Digg - AI news, before it trends

Today's Highlights

In case you missed it

Google DeepMind's LLM-Lean agent loop resolves 9 open Erdős math problems and proves 44 OEIS conjectures

RH#887|@ROBINHANSON

B(#839|@BEFFJEZOS

ZN#537|@ZACHARYNADO

SK#505|@SEBKRIER

GM#153|@GARYMARCUS

It’s a new day at Digg. Fresh stories are clustering now. 30% complete.

Top Stories

Posts:—|Clusters:—|Next Crawl:—

10

Pope XIV announces partnership with Anthropic to address the ethical and moral implications of advanced AI — The Pope formally accepted Anthropic's invitation to collaborate

5h|3.2M13.1k3.8k

SM#1811|@SAMMCALLISTER

CE#1598|@INTUITMACHINE

CH#1496|@KIMMONISMUS

20

Sakana AI urges Japan to combine foreign frontier models with domestic tech to protect national data sovereignty — Prime Minister Sanae Takaichi hosted the deep-tech strategy roundtable.

9h|494.6k8.8k399

T🐻#1476|@TKASASAGI

HA#18|@HARDMARU

30

Y Combinator's Paul Graham warns that AI-generated emails from startup founders lose all impact once detected — Graham noted a telltale journalistic tone exposes these drafts

8h|454.7k5.7k542

NR#1562|@NATANIELRUIZG

40

Tanishq Mathew Abraham, SophontAI founder, explains how his Ph.D. in deep learning-based virtual staining supports his medical AI workflow — He completed his UC Davis Ph.D. at age 19.

6h|33.4k41936

AJ#486|@JM_ALEXIA

TM#359|@ISCIENCELUVR

50

Andy Masley finds ChatGPT's water usage is 50 to 250 times lower than the viral 'bottle of water' claim — Simon Willison urges OpenAI to release retired GPT-4 specs.

7h|53.8k51770

AM#1735|@ANDYMASLEY

60

Leonardo.AI co-founder Ethan Smith warns that 10x training speedup claims in recent diffusion papers may rely on flawed metrics — Prior studies confirm similar speedup overestimations in Transformer pre-training

10h|4.1k4621

ET#1884|@TORCHCOMPILED

PM#713|@PMINERVINI

Github Stars

(7 days)

manaflow-ai/cmux21H AGO AGO
Integrates Ghostty into a macOS terminal with vertical tabs, sidebar metadata, OSC notifications, and in-app browser for AI coding agents.
759
660
30
495
203
18.9k stars
earendil-works/pi2D AGO AGO
Provides AI agent toolkit with coding CLI, unified LLM API, TUI/web UI libs and Slack bot support.
462
219
53.7k stars
…-group/InferenceBench4D 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-Text6D 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 24, 2026.

Posts: 2,232|Clusters: 591|Stories frozen at midnight PT

10

Google DeepMind's LLM-Lean agent loop resolves 9 open Erdős math problems and proves 44 OEIS conjectures — Each proof cost a few hundred dollars to generate

1d|2.1M13.8k1.8k

CE#1598|@INTUITMACHINE

CH#1496|@KIMMONISMUS

HB#1459|@HAYDNBELFIELD

ST#1446|@_ONIONESQUE

PS#1371|@RECURSEPARADOX

RH#887|@ROBINHANSON

B(#839|@BEFFJEZOS

ZN#537|@ZACHARYNADO

SK#505|@SEBKRIER

GM#153|@GARYMARCUS

20

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.

1d|4.7M#1LIKED61.7k2.3k

CB#999|@CODE_STAR

B(#839|@BEFFJEZOS

TM#359|@ISCIENCELUVR

FF#331|@FRANCOISFLEURET

GM#153|@GARYMARCUS

EM#76|@ELONMUSK

30

Box CEO Aaron Levie argues CEOs overestimate AI readiness because they only see prototypes and ignore implementation work — Steven Sinofsky compared the gap to 1990s office software.

14h|1M10.8k2.3k

SS#1992|@STEVESI

AI#1777|@ALEXOLEGIMAS

CE#1598|@INTUITMACHINE

NR#1562|@NATANIELRUIZG

MS#1005|@MADIATOR

JL#677|@JERRYJLIU0

SK#505|@SEBKRIER

MR#183|@MAITHRA_RAGHU

40

Aidan Clark argues AGI pretraining is now restricted to six industry labs with massive compute resources — Will Depue says private compute will dictate scientific progress

5d|494.6k3.7k844

AD#867|@DORIALEXANDER

BB#840|@WGRATHWOHL

EL#716|@ELIEBAKOUCH

C(#587|@STEPHENLCASPER

AC#263|@ANDREW_N_CARR

WD#254|@WILLDEPUE

AC#241|@_AIDAN_CLARK_

RA#83|@_AROHAN_

NL#64|@NATOLAMBERT

50

AI-generated parody video depicts Andrej Karpathy joining Anthropic in a scene mimicking The Office — The 103-second clip features an AI-generated Steve Carell.

16h|1.6M16k3.8k

DE#1943|@DANIEL_271828

SW#1814|@SARAHDINGWANG

CH#1496|@KIMMONISMUS

MS#1005|@MADIATOR

PS#495|@STEIPETE

60

David Sacks argues software engineering job postings are rising because cheap AI code generation has increased overall software demand — Indeed data shows developer jobs rising after September 2025.

14h|894.6k6.6k1.5k

MC#1818|@PIRROH

B(#839|@BEFFJEZOS

EL#475|@OMARSAR0

BP#357|@BORISMPOWER

C🤗#68|@CLEMENTDELANGUE

70

Nathan Lambert, AI2 post-training lead, clarifies his upcoming book covers all post-training methodologies and secures posttrainingbook.com — The new domain redirects to his ongoing writing at rlhfbook.com

12h|67.1k433260

CR#1444|@CWOLFERESEARCH

FB#1153|@XEOPHON

CB#999|@CODE_STAR

EL#716|@ELIEBAKOUCH

WB#339|@WILLCCBB

AC#263|@ANDREW_N_CARR

NL#64|@NATOLAMBERT

80

Chinese creator CuiMao releases a fictional cinematic parody about Anthropic and CEO Dario Amodei made with seedance2 AI — The dramatic fan-fiction references the floral Claude logo.

22h|952.7k3.4k1.2k

ZE#1497|@ZEPHYR_Z9

CB#999|@CODE_STAR

JM#931|@VENTURETWINS

SR#604|@CLONEOFSIMO

SY#238|@SONGLINYANG4

90

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

21h|331.8k7.3k388

YI#936|@AGIHIPPO

PM#713|@PMINERVINI

LB#567|@SRCHVRS

🇺🇦D#387|@DBAHDANAU

AK#228|@BLACKHC

100

Every co-founder Dan Shipper predicts companies will deploy a single AI super-agent in Slack to manage knowledge work — Automation will scale product managers rather than replacing them.

14h|1.2M2.2k2.9k

DS#1410|@DANSHIPPER

LR#1176|@LENNYSAN

110

George Hotz, comma.ai and tinygrad creator, warns of declining industry quality and a persistent wave of AI-generated 'slop — AI skeptic Gary Marcus endorsed Hotz's critique.

13h|250.6k941529

B(#839|@BEFFJEZOS

KA#488|@YACINEMTB

DW#392|@DEANWBALL

GM#153|@GARYMARCUS

120

Vatican announces Magnifica Humanitas, a papal encyclical arguing AI is not neutral and reflects its creators' values — It compares AI development to the Tower of Babel.

6d|57.9k491111

SH#1488|@HAMANDCHEESE

SÓ#1457|@S_OHEIGEARTAIGH

CF#1356|@CFGEEK

TH#1282|@TIMHWANG

YS#638|@YONASHAV

AC#517|@ANDREWCURRAN_

⿻A#361|@IAMTRASK

130

Anthropic bans users under 18 from accessing its AI services, drawing criticism from Stability AI co-founder Emad Mostaque — Critics argue the ban deprives teenagers of programming resources.

9h|61.7k59468

𝚟⟢#1753|@VIEMCCOY

MN#1179|@MENHGUIN

AE#719|@ADRIENLE

SK#505|@SEBKRIER

DW#392|@DEANWBALL

EM#167|@EMOSTAQUE

140

Elon Musk says xAI finished training its 1.5-trillion-parameter Grok V9-Medium model using Cursor data for supplementary training — Public release is scheduled within two to three weeks.

1h|919.7k9k495

EL#716|@ELIEBAKOUCH

T(#420|@TEORTAXESTEX

EM#76|@ELONMUSK

150

BlackRock CEO Larry Fink Says AI Infrastructure Will Tap Pensions And Savings

5h|675.7k11.4k4.1k

GM#153|@GARYMARCUS

160

Google DeepMind CEO Demis Hassabis says solving mathematical benchmarks like Erdős problems does not constitute true AGI — DeepMind's AI systems recently solved eight Erdős problems.

6h|137k1.6k399

ST#1446|@_ONIONESQUE

T(#420|@TEORTAXESTEX

DA#156|@DAVIDSHOLZ

170

Vaibhav Srivastav, OpenAI developer experience lead, shares a workflow to turn recurring Codex patterns into reusable skills or subagents — Greg Brockman highlighted the method as a self-improvement prompt.

1d|864.1k5.6k#1BOOKMARKED10.7k

V(#1853|@REACH_VB

180

Stanford NLP's Aryaman Arora disputes Greg Brockman's claim that Codex is open source, pointing to its proprietary underlying model — Brockman cited a terminal-based coding agent repository as evidence.

1d|183.9k1.6k272

V(#1853|@REACH_VB

FB#1153|@XEOPHON

AA#678|@ARYAMAN2020

190

xAI releases Grok Build 0.1.219, fixing kitty terminal layout issues and paste-related data loss — Elon Musk highlighted the team's seven-day-a-week development schedule.

6h|#1VIEWS6.3M20.4k458

KA#488|@YACINEMTB

EM#76|@ELONMUSK

200

AMP PBC founder Anjney Midha says cloud GPU rental prices have more than doubled since January 2026 — B200 hourly rental rates reached $4.89 by July 2026

12h|202.6k734253

BI#1402|@BILALTWOVEC

CB#999|@CODE_STAR

AM#527|@ANJNEYMIDHA

TM#359|@ISCIENCELUVR

210

Former Wired editor Chris Anderson argues intelligence has infinite demand, drawing pushback from writer Noah Smith — Smith questions why highly intelligent academics remain relatively low-paid.

14h|158.8k734382

CA#1839|@CHR1SA

CE#1598|@INTUITMACHINE

AD#867|@DORIALEXANDER

220

Andrej Karpathy recalls anecdote that computer scientist Geoffrey Hinton once held the official title of intern at Google — The title reflected Hinton's reporting relationship under Jeff Dean.

14h|241.8k1.5k120

KD#908|@CSPROFKGD

EL#716|@ELIEBAKOUCH

PM#318|@DOCMILANFAR

230

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.

23h|42.8k19957

LN#1709|@CLASHLUKE

AN#565|@UNIXPICKLE

AC#263|@ANDREW_N_CARR

EM#167|@EMOSTAQUE

DA#156|@DAVIDSHOLZ

240

XY Han sparks debate over whether mathematician Paul Erdős would use advanced AI or reject it for human connection — Jeremy Cohen argues Erdős would use AGI for chores.

16h|46.1k1k48

BI#1402|@BILALTWOVEC

B(#839|@BEFFJEZOS

JC#696|@DEEPCOHEN

NS#227|@NSAPHRA

250

Google DeepMind philosopher Henry Shevlin argues AI-generated images of empty indoor waterparks act as a psychological 'cognitive exploit — The images trigger tension between childhood nostalgia and isolation.

6h|809.6k4.5k2.4k

HB#1022|@HERBIEBRADLEY

260

Y Combinator's Garry Tan says his open-source Claude Code setup GStack has surpassed 100,000 GitHub stars — The setup packages 23 tools simulating software development roles

11h|118.7k1.3k439

BC#322|@BCHERNY

GT#270|@GARRYTAN

270

Boris Cherny, Anthropic Claude Code creator, recommends enabling auto mode to run the CLI tool without interrupting permission prompts — This setting allows developers to run parallel background sessions

18h|671.8k4.4k2.6k

PW#1359|@PETERWILDEFORD

BC#322|@BCHERNY

280

George Hotz says AI coding agents bypass problem-solving by commenting out failing tests — Marcus warns this undermines the economic case for AI.

9h|199.6k1.4k648

RO#1220|@KRISHNANROHIT

GM#153|@GARYMARCUS

290

Bilawal Sidhu, former Google PM, demonstrates a generative AI workflow integrating physical environments with 3D computer graphics primitives — The 10-second demo shows wireframes tracing a stone monument.

14h|30.8k374184

FO#1996|@FOFRAI

BS#1909|@BILAWALSIDHU

LB#55|@GIFFMANA

300

The UK AI Safety Institute has scaled to become the world's largest and best-funded government AI security group — Its technical staff conducts safety evaluations and red-teaming operations.

12h|19k38465

XD#1905|@ALXNDRDAVIES

SÓ#1457|@S_OHEIGEARTAIGH

JW#979|@J_ASMINEWANG

IH#799|@SOUNDBOY