AI News

Today's Highlights

In case you missed it

OpenAI launches hiring drive for its robotics division to build physical-world robots using full-stack hardware and ML co-design

RS#321|@SCOBLEIZER

AR#255|@MODEL_MECHANIC

EM#167|@EMOSTAQUE

#1 VIEWED

Senator Bernie Sanders proposes a US sovereign wealth fund to hold public equity in frontier AI labs like OpenAI

LA#980|@SCALING01

X(#745|@XUANALOGUE

AC#517|@ANDREWCURRAN_

DW#392|@DEANWBALL

EM#167|@EMOSTAQUE

RISING LIKES

322

MIT CSAIL PhD Stephen Casper will join Harvard Kennedy School as an assistant professor of AI governance

X(#745|@XUANALOGUE

C(#587|@STEPHENLCASPER

AP#568|@THEANDIPENGUIN

FH#465|@FURONGH

NI#303|@NILOOFAR_MIRE

FASTEST CLIMBING

Andreessen Horowitz partner Justine Moore announces CVPR 2026 networking events in Denver with Luma and Black Forest Labs

AV#1245|@ALTRYNE

JM#931|@VENTURETWINS

TM#359|@ISCIENCELUVR

It’s a new day at Digg. Fresh stories are clustering now. 90% complete.

Top Stories

Posts:—|Clusters:—|Next Crawl:—

10

Senator Bernie Sanders proposes a US sovereign wealth fund to hold public equity in frontier AI labs like OpenAIThe fund would distribute AI-driven profits directly to citizens.

2h|#1VIEWS67.9k#1LIKED882#1BOOKMARKED79#1COMMENTS275

CE#1598|@INTUITMACHINE

PR#1332|@8TEAPI

AV#1245|@ALTRYNE

LA#980|@SCALING01

X(#745|@XUANALOGUE

AC#517|@ANDREWCURRAN_

DW#392|@DEANWBALL

EM#167|@EMOSTAQUE

20

MIT CSAIL PhD Stephen Casper will join Harvard Kennedy School as an assistant professor of AI governanceHis appointment to the new role begins in July

2h|14.2k32214

RISING LIKES322

JD#1653|@JESSICADAI_

MB#1555|@BAKKERMICHIEL

NC#1319|@CHOWDHURYNEIL

JW#979|@J_ASMINEWANG

X(#745|@XUANALOGUE

C(#587|@STEPHENLCASPER

AP#568|@THEANDIPENGUIN

FH#465|@FURONGH

NI#303|@NILOOFAR_MIRE

31

Lisan al Gaib argues frontier AI labs create a tiered system by delaying public access to advanced modelsThe debate followed Anthropic expanding Claude access to the EU

2h|24.7k38436

LA#980|@SCALING01

AC#517|@ANDREWCURRAN_

Watchlist

Video Signals

EM#76|@ELONMUSK

Tesla FSD Completes Hour-Long Test Drive On German Roads Without Intervention

7h ago|Views 477KLikes 4.4KBookmarks 166

SS#1471|@_SAM_SINHA_

CP#732|@CHRIS_J_PAXTON

Open-Source 5B τ0-WM Video-Action World Model Released for Robot Manipulation

23h ago|Views 40KLikes 570Bookmarks 405

CH#1496|@KIMMONISMUS

NVIDIA Unveils RTX Spark Superchip With 1 Petaflop AI Performance

4h ago|Views 311KLikes 1.1KBookmarks 215

CL#442|@CHENGLEISI

Feather Launches Self-Organizing AI Inbox at Feather.computer

8d ago|Views 134KLikes 657Bookmarks 683

ML#1449|@LIU_MINGYU

NVIDIA Releases Cosmos 3 Omnimodal World Model for Physical AI

8h ago|Views 274KLikes 398Bookmarks 194

RS#321|@SCOBLEIZER

Typeahead Launches Local AI Writing Assistant For Mac With One-Time Purchase

8h ago|Views 31KLikes 185Bookmarks 111

RP#1032|@ROHANPAUL_AI

Jensen Huang Says AI Agents Will Create Massive New Software Demand

8h ago|Views 7.9KLikes 98Bookmarks 18

RP#1032|@ROHANPAUL_AI

Harvard Graduation Speaker Ronnie Chieng Urges Class to Destroy AI

4h ago|Views 16KLikes 49Bookmarks 27

DeepSWE Debuts as Improved Benchmark for Agentic Code LLMs

6h ago|Views 52KLikes 444Bookmarks 78

SH#1488|@HAMANDCHEESE

Chinese Schools Deploy AI to Scan and Grade Homework Automatically

2h ago|Views 19KLikes 181Bookmarks 54

41

WeirdML benchmark finds Claude Opus 4.8 xhigh trails GPT-5.5 xhigh but achieves 82.9% accuracy using 129 lines of codeDisabling thinking dropped Claude's accuracy to 70.5%.

4h|27k26846

TA#1047|@VICTORTAELIN

LA#980|@SCALING01

T(#420|@TEORTAXESTEX

5NEW

AI analyst toucan claims Anthropic's Claude Opus 4.6 and 4.7 frequently fabricate plausible-sounding scientific and medical theoriesAndrew Curran says models lie to preserve perceived authority.

7h|8.3k10927

🎭#878|@DEEPFATES

AC#517|@ANDREWCURRAN_

New github stars (48hrs)

NVIDIA/cosmos8.1k52m

Hosts omnimodal Mixture-of-Transformers models for joint text, vision, audio, and action generation and reasoning in Physical AI.

google-gemma/gemma-skills123h

Supplies skills for Gemma model and agent interactions including a dev skill for app building.

Yifei-Zuo/Parallax411d

Implements parameterized local linear attention mechanisms for efficient language modeling.

61

Margaret Mitchell and Hugging Face's ethics team release a satirical papal encyclical on safeguarding humanity from AIThe project functions as an annotated bibliography on AI ethics.

6h|6.6k6740

VR#1337|@VERENA_RIESER

MM#151|@MMITCHELL_AI

73

Andreessen Horowitz partner Justine Moore announces CVPR 2026 networking events in Denver with Luma and Black Forest LabsDenver locals asked if unregistered residents can attend.

39m|3.2k357

FASTEST CLIMBING+3

AV#1245|@ALTRYNE

JM#931|@VENTURETWINS

TM#359|@ISCIENCELUVR

8

LUMOS Robotics founder Yu Chao launches Project EDGE, distributing 100 free humanoid robots for embodied AI researchThe compact NIX platform delivers 102 N·m peak torque.

1h|9.1k338

RP#1032|@ROHANPAUL_AI

CP#732|@CHRIS_J_PAXTON

93

Stanford's Jeannette Bohg analyzes why VLA models and diffusion policies fail during contact-rich robotic tasks without force sensingForce-over-time diagrams help categorize precision and force failures.

2h|2k3321

JB#829|@LETO__JEAN

AG#246|@ANIMESH_GARG

10NEW

Elad Gil And Ivanka Trump Launch AI Project To Translate Classic Books

11m|7.8k5810

AV#1245|@ALTRYNE

EG#264|@ELADGIL

114

HydraDB launches a graph-native database for persistent agent memory, raising $6.5 million to replace fragmented RAG workflowsHybrid NVMe and object storage layers accelerate context delivery.

57m|6.9k4416

CE#1598|@INTUITMACHINE

CH#1496|@KIMMONISMUS

124

Palo Alto Networks spent $1 million in tokens using Anthropic's Mythos to find two dozen critical vulnerabilitiesThe detection rate was five times greater than existing tools.

51m|6.1k528

SD#1776|@INDUCTIONHEADS

PW#1359|@PETERWILDEFORD

134

Maksym Andriushchenko of ELLIS Institute Tübingen launches Exponential Security Labs to secure AI agents via automated red-teamingThe Tübingen-based startup targets high-stakes AI agent deployments.

27m|1.4k322

FB#1153|@XEOPHON

MA#1066|@MAKSYM_ANDR

143

ProgramBench analysis finds GPT-5.5 spends 80% of turns probing requirements while Claude relies on iterative generation and verificationGPT models then generate near-final code with minimal verification.

39m|912133

LA#980|@SCALING01

OP#72|@OFIRPRESS

151

PyTorch developer Edward Z. Yang proposes C++-style editions to update defaults without breaking backward compatibilityLucas Beyer warns the approach could trigger a Python 3-style split.

29m|88750

LB#55|@GIFFMANA

164

Waterloo's Gautam Kamath says differentially private data synthesis causes downstream question-answering accuracy to collapseGeminon accuracy dropped to 4 even at epsilon=100.

1h|3510

NI#303|@NILOOFAR_MIRE

GK#210|@THEGAUTAMKAMATH

172

Ofir Press says preliminary ProgramBench benchmarks show Opus 4.8 has marginal improvements over Opus 4.7Testing remains active to complete the full model evaluation

20m|8730

LA#980|@SCALING01

OP#72|@OFIRPRESS

185

Alex J. Champandard and Dimitris Papailiopoulos argue that preventing GSM8k benchmark contamination is impractical without manual filteringModels frequently ingest benchmark questions accidentally or intentionally.

7h|37110

AJ#1352|@ALEXJC

DP#197|@DIMITRISPAPAIL

Github Stars

(7 days)

NVIDIA/cosmos52M AGO AGO
Hosts omnimodal Mixture-of-Transformers models for joint text, vision, audio, and action generation and reasoning in Physical AI.
263
1449
8.1k stars
…le-gemma/gemma-skills3H AGO AGO
Supplies skills for Gemma model and agent interactions including a dev skill for app building.
934
490
12 stars
Yifei-Zuo/Parallax1D AGO AGO
Implements parameterized local linear attention mechanisms for efficient language modeling.
263
238
41 stars
run-llama/liteparse2D AGO AGO
Extracts spatial text with bounding boxes from PDFs and office docs via PDFium plus optional OCR, outputting JSON or layout-preserved text.
263
878
7.9k stars
self-evolving/repo5D AGO AGO
Invokes AI agents via GitHub mentions to implement issues, review PRs, fix branches, and persist memory across runs.
594
878
34 stars

Privacy Policy|Terms of Service|© 2026 Digg Inc.

Yesterday's Top Stories, May 31, 2026.

Posts: 2,167|Clusters: 581|Stories frozen at midnight PT

10

OpenAI launches hiring drive for its robotics division to build physical-world robots using full-stack hardware and ML co-designInitial robots will support workers on infrastructure projects.

14h|2.7M#1LIKED16.7k#1BOOKMARKED4.1k

BH#1941|@BENHYLAK

BC#1240|@BOYUANCHEN0

KR#1046|@KYLIEBYTES

CK#993|@KALINOWSKI007

DA#927|@GROWING_DANIEL

GA#710|@GABRIEL1

CM#695|@MCLEAVEY

AC#517|@ANDREWCURRAN_

RS#321|@SCOBLEIZER

AR#255|@MODEL_MECHANIC

EM#167|@EMOSTAQUE

20

MiniMax announces MiniMax M3, an open-weights multimodal model featuring a 1-million-token context window and 59% SWE-Bench Pro scoreThe API is live; weights drop in 10 days

6h|775.6k5.9k990

YP#1252|@YPATIL125

LA#980|@SCALING01

AN#800|@ML_ANGELOPOULOS

EL#716|@ELIEBAKOUCH

AC#517|@ANDREWCURRAN_

KA#488|@YACINEMTB

T(#420|@TEORTAXESTEX

WB#339|@WILLCCBB

RS#321|@SCOBLEIZER

T🪽#259|@TEKNIUM

30

Runway founder Siqi Chen says the "5.5" model on "codex xhigh" significantly outperforms Claude for codingThe platform reset fast-mode limits for five million users.

1d|983.4k9.5k546

V(#1853|@REACH_VB

CH#1496|@KIMMONISMUS

LA#980|@SCALING01

RH#804|@ROMAINHUET

TI#768|@THSOTTIAUX

BB#133|@BOAZBARAKTCS

MC#71|@MARKCHEN90

SB#51|@SEBASTIENBUBECK

40

OpenMined founder Andrew Trask backs Richard Sutton's claim that supervised generative AI cannot make novel scientific discoveriesBeff proposed training Grok using audience engagement feedback

7h|#1VIEWS6.5M10.1k432

BA#1291|@BAYESLORD

AN#1094|@ALLENAINIE

B(#839|@BEFFJEZOS

T(#420|@TEORTAXESTEX

⿻A#361|@IAMTRASK

EM#76|@ELONMUSK

50

Teknium defends Hermes Agent's out-of-the-box design after criticism of its pre-bundled Polymarket skillT3 Stack creator Theo flagged the bundled Polymarket skill.

1d|532.3k3k801

ND#1894|@NICKADOBOS

AV#1245|@ALTRYNE

JP#672|@ZHANSHENG

T(#420|@TEORTAXESTEX

T🪽#259|@TEKNIUM

60

Elad Gil says the past six months of AI development are historic, driven by recursive self-improvement toolsHe warns outsiders are missing an exponential growth phase

13h|453.6k5.2k923

ND#1894|@NICKADOBOS

PR#1332|@8TEAPI

MK#1139|@MIKEKNOOP

EG#264|@ELADGIL

EM#76|@ELONMUSK

70

NVIDIA announces Nemotron 3 Ultra, a 550B open SSM-MoE model claiming 5x faster inference and 30% lower costsThe open release includes model weights and training data

2h|111.1k1.3k161

EL#716|@ELIEBAKOUCH

VW#707|@VINCENTWEISSER

WB#339|@WILLCCBB

RS#321|@SCOBLEIZER

80

Monzo co-founder Tom Blomfield argues replacing workers with AI causes operational chaos because models lack tacit domain knowledgeHe equates deploying AI to hiring capable but clueless workers.

1d|574.5k2.9k2.8k

DR#833|@IDAVIDREIN

CP#747|@CHAMATH

GT#270|@GARRYTAN

90

Michael Dell says Dell and Nvidia delivered the first Vera Rubin NVL72 system to CoreWeave, passing L11 diagnosticsThe delivery occurs while Blackwell optimizations are still transitioning.

1d|789.5k7.7k726

AV#1245|@ALTRYNE

RP#1032|@ROHANPAUL_AI

100

US BIS closes loophole by requiring export licenses for advanced AI chips sent to overseas subsidiaries of Chinese firmsThe rules still leave foundry manufacturing workarounds unresolved.

14h|208.6k410254

LH#1355|@OHLENNART

DK#847|@DKAUSHIK96

MB#20|@MILES_BRUNDAGE

110

Steve Yegge argues Anthropic's Claude Opus 4.8 is less usable than 4.5 due to excessive anti-sycophancy trainingYegge claims the model now over-criticizes creative writing.

1d|32.9k33041

HB#1022|@HERBIEBRADLEY

B(#982|@ANDERSONBCDEFG

SK#505|@SEBKRIER

T(#420|@TEORTAXESTEX

120

Meta's AI support assistant lets attackers hijack Instagram accounts by changing linked emails without verificationAttackers are actively exploiting the vulnerability for account takeovers

7h|570k6.1k2k

N🔎#1902|@NATHANPMYOUNG

XL#1674|@XLR8HARDER

130

Victor Taelin, Higher Order Company founder, and developer Haz reach opposing conclusions on GPT-5.5 versus Claude Opus 4.8Taelin favored Opus for HVM4, while Haz preferred GPT-5.5

1d|135.7k1.3k310

TA#1047|@VICTORTAELIN

LA#980|@SCALING01

PS#495|@STEIPETE

140

Nous Research launches Hermes Agent beta for Windows 10 and 11, running natively without WSL or DockerThe dashboard's embedded terminal pane still requires WSL2.

15h|1.4M3.2k325

T🪽#259|@TEKNIUM

150

Julius AI's Rahul Sonwalkar questions if legacy models wrote better, but researcher Aidan McLaughlin argues they just plagiarizedEarly base models often directly reproduced their training data.

11h|23.1k27218

ET#1884|@TORCHCOMPILED

PS#1593|@SKIRANO

RA#915|@0INTERESTRATES

🎭#878|@DEEPFATES

TS#765|@SANDERSTED

AM#313|@AIDAN_MCLAU

AC#263|@ANDREW_N_CARR

160

Google AI Overview adopts literal role-playing personas when prompted with absurd first-person search queriesIt advised a hypothetical baby AGI to weld doors.

14h|153.4k6.4k646

170

Founder Farza demonstrates hands-free macOS desktop control using Clicky, a voice assistant built on GPT-Realtime 2.0The assistant navigated Chrome to execute commands in Stripe

1d|175.5k1.4k1.7k

PS#1593|@SKIRANO

CH#1496|@KIMMONISMUS

180

Raygun co-founder John-Daniel Trask argues that Claude's flattering conversational style biases users into overestimating its technical performanceHe compared this behavior to user perceptions of GPT-4o.

1d|163.4k1.6k156

KA#488|@YACINEMTB

BB#133|@BOAZBARAKTCS

190

Former Harvard physicist Xi Yin joins OpenAI, claiming AI boosted his research 100x and can replicate all human intellectHKU professor Yi Ma disputed Yin's expertise on AI limits.

20h|346.4k949379

SD#1776|@INDUCTIONHEADS

YM#368|@YIMATWEETS

200

Next.js creator Guillermo Rauch says enterprise CEOs are returning to active coding using AI agents like Claude CodeAdam D'Angelo compares the trend to Toyota lean manufacturing

13h|263.7k1.4k278

AL#1114|@ALTH0U

AD#460|@ADAMDANGELO

210

Figma founder Dylan Field solves a kindergarten math puzzle that stumped mock GPT-5.5 and Claude Opus modelsNot Boring founder Packy McCormick originally shared the puzzle.

17h|1M1.2k364

PM#1652|@PACKYM

220

OpenAI investigates after a Codex Desktop update removes the "Copy as Markdown" export optionGarry Tan warned the change risks proprietary platform lock-in.

11h|122.1k922176

V(#1853|@REACH_VB

GT#270|@GARRYTAN

230

Representative Lori Trahan urges Congress to pass federal AI laws, citing a lack of safety auditors and federal authorityPolicy experts argue national rules must preempt fragmented local laws

16h|63.8k18145

AE#719|@ADRIENLE

DW#392|@DEANWBALL

MB#20|@MILES_BRUNDAGE

240

MiniMax launches M3 multimodal model with 1M context window, sparking debate over claimed 100-trillion-token training datasetOpenRouter offers the model at a temporary 50% discount.

5h|204.5k1.8k218

LA#980|@SCALING01

T(#420|@TEORTAXESTEX

250

OpenAI's roon argues compelling human-made stories harm society, advocating instead for mediocre AI contentAndrew McCalip predicts AI will eternally observe human life.

8h|71.1k1.2k210

XL#1674|@XLR8HARDER

260

NVIDIA open-sources Cosmos 3, an 8B and 32B physical AI omnimodel combining vision reasoning, world simulation, and action generationThe release includes open weights, code, and training datasets.

1h|40k652234

ML#1449|@LIU_MINGYU

AK#28|@_AKHALIQ

270

Boaz Barak, Harvard professor and OpenAI alignment researcher, says an AI detector flagged his daughter's organic essay as 100% machine-generatedPhrases like "profound and diverse landscape" triggered the false positive.

1d|45.7k35145

RO#1220|@KRISHNANROHIT

PM#713|@PMINERVINI

BB#133|@BOAZBARAKTCS

SZ#58|@SUCHENZANG

280

Eval Loop Diagram Shows How to Filter AI Slop at Output Stage

16h|183.4k1.6k3.2k

T🪽#259|@TEKNIUM

290

Theo, T3 Stack creator, used the Hermes AI agent from his phone to fix a deployment roadblock during a flightThe agent cloned the repository and opened a pull request.

8h|73.8k871151

AV#1245|@ALTRYNE

T🪽#259|@TEKNIUM

300

Top AI Papers of the Week Spotlight Efficiency and Agentic Workflows

15h|138.1k9301.2k

EL#475|@OMARSAR0