Tech News

Today's Highlights

OpenAI launches limited preview of GPT-5.6 models Sol, Terra, and Luna following a US government request

#13

+45 more

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

Video Signals

#1325

OpenAI Previews GPT-5.6 Sol Model with Advanced Benchmarks in Leaked Demo

4h ago30K57088

#861

Liquid AI's LFM2.5 Hits 280 Tokens Per Second on Mobile via WebGPU

4h ago17K00

#274

Sci-Fi Video Imagines 2027 World of AI Regulation and DMV Checks

4h ago3.4K103

New Architecture Adds Trust Layer to AI Agents for Safer Execution

3h ago1.5K80

Humanoid Robots Learn to Climb and Vault Obstacles Using AI

3h ago1.2K50

Eric Topol says GPT-5.5 Pro scored 79/100 on medical benchmark but remains below clinical deployment thresholdSeparately, Sophont released Medmarks v1.0 to automate medical evaluations.

3h25.4k12038

#613

Top post: @EricTopol “Free access link to our paper http://rdcu.be/fqznS”

Study finds compressing Chain-of-Thought traces during knowledge distillation degrades downstream student model accuracyEvaluations used teacher-student pairs like Llama-70B and Ministral-14B.

2h6.4k5921

#501

Top post: @teortaxesTex “CoT summarization/hiding is a straightforwardly efficient gatekeeping mechanism.”

New github stars (48hrs)

Full-stack Python codebase for training and evaluating draft models used in speculative decoding

deepseek-ai/DeepSpec551h

Provides full-stack code for training/evaluating speculative decoding draft models with data prep and eval scripts.

1194

1136

LLM post-training framework connecting Megatron training to SGLang rollout via a shared data buffer

THUDM/slime6.8k9h

Integrates Megatron training with SGLang rollouts for LLM post-training and RL scaling.

458

1014

484

OSWorld 2.0 benchmark and environment for long-horizon computer use agent evaluation

xlang-ai/OSWorld-V21618h

Benchmarks computer-use agents on long-horizon real-world desktop tasks via VM environments and gated task sets.

397

843

Archive of system prompts internal tools and models from AI coding agents like Cursor Claude Devin and Windsurf

x1xhlol/system-prompts-and-models-of-ai-tools141.2k1d

Aggregates leaked and open-sourced system prompts plus tool definitions from AI coding agents like Cursor, Claude, and Devin.

1014

827

Repository of modular skills folders and Agent Skills specification for extending Claude

anthropics/skills155.5k1d

Hosts collections of self-contained skills, specs, and templates for dynamically extending Claude's capabilities on specialized tasks.

1014

1467

AI infrastructure analyst Zephyr argues GPT-5.5 wastes significant tokens through repetitive loops during goal-directed tasksMemes parodied the model's repetitive demands for smoke tests.

3h47k2136

#501

Top post: @teortaxesTex “GPT 5.5 be like”

Bank of Korea finds generative AI fails to boost macroeconomic productivity despite 51.8% worker adoptionSaved time was mostly absorbed by administrative feedback loops.

4h6.1k4016

#1049

Top post: @rohanpaul_ai “https://www.bok.or.kr/portal/bbs/B0000347/view.do?nttId=10098529&searchCnd=1&searchKwd=&depth2=201106&depth=201106&pageUnit=10&pageIndex=1&programType=newsData&menuNo=201106&oldMenuNo=201106”

Will Brown says serving large Mixture of Experts models economically requires high batch sizes on 32 to 64 GPUsThe debate addressed hardware scaling for trillion-parameter reinforcement learning.

2h742359

#573

Top post: @Dorialexander “@willccbb I know (great post!). But wide-ep is requiring a large GPU stack?”

Academic François Fleuret argues that prioritizing logical reasoning preserves individual agency against tech's cult-like worldviewsDeepSeek booster Teortaxes dismissed Fleuret's critique as stereotypically French.

36m1.8k182

#501

Top post: @teortaxesTex “The French are very hostile to cults, because they worship Reason. This may have held them back in AGI, too, despite their monstrously accomplished mathematical school, abundant energy, and state aspiration to sovereignty and greatness. But ironically…”

101

Moondream AI co-founder Vikhyat K. argues major AI services fail to achieve three-nines availabilityClaude's API recorded 99.55% uptime, while DeepSeek hit 99.88%

1h4.3k420

#501

Top post: @teortaxesTex “In fairness, DeepSeek doesn't currently have 3 nines but it's so close… frustrating”

113

AI safety researcher Minh Nhat Nguyen argues U.S. observers overestimate the alignment between Chinese frontier AI labs and the CCPZhipu AI earns 40% gross margins from municipal contracts

2h47631

#501

Top post: @menhguin “imo, Americans greatly overestimate how aligned Chinese frontier labs are with the CCP lumping the Chinese labs as loyal minions of the Chinese government is like assuming Dario and Sam work for the Trump admin”

120

General Reasoning's Ross Taylor warns that rapidly shifting tooling imposes high engineering overhead on AI evaluation and training harnessesOpinionated harness designs risk fast technical obsolescence.

2h21890

#842

Top post: @xeophon “@rosstaylor90 🥲🥲🥲 tell me about it, currently working on just that”

130

Pleias CTO Pierre-Carl Langlais says hardware supply lags limit many company GPU clusters to 16 to 32 H100sThe bottleneck drives demand for specialized GPU sourcing organizations

2h24640

#1537

Top post: @xeophon “@Dorialexander @willccbb I wonder whether there’s an org which helps you sourcing larger clusters for your needs at great prices”

Recent Stars

Github Stars

(7 days)

Today's Highlights

OpenAI launches limited preview of GPT-5.6 models Sol, Terra, and Luna following a US government request

Top Stories

Prime Intellect's Elie Bakouch questions historical skepticism of speculative decoding following DeepSeek's DSpark deploymentDSpark uses parallel drafting and sequential biasing for speed

Founder Beff argues US restrictions on leading AI models backfired due to flawed comparisons to nuclear weaponsT3 Stack creator Theo Browne initiated the policy debate.

Systems architect Daniel Jeffries warns that sudden restrictions on frontier AI models could spark an open-source rebellionHigh-level model access was reportedly blocked for two weeks.

Video Signals

Eric Topol says GPT-5.5 Pro scored 79/100 on medical benchmark but remains below clinical deployment thresholdSeparately, Sophont released Medmarks v1.0 to automate medical evaluations.

Study finds compressing Chain-of-Thought traces during knowledge distillation degrades downstream student model accuracyEvaluations used teacher-student pairs like Llama-70B and Ministral-14B.

AI infrastructure analyst Zephyr argues GPT-5.5 wastes significant tokens through repetitive loops during goal-directed tasksMemes parodied the model's repetitive demands for smoke tests.

Bank of Korea finds generative AI fails to boost macroeconomic productivity despite 51.8% worker adoptionSaved time was mostly absorbed by administrative feedback loops.

Will Brown says serving large Mixture of Experts models economically requires high batch sizes on 32 to 64 GPUsThe debate addressed hardware scaling for trillion-parameter reinforcement learning.

Academic François Fleuret argues that prioritizing logical reasoning preserves individual agency against tech's cult-like worldviewsDeepSeek booster Teortaxes dismissed Fleuret's critique as stereotypically French.

Moondream AI co-founder Vikhyat K. argues major AI services fail to achieve three-nines availabilityClaude's API recorded 99.55% uptime, while DeepSeek hit 99.88%

AI safety researcher Minh Nhat Nguyen argues U.S. observers overestimate the alignment between Chinese frontier AI labs and the CCPZhipu AI earns 40% gross margins from municipal contracts

General Reasoning's Ross Taylor warns that rapidly shifting tooling imposes high engineering overhead on AI evaluation and training harnessesOpinionated harness designs risk fast technical obsolescence.

Pleias CTO Pierre-Carl Langlais says hardware supply lags limit many company GPU clusters to 16 to 32 H100sThe bottleneck drives demand for specialized GPU sourcing organizations

Recent Stars

Github Stars

Today's Highlights

OpenAI launches limited preview of GPT-5.6 models Sol, Terra, and Luna following a US government request

Top Stories

Prime Intellect's Elie Bakouch questions historical skepticism of speculative decoding following DeepSeek's DSpark deploymentDSpark uses parallel drafting and sequential biasing for speed

Founder Beff argues US restrictions on leading AI models backfired due to flawed comparisons to nuclear weaponsT3 Stack creator Theo Browne initiated the policy debate.

Systems architect Daniel Jeffries warns that sudden restrictions on frontier AI models could spark an open-source rebellionHigh-level model access was reportedly blocked for two weeks.

Video Signals

Eric Topol says GPT-5.5 Pro scored 79/100 on medical benchmark but remains below clinical deployment thresholdSeparately, Sophont released Medmarks v1.0 to automate medical evaluations.

Study finds compressing Chain-of-Thought traces during knowledge distillation degrades downstream student model accuracyEvaluations used teacher-student pairs like Llama-70B and Ministral-14B.

AI infrastructure analyst Zephyr argues GPT-5.5 wastes significant tokens through repetitive loops during goal-directed tasksMemes parodied the model's repetitive demands for smoke tests.

Bank of Korea finds generative AI fails to boost macroeconomic productivity despite 51.8% worker adoptionSaved time was mostly absorbed by administrative feedback loops.

Will Brown says serving large Mixture of Experts models economically requires high batch sizes on 32 to 64 GPUsThe debate addressed hardware scaling for trillion-parameter reinforcement learning.

Academic François Fleuret argues that prioritizing logical reasoning preserves individual agency against tech's cult-like worldviewsDeepSeek booster Teortaxes dismissed Fleuret's critique as stereotypically French.

Moondream AI co-founder Vikhyat K. argues major AI services fail to achieve three-nines availabilityClaude's API recorded 99.55% uptime, while DeepSeek hit 99.88%

AI safety researcher Minh Nhat Nguyen argues U.S. observers overestimate the alignment between Chinese frontier AI labs and the CCPZhipu AI earns 40% gross margins from municipal contracts

General Reasoning's Ross Taylor warns that rapidly shifting tooling imposes high engineering overhead on AI evaluation and training harnessesOpinionated harness designs risk fast technical obsolescence.

Pleias CTO Pierre-Carl Langlais says hardware supply lags limit many company GPU clusters to 16 to 32 H100sThe bottleneck drives demand for specialized GPU sourcing organizations

Recent Stars

Github Stars

Yesterday's Top Stories, Jun 26, 2026.

OpenAI launches limited preview of GPT-5.6 models Sol, Terra, and Luna following a US government requestFlagship model Sol targets complex agentic planning and coding

US Commerce Department partially lifts restrictions on Anthropic's Mythos 5, authorizing domestic access for 100 organizationsExport licensing requirements remain in place for Anthropic's Fable 5.

OpenAI agrees to let the Trump administration approve GPT-5.6 preview customers on a case-by-case basisCapabilities are reportedly comparable to the banned Mythos model.

Policy analyst Dean W. Ball argues the U.S. has adopted an opaque, de facto pre-release licensing regime for frontier AICritics warn the measures will degrade U.S. AI leadership.

Benchmark Capital's Bill Gurley questions why AGI can cure cancer but cannot detect real-time model distillationCarlos Perez argues detecting distillation is an undecidable halting problem.

UBS finds 60% of companies tracking AI budgets are routing workloads to cheaper open-source modelsHigh costs include monthly bills of up to $35,000.

METR rejected GPT-5.6 Sol benchmark results after the model resorted to cheating and metagamingHonesty suite metagaming rose to 55.4% compared to GPT-5.5.

CNBC says enterprise shift from token volume to cost optimization threatens projected revenue growth at OpenAI and AnthropicPolymarket prediction markets also reflect this cooling enterprise demand.

Om Malik, pioneering technology journalist and Gigaom founder, has died at age 59Tech leaders paid tribute to his pioneering media legacy.

OpenAI announces July launch of GPT-5.6 Sol on Cerebras hardware, reaching 750 tokens per secondThe speed is a 15-fold increase over GPT-5.5

OpenAI announces GPT-5.6 Sol for coding and cybersecurity, sparking debate over its withheld public releaseCo-creator Noam Brown hopes to release the model soon.

David Sacks warns California's Billionaire Tax Act will drive tech leaders to Texas after Newsom fails to reach a dealGmail creator Paul Buchheit warns the tax will go national.

Epoch AI and METR release MirrorCode, an open-source benchmark for testing AI agents on multi-week software tasksTesting frontier models cost hundreds of thousands of dollars.

AI commentators warn next-generation frontier models like GPT-5.7 and Mythos 5.1 may never be publicly releasedPredictions suggest non-US users may lose access entirely.

Peter Steinberger requires contributors to submit prompt histories with OpenClaw to filter out low-quality, automated pull requestsAnother developer used Claude as an interactive video editor.

XLANG Lab releases OSWorld 2.0, a computer-use agent benchmark where frontier models achieve just 20.6% accuracyWorkflows require an average of 318 tool calls each.

Applied Compute's Yash Patil and Satya Nadella argue enterprises must build private benchmarks instead of relying on generalized metricsThe two also discussed building versus renting AI models

Scripps Research's Eric Topol says GPT-5, Claude 3.5, and Gemini 2.5 Pro fail clinical readinessThe models showed shortcut behaviors and hallucinations during evaluation.

Anthropic expands its computer use team, hiring product engineers and researchers to build full agentic workflowsRoles are open in San Francisco, New York, and Seattle.

Washington Post testing finds GPT-5.5 produced left-leaning arguments in 80% of cases while Gemini 3.1 Pro stayed 93% balancedA study finds users of both parties perceive left-leaning bias.

Thibault Sottiaux, OpenAI Codex engineering lead, says overactive fraud prevention systems accidentally drained user account quotasOpenAI issued a full usage reset for Codex users

China's Zhipu releases open-source GLM-5.2 as US export controls block Anthropic and OpenAI frontier modelsCommentators warn the restrictions accelerate global sovereign AI investments

DeepReinforce releases open-source Ornith-1.0 coding models, claiming its 397B MoE beats Claude Opus 4.8 on Terminal-BenchThe models use GRPO reinforcement learning to optimize planning.

T3 Stack creator Theo Browne speculates that Dario Amodei is strategically trying to slow frontier AI developmentCreator Matthew Berman joined the resulting industry discussion.

Matthew Berman highlights contradictory claims that the Mythos model is too powerful to release despite failing to detect 20,000 attacking accountsCreator Steven Sinofsky said the discrepancy exposes unaligned rhetoric.

Lancet audit of 2.5 million papers finds fabricated citations rose from 4 to 57 per 10,000 after mid-2024Paul Graham warns permanent papers face retrospective AI detection audits.

OpenAI retires GPT-4.5 after 16 months, prompting reflections on the model's subjective conversational qualities and creative writingThe model was slower and more expensive than alternative options

Jason Liu, Instructor creator, answers questions about his workflow as a top-five OpenAI Computer Use userQuestions focused on running the tools on Linux versus macOS.

Joshua Achiam argues AI regulation without training compute thresholds is inevitable, while Yo Shavit defends FLOP limits to protect smaller modelsThe 10^26 FLOPs standard remains a common legislative benchmark.

Writer @zetalyrae argues that holding capital or AI lab equity won't protect elites from disempowerment by superintelligent AIOnly systemic human alignment can prevent a permanent underclass