Tech News

Today's Highlights

Noam Shazeer leaves Google to join OpenAI, two years after Google spent $2.7 billion to bring him back

+45 more

It’s a new day at Digg. Fresh stories are clustering now. 70% complete.

Video Signals

#1848

Builder Releases Interactive 3D Midjourney CT Scanner Model

5h ago36K656241

JD Vance Labels AI A Communist Technology Over Surveillance Fears

3h ago48K534125

#1360

Midjourney Scanner Delivers 60-Second Full-Body Imaging Cheaper Than MRI

6h ago321K31458

NVIDIA Unveils Isaac GR00T Open Humanoid Robot Reference Design

6h ago1.2K183

AI-Generated Messi Free Kick Video Prompts Authenticity Questions

45m ago5.5M173

#1419

X Algorithm Floods User For You Feed With Consecutive Replies

12m ago4310

Humanoid Robot Performs Front and Back Flips on Outdoor Field

3h ago35640

Grok Imagine API Generates Cinematic Videos From Text Prompts

5h ago24331

Uber Launches Autonomous Solutions to Scale Global Robotaxi Services

1h ago19620

Amorepacific And MIT Launch AI Skinsight Patch For Real-Time Skin Monitoring

3h ago19511

General Reasoning's Ross Taylor says updated KellyBench shows leading open-source models remain six months behind closed frontier modelsTaylor says benchmark optimization pressure hides the true performance gap

3h9.3k659

#501

Top comment: @teortaxesTex “@ThainRos I don't measure against only public models”

Midjourney launches Midjourney Medical to deploy 50,000 rapid ultrasound scanners with Butterfly HealthThe proprietary hardware completes full-body scans in 60 seconds

47m18.7k10045

#593

Top comment: @godg641 “@pitdesi 🎉🎉😁🔥”

New github stars (48hrs)

Terminal AI coding assistant that inspects codebases edits files and runs multi-step workflows

Noumena-Network/code691d

Provides terminal AI coding assistant for codebase inspection, file edits, command execution and multi-step workflows.

573

1136

Hardware-isolated sandbox with allowlist firewall and monitor for safe Claude Code sessions

alexander-turner/claude-guard291d

Wraps Claude Code sessions with deterministic checks on commands and writes to limit agent access.

1278

570

Indian Prime Minister Narendra Modi meets Mistral AI founder Arthur Mensch to discuss potential AI partnershipsThe meeting sparked developer speculation of a Sarvam AI merger.

58m2.2k172

#256

Top comment: @filibluster “@narendramodi India has 500 million tech support scammers and doesn't have a single frontier AI lab. That's kind of embarrassing for a country of your size. You should stop fake university scams and from letting your best and brightest leave if you want your country to prosper.”

Sutter Hill's Kevin Kwok and fofr question the skull-penetration capabilities of David Holz's open-scanner medical hardwareThe design aims to keep patients outside traditional scanning tubes.

5h775161

#117

Top comment: @DavidSHolz “@fofrAI @ruima hold ur breath ig”

Midjourney releases a fictional concept and CGI video for a hypothetical radiation-free "Ultrasonic CT" scannerThe fictional plan includes a flagship spa in San Francisco.

1h000

#617

Top comment: @IterIntellectus “and ultrasound can also be used to precisely target tissue which means that the line between imaging and treatment is starting to get very interesting scan, map the body, find any issue, stimulate, ablate, modulate, or repair with sound”

A satirical Effective Altruism memo addressed to Samuel Hammond mocks shifting AI safety arguments for Donald TrumpIt followed rumors of a White House frontier AI pause.

6h613110

#1315

101

Simo Ryu argues raw chain-of-thought traces are unreadable and only look legible due to UI summarization layersRaw internal traces contain excessively long and bizarre terminology

2h1.1k160

#501

Top comment: @cloneofsimo “@teortaxesTex I dont mean the summarized CoTs, i mean the raw ones. Raw ones are so long and full of weird words, unlike the summarized ones we see on interfaces”

111

Developer @xlr8harder jokes about tool-use inefficiency after an AI assistant searches the web to find the current timeFlorian Brand joked the model mimicked a European employee

4h1.3k320

#1778

Top comment: @xeophon “@xlr8harder using an european model and it wants to clock out”

126

OpenAI develops a safety simulation pipeline that replays real ChatGPT conversations to predict model failures with 92% accuracyIts failure estimates are 1.5 times closer to actual outcomes.

4h000

#1249

Top comment: @MicahCarroll “@Marcus_J_W @hannahsheahan @CJKRaymond @tomekkorbak The idea is simple: we take de-identified conversations from a previous deployment, keep the conversation so far fixed, and ask the unreleased model to produce the next assistant turn.”

131

NCODE builder @_xjdr is investigating an email signup bug reported by open-source developer @xlr8harderThe signup attempt returned a generic invalid signup error.

17m26850

#830

Top comment: @xlr8harder “@_xjdr I don't get a useful error message to share, unfortunately, just: >invalid signup request”

141

Prime Intellect's Florian Brand jokes about whether General Reasoning is a hedge fund, drawing a shushing emoji from Ross TaylorThe exchange provided no concrete technical or organizational details.

2h16330

#842

Top comment: @rosstaylor90 “@xeophon 🤫”

Recent Stars

Github Stars

(7 days)

Today's Highlights

Noam Shazeer leaves Google to join OpenAI, two years after Google spent $2.7 billion to bring him back

Top Stories

Open-source model GLM 5.2 scores 67.3% on WeirdML v2 benchmark, placing it between Claude Opus 4.5 and Gemini 3 ProWeirdML v2 tracks API costs across 19 evaluation tasks

Anthropic's Chris Ciauri says blocked frontier models Mythos and Fable 5 will regain availability in South Korea within daysAccess restoration follows positive G7 negotiations with President Trump

Elon Musk predicts a Chinese 'Fable-class' AI by Q1 2027, prompting Tsinghua University's Jie Tang to say it will arrive soonerChinese AI organization Z.ai accepted the Q1 2027 challenge.

Video Signals

General Reasoning's Ross Taylor says updated KellyBench shows leading open-source models remain six months behind closed frontier modelsTaylor says benchmark optimization pressure hides the true performance gap

Midjourney launches Midjourney Medical to deploy 50,000 rapid ultrasound scanners with Butterfly HealthThe proprietary hardware completes full-body scans in 60 seconds

Indian Prime Minister Narendra Modi meets Mistral AI founder Arthur Mensch to discuss potential AI partnershipsThe meeting sparked developer speculation of a Sarvam AI merger.

Sutter Hill's Kevin Kwok and fofr question the skull-penetration capabilities of David Holz's open-scanner medical hardwareThe design aims to keep patients outside traditional scanning tubes.

Midjourney releases a fictional concept and CGI video for a hypothetical radiation-free "Ultrasonic CT" scannerThe fictional plan includes a flagship spa in San Francisco.

A satirical Effective Altruism memo addressed to Samuel Hammond mocks shifting AI safety arguments for Donald TrumpIt followed rumors of a White House frontier AI pause.

Simo Ryu argues raw chain-of-thought traces are unreadable and only look legible due to UI summarization layersRaw internal traces contain excessively long and bizarre terminology

Developer @xlr8harder jokes about tool-use inefficiency after an AI assistant searches the web to find the current timeFlorian Brand joked the model mimicked a European employee

OpenAI develops a safety simulation pipeline that replays real ChatGPT conversations to predict model failures with 92% accuracyIts failure estimates are 1.5 times closer to actual outcomes.

NCODE builder @_xjdr is investigating an email signup bug reported by open-source developer @xlr8harderThe signup attempt returned a generic invalid signup error.

Prime Intellect's Florian Brand jokes about whether General Reasoning is a hedge fund, drawing a shushing emoji from Ross TaylorThe exchange provided no concrete technical or organizational details.

Recent Stars

Github Stars

Today's Highlights

Noam Shazeer leaves Google to join OpenAI, two years after Google spent $2.7 billion to bring him back

Top Stories

Open-source model GLM 5.2 scores 67.3% on WeirdML v2 benchmark, placing it between Claude Opus 4.5 and Gemini 3 ProWeirdML v2 tracks API costs across 19 evaluation tasks

Anthropic's Chris Ciauri says blocked frontier models Mythos and Fable 5 will regain availability in South Korea within daysAccess restoration follows positive G7 negotiations with President Trump

Elon Musk predicts a Chinese 'Fable-class' AI by Q1 2027, prompting Tsinghua University's Jie Tang to say it will arrive soonerChinese AI organization Z.ai accepted the Q1 2027 challenge.

Video Signals

General Reasoning's Ross Taylor says updated KellyBench shows leading open-source models remain six months behind closed frontier modelsTaylor says benchmark optimization pressure hides the true performance gap

Midjourney launches Midjourney Medical to deploy 50,000 rapid ultrasound scanners with Butterfly HealthThe proprietary hardware completes full-body scans in 60 seconds

Indian Prime Minister Narendra Modi meets Mistral AI founder Arthur Mensch to discuss potential AI partnershipsThe meeting sparked developer speculation of a Sarvam AI merger.

Sutter Hill's Kevin Kwok and fofr question the skull-penetration capabilities of David Holz's open-scanner medical hardwareThe design aims to keep patients outside traditional scanning tubes.

Midjourney releases a fictional concept and CGI video for a hypothetical radiation-free "Ultrasonic CT" scannerThe fictional plan includes a flagship spa in San Francisco.

A satirical Effective Altruism memo addressed to Samuel Hammond mocks shifting AI safety arguments for Donald TrumpIt followed rumors of a White House frontier AI pause.

Simo Ryu argues raw chain-of-thought traces are unreadable and only look legible due to UI summarization layersRaw internal traces contain excessively long and bizarre terminology

Developer @xlr8harder jokes about tool-use inefficiency after an AI assistant searches the web to find the current timeFlorian Brand joked the model mimicked a European employee

OpenAI develops a safety simulation pipeline that replays real ChatGPT conversations to predict model failures with 92% accuracyIts failure estimates are 1.5 times closer to actual outcomes.

NCODE builder @_xjdr is investigating an email signup bug reported by open-source developer @xlr8harderThe signup attempt returned a generic invalid signup error.

Prime Intellect's Florian Brand jokes about whether General Reasoning is a hedge fund, drawing a shushing emoji from Ross TaylorThe exchange provided no concrete technical or organizational details.

Recent Stars

Github Stars

Yesterday's Top Stories, Jun 17, 2026.

Noam Shazeer leaves Google to join OpenAI, two years after Google spent $2.7 billion to bring him backThe departure directly impacts Google's Gemini initiative.

Midjourney founder David Holz launches Midjourney Medical to build an ultrasonic CT scanner for 60-second full-body scansThe startup aims to deploy 50,000 scanners over six years

AI CEOs Sam Altman, Dario Amodei, Demis Hassabis, and Arthur Mensch met privately and with President Trump at the G7 summitThe two-hour session covered regulatory challenges and export restrictions.

Z.ai launches GLM-5.2, an open-weights LLM with a 1-million-token context window that beats GPT-5.5 on FrontierSWEChris Lattner's Modular Cloud provides day-zero serving.

US orders Anthropic to suspend Fable 5 and Mythos 5 models globally over foreign national access restrictionsThe government gave Anthropic a 90-minute compliance deadline.

Capital Factory co-founder and CEO Joshua Baer dies in a business jet crash near Laredo, TexasTech and investment leaders shared tributes honoring his legacy.

Anthropic's Dario Amodei and DeepMind's Demis Hassabis propose a US-led coalition to block China from frontier AI and chipsThe pitch was made during a closed-door G7 meeting.

OpenAI Codex engineering lead Thibault Sottiaux says Codex tools can now be configured to run any open-source modelUsers can route client traffic to self-hosted local endpoints.

Elon Musk predicts complete AI-generated movies by the end of the year following a brief xAI video model teaserThe prediction followed a trailer showcasing xAI's video model.

Cal Newport argues AI companies must either halt development or accept full product liability instead of warning about existential doomHugging Face founder Clement Delangue supported ending "doom marketing.

Former OpenAI Sora developer Gabriel Petersson leaves to found a startup building a final product prior to AGIThe startup is flying candidates to Sweden for interviews

Former Webflow CTO Bryant Chou launches Ploy with $27 million in seed funding to automate website operationsEarly adopters of the platform include Hex and Clay

OpenAI joins the Rust Foundation as a Platinum Member, committing $600,000 to support open-source maintainersOpenAI is adopting Rust for its systems programming infrastructure.

David Sacks, former White House AI czar, warns defensive teams to secure codebases before Anthropic Mythos-level capabilities diffuse to ChinaHe rejected media reports claiming he dismissed these cyber threats

Snap CEO Evan Spiegel debuts bulky new Specs AR glasses, triggering design criticism and a stock declineCritics compared the bulky frames to IMAX 3D glasses.

Leaked Anthropic Slack messages reveal employee panic over Trump administration restrictions and a disputed $200 million defense contractOver-aggressive cyber protections in Fable 5 drew industry mockery

Martin Shkreli claims WSJ reporters hold liberal biases while its favorable editorial profile of Elon Musk is accurateMusk joked the paper might not be so unfavorable

Trump administration demands Anthropic guarantee Fable 5 cannot be jailbroken, a mandate experts call technically impossibleThe mandate effectively requires solving the AI alignment problem.

Block deploys Builderbot, an internal AI agent system that makes 15% of its production code changesThe system merges 1,500 pull requests weekly via Slack.

Illinois Governor J.B. Pritzker signs a 0.2% cryptocurrency transaction tax that applies to personal wallet transfersThe Crypto Council calls it the most punitive US tax.

xAI releases Grok 4.3 on Amazon Bedrock to bring real-time X data to AWS developersAzure, Google Vertex, and Oracle also host the model.

Odyssey ML raises $310 million Series B to develop physical-world simulation modelsInvestors in the round include Amazon, GV, and AMD.

Robotics startup XDOF exits stealth with $70 million and releases ABC-130K, the largest open-source bimanual robot manipulation datasetThe dataset was gathered using the YAM hardware platform.

PPO creator John Schulman reveals the foundational Proximal Policy Optimization paper was rejected from NIPS 2017He shared the anecdote while posting new RLVR research.

Will Manidis, ScienceIO founder, argues SpaceX could leverage its cheap capital to acquire Stripe and FigmaLevelsio suggested Stripe could power X as an everything app.

Mark Zuckerberg's memo urging Meta employees to "have fun" at an AI hackathon draws pushback following 8,000 layoffsThe internal memo aimed to revive Meta's "hacker culture.

Next.js creator Guillermo Rauch's Vercel launches Eve, a file-based AI agent framework with a unified Agent StackIt uses Markdown for instructions and TypeScript for tools.

OpenAI releases LifeSciBench, a benchmark of 750 expert tasks to evaluate AI in life science researchThe benchmark covers seven distinct biological research workflows

AI commentator @teortaxesTex says GLM 5.2 has a seven-month developmental lag, predicting Chinese frontier parity by late 2026Elon Musk expects parity to take until early 2027.

Lenny Rachitsky, creator of Lenny's Podcast, teases upcoming guest lineup featuring AI leaders from OpenAI, Anthropic, and Google DeepMindGuests include Jeff Dean and Netflix's Elizabeth Stone.