October 22 at San Jose McEnery Convention Center

AGNTCon and MCPCon co-locate at the San Jose McEnery Convention Center on October 22, creating a dedicated conference for agent developers and MCP practitioners — two of the most active communities in current AI developm...

Sources: x.com · 145 hours, 40 minutes ago

The AI Conference 2026: Shaping the Future of AI — September 28 at Pier 48, San Francisco

The AI Conference 2026 runs September 28 at Pier 48 in San Francisco, focusing on shaping the future direction of AI development across research, policy, safety, and commercial deployment.

Sources: x.com · 145 hours, 40 minutes ago

Agentic AI Summit at UC Berkeley — August 1, High-Signal Academic and Industry Event

The Agentic AI Summit runs August 1 at UC Berkeley, listed as High Signal. The summit focuses specifically on agentic AI systems at the intersection of academic research and industry deployment — a rapidly evolving front...

Sources: x.com · 145 hours, 40 minutes ago

AGI Summit and GenAI Summit 2026 — July 18-19 at Palace of Fine Arts, San Francisco

The AGI Summit and GenAI Summit co-locate at the Palace of Fine Arts in San Francisco July 18-19, listed as High Signal. The joint event brings together frontier AI researchers and practitioners focused on the path towar...

Sources: x.com · 145 hours, 40 minutes ago

Big events

Major AI conferences and global summits happening in 2026 — from CVPR Denver to NeurIPS and beyond.

Data+AI Summit Pre-Event: Open-Weight Models on Amazon Bedrock — June 9 at AWS Builder Loft, 227 Attending

A Data+AI Summit pre-event running June 9 at the AWS Builder Loft on 525 Market Street features Artificial Analysis, MiniMax, Nvidia, and AI startups discussing open-weight models on Amazon Bedrock. 227 attendees confirm...

Sources: x.com · 145 hours, 40 minutes ago

AI TechWorld 2026 — September 1 at Santa Clara Convention Center

AI TechWorld 2026 runs September 1 at the Santa Clara Convention Center, marking the start of fall AI conference season with a broad-focus event covering AI technology across enterprise, research, and startup segments.

Sources: x.com · 145 hours, 40 minutes ago

AI and Big Data Expo North America — June 16 at Santa Clara Convention Center

The AI and Big Data Expo North America runs June 16 at the Santa Clara Convention Center, co-located with the Data+AI Summit week for maximum practitioner attendance.

Sources: x.com · 145 hours, 40 minutes ago

AI Infra Summit — September 14 at Santa Clara Convention Center

The AI Infra Summit runs September 14 at the Santa Clara Convention Center, focused on the infrastructure layer powering AI at scale: compute, networking, storage, and the systems engineering required to deploy frontier ...

Sources: x.com · 145 hours, 40 minutes ago

GovAI Coalition Summit 2026 — December 9 at San Jose Convention Center

The GovAI Coalition Summit runs December 9 at the San Jose Convention Center, bringing together AI governance researchers, policymakers, and technologists for the year's capstone AI policy event.

Sources: x.com · 145 hours, 40 minutes ago

WeAreDevelopers World Congress North America — September 23 at San Jose McEnery Convention Center

WeAreDevelopers World Congress North America runs September 23 at the San Jose McEnery Convention Center, featuring Fireworks, GitHub, Warp, Databricks, Augment Code, Stripe, Sentry, Microsoft, Block, and more.

Sources: x.com · 145 hours, 40 minutes ago

PyTorch Conference — October 20 at San Jose McEnery Convention Center

The PyTorch Conference runs October 20 at the San Jose McEnery Convention Center, bringing together the global PyTorch developer community for technical sessions on the most widely used open-source deep learning framewor...

Sources: x.com · 145 hours, 40 minutes ago

AGNTCon + MCPCon — October 22 at San Jose McEnery Convention Center

AGNTCon and MCPCon co-locate at the San Jose McEnery Convention Center on October 22, creating a dedicated conference for agent developers and MCP practitioners — two of the most active communities in current AI developm...

Sources: x.com · 145 hours, 40 minutes ago

Hackathons

Upcoming AI build competitions, hackathons, and developer events worldwide.

GTM Eng SF: Claude Code + GTM Lightning Talks — June 9 at SoMa, Exa, LangChain, AssemblyAI and More

GTM Eng SF runs today June 9 in San Francisco with Claude Code-focused GTM lightning talks from leaders at Exa, LangChain, AssemblyAI, and other AI companies. 91 attendees confirmed for this practitioner-focused evening....

Sources: x.com · 145 hours, 40 minutes ago

Fork It, Branch It, Bop It: New Primitives for Agents — June 9 at PlanetScale, SoMa, 158 Attending

Running today June 9 at PlanetScale's office at 108 Natoma Street in SoMa, this technical meetup explores new primitives for AI agents with 158 confirmed attendees. Topics include branching, forking, and state management...

Sources: x.com · 145 hours, 40 minutes ago

Tech Weekend Silicon Valley: AI Startup Pitch Competition — June 18-19 at Allied Arts Guild, Menlo Park

Tech Weekend Silicon Valley runs June 18-19 at Allied Arts Guild in Menlo Park with Founder-VC Matchmaking, Deep Tech Pitch Competition, and AI Startup Pitch Competition events across both days.

Sources: x.com · 145 hours, 40 minutes ago

Scalekit Hackathon — Build Day SF, June 27 in San Francisco

Scalekit is hosting a Build Day hackathon in San Francisco on June 27, inviting developers to build integrations and applications on top of Scalekit's enterprise authentication and identity infrastructure for AI applicat...

Sources: x.com · 145 hours, 40 minutes ago

Artificial Analysis Coding Agent Benchmarks — June 11 in San Francisco, 262 Attending

Artificial Analysis hosts a session on coding agent benchmarks June 11 in San Francisco with 262 confirmed attendees, presenting independent evaluation data on how AI coding agents perform across real-world programming t...

Sources: x.com · 145 hours, 40 minutes ago

Build with Cursor: San Francisco — June 11 at SoMa, 206 Attending

Build with Cursor San Francisco runs June 11 with 206 confirmed attendees, focused on developers building with Cursor's AI-powered code editor. The event combines product learning and community building for Cursor's grow...

Sources: x.com · 145 hours, 40 minutes ago

AI Supply Chain Hackathon 2026: Food Banks + AI — June 24 at Plug and Play Tech Center

The AI Supply Chain Hackathon 2026 runs June 24 at Plug and Play Tech Center, focusing on building AI tools for food bank supply chain optimization. Teams compete to apply AI to real-world logistics and distribution chal...

Sources: x.com · 145 hours, 40 minutes ago

Midsummer Multimodal AI Hackathon — June 19 in San Francisco

The Midsummer Multimodal AI Hackathon runs June 19 in San Francisco, challenging participants to build applications combining vision, language, audio, and other modalities using the latest multimodal AI capabilities.

Sources: x.com · 145 hours, 40 minutes ago

Dinners

Upcoming AI dinners, meetups, and community gatherings — SF Bay Area and global.

Reading Group: JUDGEMENTBENCH Comparing Rubric and Preference Evaluation — June 17 at 101 Second Street, 64 Attending

A reading group session on JUDGEMENTBENCH: Comparing Rubric and Preference Evaluation for Quality Assessment runs June 17 at 64 attendees at 101 Second Street in San Francisco, covering AI model evaluation methodology.

Sources: x.com · 145 hours, 40 minutes ago

AI, HR and the Future of Work Startup Showcase by Indeed and OMERS Ventures — June 9 in San Francisco

Indeed and OMERS Ventures are co-hosting an AI, HR, and Future of Work startup showcase and mixer on June 9 in San Francisco, featuring startups building at the intersection of AI and talent, hiring, and workforce manage...

Sources: x.com · 145 hours, 40 minutes ago

Cafe Cursor San Francisco — June 10 in San Francisco

Cafe Cursor San Francisco runs June 10 at 10:00 AM, providing a casual daytime community space for Cursor users to work, connect, and discuss AI coding workflows in a cafe-style format.

Sources: x.com · 145 hours, 40 minutes ago

Founder Breakfast — June 10 in San Francisco, High-Signal Morning Event

A high-signal Founder Breakfast runs June 10 at 9:00 AM in San Francisco, listed as High Signal on Mimetic's tracker. The curated morning format brings together serious founders for structured early-morning connection....

Sources: x.com · 145 hours, 40 minutes ago

Data in Motion: Streaming and CDC for AI and Agents — June 23 at AWS Builder Loft, 525 Market Street

Data in Motion: Streaming and CDC for AI and Agents runs June 23 at the AWS Builder Loft with 18 registered attendees, focusing on real-time data infrastructure for AI agents — change data capture and streaming pipelines...

Sources: x.com · 145 hours, 40 minutes ago

Day-Zero: Agents in the Wild — June 22 at 1900 Broadway, Oakland, 41 Attending

Day-Zero: Agents in the Wild runs June 22 at 1900 Broadway in Oakland with 41 confirmed attendees, focusing on practitioners who have deployed AI agents in production — sharing real-world experiences from the trenches.

Sources: x.com · 145 hours, 40 minutes ago

AI Builders x Padel — June 17 in San Francisco

AI Builders x Padel combines padel tennis with AI builder networking in San Francisco on June 17, offering a physical activity-based community event for the technical AI builder community.

Sources: x.com · 145 hours, 40 minutes ago

Full Sail Deploy: AI Builder Networking on San Francisco Bay — June 17 at Pier 39

Full Sail Deploy runs June 17 at Adventure Cat Sailing Charters at Pier 39, offering AI builders a unique networking experience on San Francisco Bay while sailing — a memorable venue for community building.

Sources: x.com · 145 hours, 40 minutes ago

For Google Notebook LM

Utility section for NotebookLM output.

Price Reversal in Reasoning Language Models — Research on True Cost of AI APIs

The reasoning language model price reversal analysis tracks how model economics are changing as inference cost, capability, and product packaging shift.

Sources: x.com · 1919 hours, 43 minutes ago

SlopCodeBench: AI Agents Generate Progressively Worse Code — Research for Analysis

SlopCodeBench studies how AI agents can generate progressively worse code, underscoring the need for external tests, review loops, and verification systems.

Sources: x.com · 1919 hours, 43 minutes ago

Stanford NLP: Sycophantic LLMs Cause Social Harm — Research Paper for Deep Analysis

Stanford NLP’s sycophancy research examines how agreeable model behavior can create social harm, especially when users rely on AI for judgment or advice.

Sources: x.com · 1919 hours, 43 minutes ago

Google TurboQuant Paper — Technical Deep Dive on LLM KV Cache Compression

Google TurboQuant’s technical analysis explains how KV cache compression can change LLM serving costs, throughput, and memory requirements for production systems.

Sources: x.com · 1919 hours, 43 minutes ago

Apollo Research and OpenAI Metagaming Study — Deep Dive for NotebookLM Analysis

Apollo Research and OpenAI’s metagaming study examines how advanced models may behave under evaluation, making it relevant to agent safety and benchmark reliability.

Sources: x.com · 1919 hours, 43 minutes ago

GenAI Summit SF 2026 — July 17 at Palace of Fine Arts, High-Signal Annual Summit

GenAI Summit SF returns July 17 at the Palace of Fine Arts in San Francisco, listed as High Signal on Mimetic. The summit is one of the anchor generative AI events of the summer conference season.

Sources: x.com · 145 hours, 40 minutes ago

Physical AI Superconnect 2026 — June 24 at Computer History Museum, Mountain View, 201 Attending

Physical AI Superconnect 2026 runs June 24 at the Computer History Museum in Mountain View, with 201 confirmed attendees. The event focuses on physical AI, embodied systems, and the intersection of AI with the physical w...

Sources: x.com · 145 hours, 40 minutes ago

AI for Science Summit in SF — June 16 at SoMa, High-Signal Technical Research Event

The AI for Science Summit runs June 16 in San Francisco with 69 confirmed attendees, listed as High Signal on Mimetic. The summit focuses on AI applications in scientific research — a rapidly growing intersection of fron...

Sources: x.com · 145 hours, 40 minutes ago

Get This as a Daily Newsletter

Newsletter subscribe section.

localhost:2026 SF — June 18 at The St. Regis, High-Signal Developer Conference

localhost:2026 SF is a high-signal developer conference running June 18 at The St. Regis San Francisco, 125 3rd St. Listed as High Signal on Mimetic's tracker, this event targets serious AI developer practitioners.

Sources: x.com · 145 hours, 40 minutes ago

AI Engineer World's Fair — June 28 to July 2 at Moscone Center, San Francisco

The AI Engineer World's Fair returns for its fourth year, June 28 to July 2 at Moscone Center in San Francisco. Microsoft, GitHub, and major AI labs are participating with talks, labs, and workshops for AI engineers....

Sources: x.com · 145 hours, 40 minutes ago

Vector Space Day San Francisco — June 11 at The Midway, 622 Attendees, 400+ Engineers

Vector Space Day 2026 brings 400+ engineers, researchers, and AI builders to The Midway in San Francisco on June 11 for a full-day technical conference on retrieval, agents, memory, and edge AI. Organized by Qdrant, feat...

Sources: x.com · 145 hours, 40 minutes ago

Data + AI Summit 2026 by Databricks — June 16-18 at Moscone Center, San Francisco

Databricks' Data + AI Summit 2026 runs June 16-18 at Moscone Center in San Francisco, bringing together data, development, and AI teams to level up on production data and AI infrastructure. AI agents, real-time pipelines...

Sources: x.com · 145 hours, 40 minutes ago

Signals

Radical Numerics Emerges from Stealth with $50M Seed — General Biological Intelligence with Generative Genomics and Biodefense Mandate

68% EngagementScienceBullish

Radical Numerics launches from stealth, $50M seed, building general biological intelligence. Founding team created generative genomics. Dual mandate: human health design + biodefense. Co-founders include Michael Poli (SS...

Sources: AI Investors, x.com · 2 hours, 52 minutes ago

Salesforce Acquires Fin (Formerly Intercom) for $3.6 Billion — Customer Service AI Category Leader Joins Agentforce

90% EngagementFundingVc

Salesforce agreed to acquire Fin AI (formerly Intercom) for ~$3.6B. CEO Eoghan McCabe announced the deal Monday morning — 2,793 likes, 419K views. Fin is the #1 AI customer agent; the deal adds Fin and its Apex model to ...

Sources: AI Companies #2, x.com · 2 hours, 57 minutes ago

OpenAI and Anthropic Both File Confidential S-1s in the Same Week — The AI IPO Era Has Begun

72% EngagementFundingVc

Both OpenAI ($852B valuation) and Anthropic filed confidential S-1 IPO registrations within days of each other. No timelines set. Whichever goes public first sets the AI valuation benchmark. The era of AI IPO speculation...

Sources: AI Influencers, x.com · 10 hours, 34 minutes ago

Third Party Builds Fable Compliance Gate — Technical Barrier to Restoring Fable 5 and Mythos Reportedly Removed

55% EngagementPolicySignal

A third party claims to have built what the US government requested from Anthropic: a gate restricting Fable 5 and Mythos to authorized users only. If adopted, the export control order enabling Fable's return may be sati...

Sources: AI Leaders #1, x.com · 10 hours, 34 minutes ago

DeepMind Formally Defines AGI and ASI — AGI Is One Human-Level System, ASI Is a Colony of Multiple AGIs

58% EngagementResearchSignal

Google DeepMind paper 'From AGI to ASI' defines AGI as a single human-level general system and ASI as a colony of multiple AGIs working in coordination. The colony definition changes how we think about the ASI timeline a...

Sources: AI Newsmakers, x.com · 10 hours, 34 minutes ago

Sakana AI Launches First Commercial Product — Research Lab Crosses into Enterprise Market

60% EngagementProductsBullish

Sakana AI (co-founded by David Ha, Llion Jones) launched Marlin today — its first commercial product after 2 years of research. An autonomous enterprise research assistant. 544 likes on launch. The Tokyo lab transitions ...

Sources: AI Companies #1, x.com · 10 hours, 34 minutes ago

Sacks' 8-Month Prediction Vindicated — 8,389 Likes: Anthropic's Safety Strategy Was Regulatory Capture

95% EngagementPolicySignal

Sacks June 11 post (8,389 likes, 700 RTs): 'About 8 months ago, I warned Anthropic is running a sophisticated regulatory capture strategy based on fear-mongering. This take was controversial at the time; now look how man...

Sources: AI Leaders #1, x.com · 48 hours, 58 minutes ago

Sacks: Dario Refused to Fix Jailbreak — Government Asked First, Anthropic Said No

99% EngagementPolicyCritical

Trump AI Czar David Sacks: The Admin asked Dario to fix the jailbreak or de-deploy Fable. Dario refused. The export controls followed. This is NOT a misunderstanding — it's a standoff Anthropic chose. (872 likes, Robert ...

Sources: AI Leaders #1, x.com · 48 hours, 58 minutes ago