Tech News

Today's Highlights

Former Anthropic and DeepMind lead Behnam Neyshabur launches Mirendil with $200 million to automate frontier AI engineering

+42 more

It’s a new day at Digg. Fresh stories are clustering now. 65% complete.

Video Signals

#1257

CRED Founder Says AI Now Writes 90% Of Code At Fintech Startup

2h ago7.8K3515

#1360

Dreamina Octo Vibe Create Workflow Generates Fantasy Short Film

4h ago9.2K3714

#710

Qualcomm Unveils Agentic AI Partnerships With Hugging Face And Modular

1h ago4.1K00

Robots Build Humanoid Robots in Automated Factory Assembly Line

1h ago84883

Figure Scales Humanoid Robot Testing With Parallel Head Calibration

2h ago1.1K93

Manufacturers Improve AI Outcomes By Cutting Latency In Operations

2h ago1.3K41

Google Unveils Gemini Omni For Chat-Based AI Video Editing

3h ago1.8K60

Paragraphica AI Camera Captures Data Instead of Reality

4h ago1.6K73

Perseverance Rover Completes First AI-Planned Drive on Mars

5h ago94710

Humanoid Robots Serve Coffee, Ice Cream in Modular Retail Kiosk

12m ago36900

Extropic AI founder Guillaume Verdon backs Elon Musk's call to rebuild X's recommendation algorithm, while t3.gg creator Theo defends itTheo argues the system adaptively tracks real-time tech developments

4h15.2k2815

#447

Top post: @beffjezos “@elonmusk @karpathy @kuroke01 @gallabytes I will literally come to X HQ, sign any NDA, and help your team design a new algorithm from first principles with LLMs + latent graph diffusion, pro bono.”

AI technologist @deepfates requests community recommendations for digital audio workstations, drawing an endorsement for AbletonCreator @viemccoy recommended the software as a comprehensive solution

5h17.3k15935

#1014

Top post: @theo “@deepfates Ableton is correct FL Studio is more fun”

New github stars (48hrs)

CEO-Bench evaluates long-horizon LLM agents via 500-day startup simulations with business databases and market dynamics

zlab-princeton/ceobench-src291d

Simulates 500-day startup operations to benchmark long-horizon LLM agents via business databases, tools, and market events.

1014

802

Swift adapter exposing llama.cpp GGUF models through Apple's LanguageModel protocol

pcuenca/LlamaLanguageModels51d

Implements Apple's Foundation Models LanguageModel API for llama.cpp GGUF models in Swift.

1014

861

Collection of normalized two-bar breakbeat audio samples

switchangel/breaks901d

Hosts normalized two-bar breakbeat WAV samples plus Strudel config file.

1014

1398

Lightweight Python library for zero-shot chain-of-thought evaluation of language models on standard benchmarks

openai/simple-evals4.5k1d

Evaluates language models on benchmarks like MMLU, GPQA, MATH, and SimpleQA using zero-shot chain-of-thought prompting.

1014

144

Lightweight graph-based vector index for private offline RAG on personal devices with 97% storage reduction

StarTrail-org/LEANN12.6k1d

Implements graph-based vector indexing with selective recomputation for compact on-device RAG.

1014

441

Meta's FAIR introduces Autodata, an agentic framework that converts inference compute into high-quality synthetic training dataThe Agentic Self-Instruct implementation coordinates four specialized subagents.

4h8k183175

#147

Top post: @iScienceLuvr “abs: https://arxiv.org/abs/2606.25996”

ByteDance and Renmin University release iLLaDA, an 8B masked diffusion language model that outperforms Qwen2.5 7BIt was trained from scratch on 12 trillion tokens.

3h6.4k6928

#501

Top post: @teortaxesTex “So far ByteDance is supreme in videogen, but open weights LLMs from startups can humble Doubao Seed. Would be funny if they make a great leap forward when they finally figure out how to build fully diffusion-based LLMs too.”

Anthropic data reveals Chinese AI firms distilled its models, with Zhipu showing characteristics of unreleased Claude 4.5Alibaba led the activity with 28.8 million detected exchanges

1h7.7k8220

#501

Top post: @teortaxesTex “"punches above its weight" here is a vague metric that seeks to capture performance divided by (available human resources, capital, compute and focus on products). Minimax is the least impressive in this bunch, but Alibaba has vastly more resources. If they need so much Claude…”

SoftBank's bizarre shareholder slide depicting Masayoshi Son as an undervalued goose becomes a viral tech memeSon used the deck to pledge ten more years.

3h1.8k141

#594

Top comment: @edzitron “these two slides run one after another.”

100

Cambridge's Seán Ó hÉigeartaigh warns the US-led Pax Silica AI agreement creates a one-sided dependency for alliesUS AI access can be cut instantly, unlike physical hardware

40m1.3k43

#501

Top post: @teortaxesTex “> I think it's exactly the right direction to take for allies for now: America builds models, they integrate with allied supply chain positions… And I think "integration" is a cope. The issue for Europeans is that the integration is one-sided in terms of what it makes viable. US AI dependencies (such as ASML), while objectively vital, have high latency measured in months-years, but Fable can be switched off with one button push. It's the usual advantage of conventional arms over strategic ones. That's not really better for the EU than the current Chinese position where they can deny the West some REEs or tungsten or diamond saws for wafers; except we know that Xi also has diamond hands, but the EU will fracture and cry uncle once some Helberg calls their bluff. What "Pax Silica" type alliance ought to entail is removing this latency advantage; something substantive like supply backed by physics. As in, actual Fable weights in EU-controlled datacenters; and forward commitments to deploy American frontier models there within X days of their certification for American customers; and liabilities for cancellation. That's where the discussion has to *start*.”

110

Pleias co-founder Pierre-Carl Langlais argues the EU's promotion of cloned AI data opt-out tools favors specific interestsThe tools replicate Spawning's original ai.txt protocol.

4h3.5k451

#1537

Top comment: @Dorialexander “Unfortunately our core issue is much worse than over-regulation: generalized grift. Policy people are daily joining mid companies they helped propel/entrench and it’s totally normalized.”

120

New donation-backed initiative launches to eliminate respiratory diseases by funding Phase 2 preventatives and far-UVC air-cleaning systemsPrime Intellect's Vincent Weisser is among the project's supporters

4h387101

#990

Top post: @EgeErdil2 “@_sholtodouglas hopefully one day people will look back on the flu the way we today look back on measles”

130

Creator @scaling01 says Anthropic publishes distillation metrics to push for AI regulation because tracking distilled tokens is technically impracticalProving token-level distillation remains a major technical challenge.

40m40910

#501

Top post: @scaling01 “@teortaxesTex Anthropic just want to push for restrictions, the actual numbers who distills how much doesn't matter to them it's just very hard to prove which tokens are used for "distillation"”

Recent Stars

Github Stars

(7 days)

Today's Highlights

Former Anthropic and DeepMind lead Behnam Neyshabur launches Mirendil with $200 million to automate frontier AI engineering

Top Stories

Hugging Face CEO Clément Delangue says the open AI platform crossed $100 million in annual run-rateThe platform remains free for 97% of its users.

Anthropic's Sam McAllister says reports of unreleased Claude Fable 5 model leaking are due to a UI bugMcAllister confirmed that zero traffic is serving the model.

LisanBench creator Lisan al Gaib teases GPT-5.6-Preview with a screenshot of an apparent internal admin routeThe JSON snippet details route paths and CDN assets.

Video Signals

Extropic AI founder Guillaume Verdon backs Elon Musk's call to rebuild X's recommendation algorithm, while t3.gg creator Theo defends itTheo argues the system adaptively tracks real-time tech developments

AI technologist @deepfates requests community recommendations for digital audio workstations, drawing an endorsement for AbletonCreator @viemccoy recommended the software as a comprehensive solution

Meta's FAIR introduces Autodata, an agentic framework that converts inference compute into high-quality synthetic training dataThe Agentic Self-Instruct implementation coordinates four specialized subagents.

ByteDance and Renmin University release iLLaDA, an 8B masked diffusion language model that outperforms Qwen2.5 7BIt was trained from scratch on 12 trillion tokens.

Anthropic data reveals Chinese AI firms distilled its models, with Zhipu showing characteristics of unreleased Claude 4.5Alibaba led the activity with 28.8 million detected exchanges

SoftBank's bizarre shareholder slide depicting Masayoshi Son as an undervalued goose becomes a viral tech memeSon used the deck to pledge ten more years.

Cambridge's Seán Ó hÉigeartaigh warns the US-led Pax Silica AI agreement creates a one-sided dependency for alliesUS AI access can be cut instantly, unlike physical hardware

Pleias co-founder Pierre-Carl Langlais argues the EU's promotion of cloned AI data opt-out tools favors specific interestsThe tools replicate Spawning's original ai.txt protocol.

New donation-backed initiative launches to eliminate respiratory diseases by funding Phase 2 preventatives and far-UVC air-cleaning systemsPrime Intellect's Vincent Weisser is among the project's supporters

Creator @scaling01 says Anthropic publishes distillation metrics to push for AI regulation because tracking distilled tokens is technically impracticalProving token-level distillation remains a major technical challenge.

Recent Stars

Github Stars

Today's Highlights

Former Anthropic and DeepMind lead Behnam Neyshabur launches Mirendil with $200 million to automate frontier AI engineering

Top Stories

Hugging Face CEO Clément Delangue says the open AI platform crossed $100 million in annual run-rateThe platform remains free for 97% of its users.

Anthropic's Sam McAllister says reports of unreleased Claude Fable 5 model leaking are due to a UI bugMcAllister confirmed that zero traffic is serving the model.

LisanBench creator Lisan al Gaib teases GPT-5.6-Preview with a screenshot of an apparent internal admin routeThe JSON snippet details route paths and CDN assets.

Video Signals

Extropic AI founder Guillaume Verdon backs Elon Musk's call to rebuild X's recommendation algorithm, while t3.gg creator Theo defends itTheo argues the system adaptively tracks real-time tech developments

AI technologist @deepfates requests community recommendations for digital audio workstations, drawing an endorsement for AbletonCreator @viemccoy recommended the software as a comprehensive solution

Meta's FAIR introduces Autodata, an agentic framework that converts inference compute into high-quality synthetic training dataThe Agentic Self-Instruct implementation coordinates four specialized subagents.

ByteDance and Renmin University release iLLaDA, an 8B masked diffusion language model that outperforms Qwen2.5 7BIt was trained from scratch on 12 trillion tokens.

Anthropic data reveals Chinese AI firms distilled its models, with Zhipu showing characteristics of unreleased Claude 4.5Alibaba led the activity with 28.8 million detected exchanges

SoftBank's bizarre shareholder slide depicting Masayoshi Son as an undervalued goose becomes a viral tech memeSon used the deck to pledge ten more years.

Cambridge's Seán Ó hÉigeartaigh warns the US-led Pax Silica AI agreement creates a one-sided dependency for alliesUS AI access can be cut instantly, unlike physical hardware

Pleias co-founder Pierre-Carl Langlais argues the EU's promotion of cloned AI data opt-out tools favors specific interestsThe tools replicate Spawning's original ai.txt protocol.

New donation-backed initiative launches to eliminate respiratory diseases by funding Phase 2 preventatives and far-UVC air-cleaning systemsPrime Intellect's Vincent Weisser is among the project's supporters

Creator @scaling01 says Anthropic publishes distillation metrics to push for AI regulation because tracking distilled tokens is technically impracticalProving token-level distillation remains a major technical challenge.

Recent Stars

Github Stars

Yesterday's Top Stories, Jun 24, 2026.

Former Anthropic and DeepMind lead Behnam Neyshabur launches Mirendil with $200 million to automate frontier AI engineeringThe startup aims to automate complex scientific pipelines like AlphaFold

OpenAI announces Jalapeño, its first custom AI chip co-developed with Broadcom in nine monthsThe design was accelerated using OpenAI's own models.

Intercept launches with $500 million from Stripe, OpenAI, and Bill Gates to eliminate respiratory virusesIt will fund broad-spectrum preventatives and air-cleaning tech.

Claude AI advises Europeans to install air conditioning, sparking debate over climate adaptation and economic productivityElon Musk cited Singapore's AC usage as a productivity driver.

Z.ai's open-source GLM-5.2 scores 22.8% on ARC-AGI-2, trailing GPT-5.5 but running 7.5 times cheaperIt is the top-performing Chinese model on ARC-AGI-2.

Claude Code CLI updates suggest Anthropic is preparing to bundle Fable 5 into standard subscriptionsAnthropic has not yet officially announced the return.

Trump administration shifts Claude Fable 5 talks to Anthropic co-founder Tom Brown, bypassing CEO Dario AmodeiOfficials reportedly found Amodei difficult during policy discussions.

Arthur Conmy, formerly of Google DeepMind, joins Anthropic to align frontier models during trainingHe argues current models cannot safely automate AGI development

Qualcomm agrees to acquire Modular, the AI software company co-founded by Chris Lattner, to unify edge-to-cloud computeThe deal targets generative and agentic AI solutions.

Figma releases Weave Generator, Figma Motion, and an integrated AI agent to build plugins and shaders directly in its canvasThe updates consolidate creative, generative, and code-based canvas workflows

AI PACs spent $20 million in a New York congressional primary, raising questions about the efficacy of industry political spendingLeading the Future spent $8 million opposing candidate Bores

Two recent xAI employees say more than half of Grok usage is driven by adult and NSFW contentThe trend mirrors parent platform X's high NSFW traffic

Andrej Karpathy proposes shifting LLMs from personal chatbots to persistent, asynchronous organizational teammatesSkeptics dismissed the concept as marketing for existing workflows.

John Carmack argues id Software should have upgraded the Doom engine instead of rewriting technology during Quake's developmentHe also admitted pushing the development team to burnout

Key Gemini model contributors Jonas Adler and Alexander Pritzel leave Google to join AnthropicThe departures continue a trend of staff leaving Google AI.

Anthropic launches Claude Tag, embedding its AI assistant into Slack channels to act as an active team memberThe integration lets users delegate tasks like writing code.

Andrej Karpathy says X's reinforcement learning algorithms drive unprecedented toxicity by rewarding negative behaviorConsequently, Karpathy has reduced his activity on the platform

Anthropic accuses Alibaba of using 25,000 fake accounts to scrape Claude to train its Qwen modelsThe scraping campaign generated 28.8 million Claude interactions.

A game engine lineage chart maps how id Software's original Quake codebases spawned Valve's Source and the Call of Duty franchiseThe genealogy tracks forks powering Half-Life, Apex Legends, and Warzone.

OpenThoughts releases OpenThinkerAgent-32B, an open terminal and coding agent model post-trained from Qwen-3-32BIt achieves a 26% score on Terminal Bench.

OpenAI's roon argues that accepting the likelihood of near-term AGI is too psychologically difficult for most intelligent observersObservers increasingly rely on coping mechanisms instead of acceptance.

Political scientist Francis Fukuyama proposes an international treaty to establish a global moratorium on superintelligent AIAnders Sandberg and Nate Soares endorsed the proposed global ban

OpenAI study finds reinforcement learning targeting beneficial traits produces robust, emergent alignmentThe alignment gains generalize and resist adversarial pressure.

Google integrates native computer use into Gemini 3.5 Flash with built-in safeguards against prompt injectionThe model scored 78.4 on the OSWorld-Verified benchmark

Nathan Lambert, AI2 post-training lead, releases an ML foundations lecture to kick off his 11-part post-training courseThe video lecture covers KL divergence, cross-entropy, and MDPs.

Zvi Mowshowitz says Polymarket odds for Claude Fable 5 returning to US users by August are overconfident due to KYC hurdlesThe Polymarket contract reached $1.9 million in trading volume.

Runlayer raises $30 million to secure AI agents by replacing standard employee logins with dedicated, scoped identitiesFelicis and Khosla Ventures led the funding round.

Blueprint founder Bryan Johnson rebrands his longevity company to Immortals to focus on biological immortalityJohnson previously founded payment platform Braintree and neurotech startup Kernel.

Notion launches External Agents to integrate Claude and Cursor directly into shared team workspacesThe integrated agents can triage bugs and generate pull requests.

Keen Technologies founder John Carmack reflects on how Jerry Pournelle's pessimistic 1970s sci-fi contrasted with the subsequent tech boomSteven Sinofsky highlighted Pournelle's influential Byte magazine column.