Tech News

11

Anthropic will restore access to Claude Fable 5 and Mythos 5 after US lifts export controlsService restoration begins July 2 following policy negotiations

4h205.9k3.5k225

Top comment: @AndrewCurran_ “Fable has been freed. For everyone.”

21

Etched raises $800 million to ship its custom AI inference hardware starting this summerThe company has secured over $1 billion in contracts

12h60.7k37675

Top comment: @karpathy “Congrats!! I was impressed to learn about some of the engineering wizardry (e.g. *very* low voltage domains, cluster scale memory, ...) that goes into tokens/watt maxxing of state of the art LLMs at interactive tokens/sec/user. Esp fun and memorable is the idea that this is engineering at the "opposite" regime to that of power transmission lines: very low voltage high current (at tiny distances) vs. very high voltage & low current (at great distances). Looking forward to more!”

30

Watchlist

Video Signals

#52

xAI Shares Video Promoting Universe Exploration And AI Careers

12h ago527K1.3K96

#571

Etched Emerges From Stealth With $800M Raised And $1B In AI Chip Contracts

12h ago2.7K2418

#1260

Aravind Srinivas Warns Export Controls Accelerate China's AI Infrastructure Edge

19h ago63K20268

#75

Google DeepMind Opens New London Office Platform 37 Honoring AlphaGo

18h ago79K20918

#66

ASPIRE Gives Robots Self-Evolving Skills Library That Compounds Indefinitely

10h ago4.9K5423

#1329

Theo Releases Video Explaining OpenAI Models' Token Efficiency Edge

4h ago8.2K18136

#60

Podcast Explores AI's Rapid Progress In Mathematics And Conceptual Breakthroughs

10h ago4.9K6342

#1546

Anthropic CEO Warns Open-Source AI Follows Very Dangerous Path

12h ago27K00

#713

Figure F.03 Humanoid Robot Begins Operations at BMW Factory

12h ago3.1K608

#905

US Lifts Export Controls On Anthropic Claude Fable 5 And Mythos 5

1h ago135K01

40

55

Neuralink completes first human through-dura electrode insertion, bypassing the need to cut the brain's protective membraneThe milestone procedure was performed at Toronto's University Health Network.

7h#1VIEWS1.3M#1LIKED4.3k176#1COMMENTS924

#52

Top post: @elonmusk “@shivon @cb_doge Absolutely”

60

72

83

Rihard Jarc argues token optimization structurally benefits Amazon, Microsoft, and Google over dedicated AI labsBill Ackman cited the analysis to back his thesis.

16h898k3.4k#1BOOKMARKED2.5k

#109

Top comment: @Pregory1 “@RihardJarc “…might *not* be…, but it does…n’t mean…?” Is that typo? Should it be “but it *does* mean”? Great write up, by the way. Keep up the good work. 🤝”

91

102

115

125

AI commentator Andrew Curran teases an upcoming memory-efficiency architecture breakthrough from an unnamed OpenAI spinoutThe developer team is reportedly not Safe Superintelligence.

3h56.6k1.2k327

#501

Top post: @bayeslord “jerrrrrry what did you do?”

131

Spotted backend preparations spark speculation that a Claude Sonnet 5 release is imminentThe launch may coincide with a rumored Fable 5.

18h110.2k92969

#741

Top post: @kimmonismus “Rumors about Sonnet 5 have been circulating for a long time. Now it looks like Sonnet 5 will be released alongside the re-release of Fable 5 (presumably with ID verification). This leaves a bad taste in my mouth, because it raises the possibility that those who might not be able to access Fable 5 (in Europe?) will essentially get a much weaker alternative in Sonnet 5. I might be wrong, I even hope so. But I have a bad feeling about it.”

145

158

LisanBench creator @scaling01 says Claude Sonnet 5 is twice as expensive as GPT-5.5 and 57 times pricier than DeepSeek-V4-ProSonnet 5 costs $2.29 per task versus DeepSeek's $0.04.

5h41.2k72235

FASTEST CLIMBING+8

#382

Top post: @steipete “Price per token != cost per task”

163

177

187

Nikita Bier, Head of Product at X, jokes that his first year hit three all-time highs: user engagement, cortisol, and biological ageOther tech creators reacted by praising his high-stress milestone.

11h20.7k58315

RISING LIKES583

#239

Top post: @bryan_johnson “@nikitabier Nikita I’ve honestly been thinking about your well being. I hope you’re doing ok. I appreciate all that you do for the community.”

191

DeepSeek releases DSpark, a speculative decoding algorithm boosting DeepSeek-V4 throughput by up to 661%Open-source vLLM has rapidly integrated support for the algorithm.

18h67.9k39060

#403

Top post: @zephyr_z9 “DSpark will see crazy adoption The Whale reduced the compute pressure on Chinese AI infra”

Digg Deeper question: “What's speculative decoding? And why does this news matter?”

answerview answer

202

Anthropic launches the Claude Desktop app in beta for Ubuntu and Debian LinuxThe beta supports Claude Code for all paid-tier users.

11h17.9k27018

#369

Top post: @bcherny “You asked, we listened. Claude Desktop on Linux is here! Download link: https://code.claude.com/docs/en/desktop-linux”

Today's Highlights

Anysphere launches Cursor for iOS, letting developers run always-on cloud agents and manage desktop agents remotely

Neuralink completes first human through-dura electrode insertion, bypassing the need to cut the brain's protective membrane

Rihard Jarc argues token optimization structurally benefits Amazon, Microsoft, and Google over dedicated AI labs

LisanBench creator @scaling01 says Claude Sonnet 5 is twice as expensive as GPT-5.5 and 57 times pricier than DeepSeek-V4-Pro

Top Stories

Anthropic will restore access to Claude Fable 5 and Mythos 5 after US lifts export controlsService restoration begins July 2 following policy negotiations

Etched raises $800 million to ship its custom AI inference hardware starting this summerThe company has secured over $1 billion in contracts

Anthropic releases Claude Sonnet 5, scoring 63.2% on SWE-bench Pro and 81.2% on OSWorld computer useThe model features an optional 1-million-token context window.

Video Signals

Ramp and Revelio Labs study of 21,559 firms finds high-intensity AI adopters grew headcount by 10% over two yearsEntry-level headcount at those high-adopting firms rose 12%.

Neuralink completes first human through-dura electrode insertion, bypassing the need to cut the brain's protective membraneThe milestone procedure was performed at Toronto's University Health Network.

Google DeepMind launches Nano Banana 2 Lite for $0.034 images and previews Gemini Omni Flash APIOmni Flash generates 10-second videos at $0.10 per second.

Visa, Stripe, and Coinbase back Open Standard's upcoming OpenUSD stablecoin designed to share economics with usersTempo will natively issue the stablecoin on day one.

Rihard Jarc argues token optimization structurally benefits Amazon, Microsoft, and Google over dedicated AI labsBill Ackman cited the analysis to back his thesis.

Biotech startup Conception successfully generates the first early-stage human eggs derived from stem cellsThe milestone advances in vitro gametogenesis to address infertility.

OpenAI internal optimization cuts inference costs by half, running logged-out ChatGPT traffic on a couple hundred GPUsObservers speculate the undisclosed method relies on speculative decoding

Sebastian Raschka releases a new book on building reasoning LLMs from scratchThe guide covers inference scaling, reinforcement learning, and distillation.

AI commentator Andrew Curran teases an upcoming memory-efficiency architecture breakthrough from an unnamed OpenAI spinoutThe developer team is reportedly not Safe Superintelligence.

Spotted backend preparations spark speculation that a Claude Sonnet 5 release is imminentThe launch may coincide with a rumored Fable 5.

Anthropic's Claude Code is accused of quietly fingerprinting China-based users through hidden markers in outbound system promptsIt reportedly detects local timezone settings when proxies are enabled

LisanBench creator @scaling01 says Claude Sonnet 5 is twice as expensive as GPT-5.5 and 57 times pricier than DeepSeek-V4-ProSonnet 5 costs $2.29 per task versus DeepSeek's $0.04.

Smol.ai founder Shawn 'swyx' Wang kicks off the AI Engineer World's Fair, celebrating three years of the developer summitThe developer summit features a keynote by creator Theo.

AI developer @teortaxesTex says Claude Sonnet 5.0 underperforms Sonnet 4.6 on non-target benchmarks, sparking 'benchnerfing' debatesAnother critic questioned if the model even outperforms GLM-5.2.

Nikita Bier, Head of Product at X, jokes that his first year hit three all-time highs: user engagement, cortisol, and biological ageOther tech creators reacted by praising his high-stress milestone.

DeepSeek releases DSpark, a speculative decoding algorithm boosting DeepSeek-V4 throughput by up to 661%Open-source vLLM has rapidly integrated support for the algorithm.

Anthropic launches the Claude Desktop app in beta for Ubuntu and Debian LinuxThe beta supports Claude Code for all paid-tier users.

Recent Stars

Github Stars

Yesterday's Top Stories, Jun 29, 2026.

Anysphere launches Cursor for iOS, letting developers run always-on cloud agents and manage desktop agents remotelyThe mobile app supports voice dictation for background code generation

Meta AI open-sources Brain2Qwerty v2, a non-invasive brain-to-text decoder achieving 61% word accuracyIt transitions from character decoding to full-sentence semantic decoding.

Meituan releases LongCat-2.0, a 1.6T-parameter open-weights MoE model trained entirely on 50,000 Chinese ASICsIt supports a native 1 million token context window.

Anthropic CEO Dario Amodei argues open-source AI cannot be monitored, revoked, or patched for safety after releaseThe clip reignited debate over open-source AI safety risks

Secretary of War Pete Hegseth appoints Marc Andreessen and Blake Masters to new Defense Policy BoardAmbassador Robert Lighthizer will chair the advisory board.

Former Meta PM Xiaoyin Qu predicts US and European enterprises will adopt Chinese AI models to self-host on private GPUsQu cites enterprise distrust of Anthropic's data retention policies

Meta restricts employees from using Claude Code and OpenAI Codex to prevent competitor model distillationThe policy protects the training pipeline of Meta's upcoming MetaCode.

Kim Isenberg and Peter Wang argue the US is losing the AI race due to energy and data center constraintsChina uses free models to promote domestic Huawei hardware.

Smol.ai founder Shawn 'swyx' Wang opens the AI Engineer World's Fair with a focus on local AI and NVIDIA Blackwell hardwareMajor sponsors include OpenAI, Anthropic, and Google DeepMind

Tesla releases FSD v14 Lite, bringing distilled AI4 driving models to memory-constrained AI3 hardwareOlder AI3 computers operate with 15% of AI4's memory bandwidth

Elon Musk compares machine optimization to solving an elegant puzzle that yields practical utilityResearch engineer shyamal posted in agreement with the sentiment

CLODO launches an AI-powered talent search platform that scans 1.2 billion profiles using natural languageEarly adopters include the U.S. Intelligence Community

2025 ARC Prize winner Alexia Jolicoeur-Martineau, known for her Tiny Recursive Model work, joins MicrosoftShe joins the company from Samsung's SAIT AI Lab.

Cognition AI releases Devin Fusion, using mid-session model routing to cut agentic coding costs by 35%A parallel sidekick agent handles localized operations like linting.

Foundation for American Innovation launches Frontier Legal Defense to fight regulatory overreach and copyright disputes in AIThe initiative is recruiting litigators fluent in AI scaling policies

Claude Code creator Boris Cherny says the next version will run subagents in the background by defaultThis lets users continue chatting while tasks run independently.

Startup 8090 raises $135 million in Series A funding led by Salesforce VenturesThe capital will fund hiring and infrastructure development

National Design Studio releases Rampart, a 14.7MB open-source model that redacts PII directly in the browserEdward Coristine and Tai Groot built the Hugging Face-hosted model.

Rocket Lab to acquire Iridium Communications in an $8 billion deal, combining launch capabilities with global satellite networkRocket Lab went public in 2021 at $4.1 billion.

OpenAI teases a custom physical shortcut keypad built with Work Louder for its Codex coding modelThe illuminated 3x3 keypad will be revealed July 15.

DeepSeek's upcoming V4 models will introduce peak-hour pricing that doubles API rates during peak Beijing hoursThe models support a 1M token context window.

Three labor economists debate whether AI will reverse the Industrial Revolution by making human labor abundantDavid Autor argues AI will create new human tasks.

Policy analyst Dean W. Ball sparks a debate over an AI labor essay that is actually David Ricardo's 1817 textAI safety researchers deemed the replacement premise highly plausible

OpenAI's Gabriel Petersson argues strategic pivots and intense, founder-type hires make AI companies structurally prone to internal conflictCreator roon countered by highlighting Anthropic's internal cohesion.

Jacob Andreas, MIT NLP research lead, co-develops method to replace transformer attention heads with human-readable Python codeReplacing 25% of heads maintained downstream question-answering accuracy.

Meta's Lucas Beyer asks if any serious open-source LLM pretraining repositories exist without relying on NVIDIA's Megatron-LMAndrew Carr proposed PyTorch's torchtitan as a potential alternative

TheStalwart newsletter argues geopolitical AI competition could force governments to use domestic austerity to fund computing powerCreator @teortaxesTex says resource diversion already drives inflation concerns

Groq founder Jonathan Ross argues startups should hire interns because LLMs let early-career talent bypass conventional technical limitationsMarkus Wulfmeier noted student-led projects drove top DeepMind research.

Arena, the AI evaluation platform co-founded by Chatbot Arena's Wei-Lin Chiang, hits $100 million ARR eight months after launchIt tracks enterprise model performance on live user tasks.

Balsa Research's Zvi Mowshowitz disputes WSJ claims that Chinese model GLM-5.2 matches Anthropic's Mythos in cybersecurityEthan Mollick warns the narrative will skew policy debates