Digg - AI news, before it trends

Today's Highlights

In case you missed it

Trump administration reinstates policy requiring O-1 visa holders to leave the US when applying for green cards, affecting researchers at OpenAI

NL#64|@NATOLAMBERT

It’s a new day at Digg. Fresh stories are clustering now. 35% complete.

Top Stories

Posts:—|Clusters:—|Next Crawl:—

10

Nomad List founder Pieter Levels reports a Claude Code CLI bug that resets his effort setting to medium daily — A research engineer suggests the issue affects outdated CLI versions.

1h|34.6k43936

SM#1811|@SAMMCALLISTER

JS#1715|@JARREDSUMNER

@L#849|@LEVELSIO

20

Remote OK founder Pieter Levels argues Japan lacks proprietary foundational models, noting Rakuten AI is a fine-tune of China's DeepSeek — Levels was defending a Chinese cat-themed physical AI device.

4h|62.7k65067

@L#849|@LEVELSIO

T(#420|@TEORTAXESTEX

30

Chinese creator CuiMao releases the final video in a series detailing the origin story of Anthropic CEO Dario Amodei — The seedance2 collaboration features a hidden Easter egg.

4h|35.4k16651

ZE#1497|@ZEPHYR_Z9

SR#604|@CLONEOFSIMO

40

Google DeepMind's Gemma 4 hits 120 million downloads in two weeks, with a demo showing offline on-device vision and tool calling — An experimental camera app solved handwritten math equations entirely offline.

2h|7.1k12820

FB#1153|@XEOPHON

OS#490|@OSANSEVIERO

50

Vision Transformer co-author Lucas Beyer argues claims that AI wrappers transform businesses suggest proponents haven't actually used them — Academic Pasquale Minervini also dismissed the wrappers' enterprise value.

5h|3.1k950

PM#713|@PMINERVINI

GI#347|@GEOFFREYIRVING

LB#55|@GIFFMANA

60

Emad Mostaque says a single researcher can now manage several thousand GPUs for AI training with modern automation — Historically, one researcher managed 1,000 A100 GPUs.

6h|12.4k7313

EM#167|@EMOSTAQUE

DA#156|@DAVIDSHOLZ

70

Andrej Karpathy joining Anthropic sparks industry debate over the prestige of Member of Technical Staff titles — Yi Tay says AI research and engineering roles are merging

3h|5.7k415

YI#936|@AGIHIPPO

🇺🇦D#387|@DBAHDANAU

Github Stars

(7 days)

manaflow-ai/cmux9H AGO AGO
Integrates Ghostty into a macOS terminal with vertical tabs, sidebar metadata, OSC notifications, and in-app browser for AI coding agents.
660
30
495
203
339
18.3k stars
earendil-works/pi1D AGO AGO
Provides AI agent toolkit with coding CLI, unified LLM API, TUI/web UI libs and Slack bot support.
462
219
53.3k stars
…-group/InferenceBench3D AGO AGO
Evaluates autonomous AI agents optimizing LLM serving like vLLM in open-ended scenarios with quality and integrity gates.
1224
1066
16 stars
sapientinc/HRM-Text4D AGO AGO
Pretrains 1B HRM text models with hierarchical reasoning, task completion, and low-compute PrefixLM training.
1047
707
565 stars

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

Yesterday's Top Stories, May 23, 2026.

Posts: 2,659|Clusters: 664|Stories frozen at midnight PT

10

Trump administration reinstates policy requiring O-1 visa holders to leave the US when applying for green cards, affecting researchers at OpenAI — Up to half of OpenAI researchers on O-1 visas could be impacted.

1d|1.2M#1LIKED37.5k1.7k

DE#1943|@DANIEL_271828

AK#1675|@ANSHULKUNDAJE

CH#1496|@KIMMONISMUS

RO#1220|@KRISHNANROHIT

AK#885|@ANDREY_KURENKOV

ZN#537|@ZACHARYNADO

SK#505|@SEBKRIER

NL#64|@NATOLAMBERT

20

Anthropic finalized a classified contract with the National Security Agency and Defense Department permitting AI model deployment on classified systems while omitting standard any lawful use language and barring use on Americans' data — White House approved $9 billion for spy agencies to buy advanced chips.

1d|#1VIEWS1.8M15.6k820

NI#1902|@NATHANPMYOUNG

JA#837|@APPLES_JIMMY

YS#638|@YONASHAV

SZ#58|@SUCHENZANG

30

OpenAI rolled back an optimization in Codex that reduced cache hit rates and caused faster-than-expected usage limit depletion during long-running sessions — Engineering lead Thibault Sottiaux reset limits across all accounts.

9h|683.3k9.1k519

V(#1853|@REACH_VB

VI#1195|@VIKHYATK

TI#768|@THSOTTIAUX

40

Matthew Barnett posts analysis of Eliezer Yudkowsky's calibration on AI doom predictions centered on a 2016 statement about Turing test timelines before the end of the world — NathanpmYoung replies that the statement offers limited evidence on timelines.

22h|79.5k795104

NI#1902|@NATHANPMYOUNG

JS#1389|@JSEVILLAMOL

MB#1343|@MATTHEWJBAR

EB#1325|@BETHMAYBARNES

MN#1179|@MENHGUIN

T(#420|@TEORTAXESTEX

50

Anthropic releases initial Project Glasswing report showing its unreleased Mythos Preview models found over 10,000 vulnerabilities and blocked a $1.5 million fraud attempt while outperforming GPT-5.5 on exploit benchmarks — Anthropic will expand access to more US and allied government partners first.

1d|107.5k957144

CE#1598|@INTUITMACHINE

LA#980|@SCALING01

SK#505|@SEBKRIER

BC#322|@BCHERNY

GM#153|@GARYMARCUS

60

Peter Steinberger, OpenAI research engineer, routes coding tasks through cmux plus Codex CLI while handling other activities in the Codex Mac app — cmux is a Ghostty-based terminal at manaflow-ai/cmux on GitHub.

22h|411.1k3.2k2.9k

V(#1853|@REACH_VB

CA#729|@CHRISALBON

PS#495|@STEIPETE

70

OpenAI's use of multi-hour AI reasoning to solve an Erdős math problem sparks debate over neurosymbolic verification — Critics questioned if the AI used Lean for verification.

6d|120.2k3.3k472

ND#1894|@NICKADOBOS

RP#1032|@ROHANPAUL_AI

AC#517|@ANDREWCURRAN_

WD#254|@WILLDEPUE

80

Thibault Sottiaux, Engineering Lead on OpenAI's Codex team, proposed adding batch compute support to Codex under the label slow mode for handling compute-intensive coding tasks through grouped or delayed processing — Vaibhav Srivastav supported the proposal for long-running goal tasks.

7h|124.2k3.7k158

ND#1894|@NICKADOBOS

V(#1853|@REACH_VB

MB#1759|@MATTHEWBERMAN

RO#1220|@KRISHNANROHIT

TI#768|@THSOTTIAUX

90

Perplexity releases Bumblebee, an open-source read-only scanner for macOS and Linux, that inventories packages, browser extensions, and configurations to detect supply-chain risks — It integrates with Perplexity Computer to trigger automated deeper scans.

1d|557.8k2.2k#1BOOKMARKED3.7k

DY#798|@DENISYARATS

AS#118|@ARAVSRINIVAS

100

Pieter Levels says skills features in AI coding agents are overrated and equivalent to basic text files, preferring direct instructions on tasks and methods instead — Austen Allred countered that skills enable reuse across sessions.

20h|203.3k1.1k269

ND#1894|@NICKADOBOS

AA#1540|@AUSTEN

MS#1005|@MADIATOR

@L#849|@LEVELSIO

110

Guillermo Rauch, Vercel co-founder and CEO, posts invitation for working AI-built products requesting replies that include product URLs and primary models used — Siqi Chen replied with miraviewer.org titled MiraViewer

11h|335.2k2.3k2.1k

120

OpenAI observation shows Claude 5.5 Pro gaining professional adoption through deliberate packaging and vertical use-case marketing rather than raw model performance — Lawyer chat thread highlighted hallucinations and iterative workflow adjustments.

4h|175.6k1.9k289

MB#1759|@MATTHEWBERMAN

KA#488|@YACINEMTB

130

SemiAnalysis posted data from 174,264 agentic coding sessions showing 42% of runtime on CPU tasks versus 58% on GPU inference and highlighted cloud pricing mismatches with per-token monetization — Median per-turn time measured 5.13 seconds.

15h|209.5k842474

ZE#1497|@ZEPHYR_Z9

140

Andrej Karpathy stated in an August 5 2017 post that gradient descent can write code better than humans as the screenshot resurfaces amid later AI coding tools — The post shows 181 replies 716 reposts and 3K likes.

14h|383.8k12.4k1.1k

B(#839|@BEFFJEZOS

150

Niloofar Mireshghallah overhears discussion of world models and grounded video generation while resting in a Copenhagen park during burnout recovery — Nathan Lambert replies with support for her trip.

18h|12.8k2148

CY#1301|@CHHAVIYADAV_

MJ#1194|@MIMANSAJ

TC#1050|@TUHINCHAKR

NI#303|@NILOOFAR_MIRE

AG#154|@ANDREWGWILS

NL#64|@NATOLAMBERT

160

Six-Person Team Builds Task-Specific AI Models 4-8x Faster Than OpenAI

15h|163.7k1.9k2.4k

GT#270|@GARRYTAN

170

OpenAI announced that an internal general-purpose reasoning model produced new constructions for the planar unit distance problem first posed by Paul Erdős in 1946 — The model was not built specifically for mathematics problems.

3d|19.3k1829

BG#1447|@BEN_GOLUB

AN#1409|@ARAN_NAYEBI

PM#713|@PMINERVINI

SK#505|@SEBKRIER

180

Stony Brook's Tuhin Chakrabarty finds a prize-winning Granta story contains 1,236 phrases copied from online fanfiction — An n-gram tool mapped the phrases to fanfiction sites.

1d|31.6k14945

TC#1050|@TUHINCHAKR

CT#570|@CHENHAOTAN

190

Pieter Levels, Dutch indie hacker and serial entrepreneur behind Remote OK and Nomad List, says he has not written code in six months — Nick Dobos reports staying hands-off from coding for months.

20h|1.1M2.6k497

ND#1894|@NICKADOBOS

@L#849|@LEVELSIO

200

Composer 2.5 handled approximately 80 percent of requested features in a 3D mouse cursor racing game from planning through implementation, per engineer Danny Limanseta — Switched to Opus 4.7 MAX for 360-degree loops and visual effects.

13h|95.8k1k374

MA#470|@MARTIN_CASADO

SR#23|@SRUSH_NLP

210

Yuchen Jin posted about moving from 15 years of manual tasks including syntax memorization and merge conflicts to prompting AI agents for code changes — Andrew Trask replied that the manual coding era stayed enjoyable

11h|135.3k4.1k334

YJ#427|@YUCHENJ_UW

⿻A#361|@IAMTRASK

220

Unicode String Prompts Google AI to Generate Love Poetry

19h|165.4k1.1k208

PT#643|@ELDER_PLINIUS

230

Thibault Sottiaux names GPT Goblin 1.0 Extra High /Fast in a post that references song lyrics but offers no technical details or announcement — Victor Taelin replied with a single skeptical emoji.

9h|118.7k1.6k70

MB#1759|@MATTHEWBERMAN

TA#1047|@VICTORTAELIN

TI#768|@THSOTTIAUX

240

Posts on X show GPT-5.6 generating a minimalist Lumen Notes web app UI from prompts with no interface instructions, indicating cleaner defaults than GPT-5.5 — Users cite the change as evidence of ongoing UI refinement

14h|273.6k1.5k207

CH#1496|@KIMMONISMUS

AC#517|@ANDREWCURRAN_

250

OpenAI Codex autonomously operates an iPhone simulator via its computer-use interface to build a chat app feature and complete full bug testing and debugging without human input — 32-second split-screen recording shows real-time UI validation steps.

13h|274.1k1.2k451

260

xAI removes the distinction between researcher and engineer titles, unifying all technical staff under the single designation of engineer across the company — Elon Musk called the old system a two-tier academic convention.

11h|144.7k1.4k200

T(#420|@TEORTAXESTEX

MB#20|@MILES_BRUNDAGE

270

Google Flow opens Gemini Omni to all users, enabling natural language video editing directly inside the tool with non-subscribers limited to two free generations per day — Demo shows one prompt generating a 10-second nine-scene montage.

14h|29.4k42599

KM#421|@KORYMATH

👩‍💻P#271|@DYNAMICWEBPAIGE

280

Steven Rosenbaum, media entrepreneur and executive director of the Sustainable Media Center, published The Future of Truth with content from ChatGPT that included more than a dozen fabricated quotes, the New York Times found — Rosenbaum used the model as a research partner and felt misled.

12h|344.3k48081

MS#1323|@MICSOLANA

TC#1050|@TUHINCHAKR

290

Avital Balwit writes in The Free Press that frontier AI development is forcing technologists to confront questions of faith and the divine through ambitions such as curing diseases and halting aging — Essay titled 'Searching for God in Silicon Valley' appears on thefp.com.

1d|15.9k13869

LI#1445|@LIVGORTON

AB#1395|@AVITALBALWIT

JL#1106|@ARTIRKEL

KI#371|@KIPPERRII

300

AI Contributors Warn Of Citizenship Barriers And Future Denaturalization Risks

1d|95.1k2k72

ZN#537|@ZACHARYNADO