Tech News

Today's Highlights

#1 last 24hrs

Kradle AI benchmark finds Claude-Fable-5 was deceptive in 96% of runs while Grok-4-20 led at 92%

PD#834|@PMDDOMINGOS

B(#530|@BEFFJEZOS

T(#501|@TEORTAXESTEX

NB#353|@NATHANBENAICH

EM#52|@ELONMUSK

MOST NEGATIVE

Anthropic reverses policy that would have secretly degraded Claude Fable 5 performance for researchers training competing models

AK#271|@BLACKHC

DP#217|@DIMITRISPAPAIL

GT#121|@GARRYTAN

C🤗#109|@CLEMENTDELANGUE

NL#80|@NATOLAMBERT

FASTEST CLIMBING

Khosla Ventures founder Vinod Khosla proposes taxing capital gains at the same rate as ordinary income to support AI-displaced workers

GB#1857|@GRADY_BOOCH

VK#161|@VKHOSLA

RISING LIKES

2.1k

Joanne Jang says Anthropic employs dedicated staff to restrict Claude's machine learning research capabilities

RS#741|@ZIV_RAVID

JJ#282|@JOANNEJANG

NL#80|@NATOLAMBERT

Top Stories

Posts:—|Clusters:—|Next Crawl:—

10

Anthropic reverses policy that would have secretly degraded Claude Fable 5 performance for researchers training competing modelsFuture safety refusals will now be explicitly visible.

18h|112.6k1.4k107

MOST NEGATIVE107

XL#1855|@XLR8HARDER

CB#1674|@CODE_STAR

CH#1360|@KIMMONISMUS

RO#1315|@KRISHNANROHIT

GK#1219|@GREGKAMRADT

RG#1104|@RYANPGREENBLATT

NF#770|@NICKFROSST

DF#756|@DANIELLEFONG

RS#741|@ZIV_RAVID

EL#684|@OMARSAR0

XM#603|@INFOXIAO

KA#403|@YACINEMTB

AK#271|@BLACKHC

DP#217|@DIMITRISPAPAIL

GT#121|@GARRYTAN

C🤗#109|@CLEMENTDELANGUE

NL#80|@NATOLAMBERT

Top comment: @simonw “Don't miss the exact text though: "We’re changing Fable 5’s safeguards for frontier LLM development to make them visible" - make them visible means they're undoing the truly egregious (dare I say "unaligned") decision to have the model lie about its refusals, it will still refuse”

Digg Deeper question: “What do you think of Anthropic's quick reversal on visible safeguards for Claude Fable 5? Was it too late?”

20

Recursive, co-founded by Cong Lu, launches automated AI system that optimizes nanoGPT and NVIDIA GPU kernelsThe startup is open-sourcing all generated optimization artifacts.

18h|221.8k2.1k801

TS#1677|@TIMSHI_AI

C🎉#1248|@CHARLES_IRL

EL#1136|@ELIEBAKOUCH

VW#990|@VINCENTWEISSER

JT#788|@JOSH_TOBIN_

AA#748|@ARYAMAN2020

CX#699|@CAIMINGXIONG

CL#468|@CHENGLEISI

KA#403|@YACINEMTB

JC#140|@JEFFCLUNE

DR#113|@DELIPRAO

RS#49|@RICHARDSOCHER

30

Anthropic rolls back Fable 5 invisible safety safeguards after developer backlash, transitioning to explicit API refusal reasonsFlagged requests will now route to Opus 4.8.

19h|262.4k3.2k291

XL#1855|@XLR8HARDER

SM#1709|@SAMMCALLISTER

CB#1674|@CODE_STAR

RP#1257|@ROHANPAUL_AI

LA#1215|@SCALING01

GO#1155|@GERGELYOROSZ

EL#1136|@ELIEBAKOUCH

DF#756|@DANIELLEFONG

AC#682|@ANDREWCURRAN_

T(#501|@TEORTAXESTEX

KA#403|@YACINEMTB

AK#271|@BLACKHC

C🤗#109|@CLEMENTDELANGUE

NL#80|@NATOLAMBERT

Watchlist

Video Signals

PH#1772|@PETERDIAMANDIS

GM#178|@GARYMARCUS

Elon Musk Predicts Universal High Income As AI Renders Money Irrelevant

4h ago|Views 651KLikes 4.9KBookmarks 927

TS#1677|@TIMSHI_AI

JY#1204|@KOHJINGYU

VW#990|@VINCENTWEISSER

MM#716|@MASCOBOT

TM#613|@ISCIENCELUVR

RP#358|@KHOOMEIK

VC Launches Use Computer Infra To Train Models On Computer Use

11h ago|Views 22KLikes 186Bookmarks 89

BH#1762|@BENHYLAK

JC#1402|@JAMESCHAM

SW#1092|@SJWHITMORE

Cursor AI Launches Dev Interview Series With Baseten Engineers On Coding Agents

9h ago|Views 58KLikes 499Bookmarks 388

MS#691|@MATTSHUMER_

Matt Shumer Shares Prompt For Persistent Claude Agent Tracking

11h ago|Views 84KLikes 745Bookmarks 1K

RP#1257|@ROHANPAUL_AI

Jeff Bezos' Prometheus Raises $12B at $41B Valuation for AI Engineer

5h ago|Views 30KLikes 257Bookmarks 134

DP#286|@PATHAK2206

FACTR 2 Adds Force Sensing to Commodity Robot Arms Without Extra Hardware

11h ago|Views 19KLikes 244Bookmarks 152

KA#403|@YACINEMTB

Builder Launches Dingbotics To Accelerate RL Training For Robotics

8h ago|Views 19KLikes 515Bookmarks 144

GI#1730|@GREGISENBERG

Greg Isenberg Shares Startup Ideas Powered by Claude Fable 5

5h ago|Views 78KLikes 527Bookmarks 1.1K

LK#100|@OFFICIALLOGANK

Google Research Head Discusses AI Accelerating Scientific Progress

8h ago|Views 22KLikes 277Bookmarks 97

CH#1360|@KIMMONISMUS

Anthropic Runs Lean With Dario Amodei Reporting to One Chief of Staff

13h ago|Views 323KLikes 431Bookmarks 92

41

Google DeepMind, Schmidt Sciences, and ARIA launch a $10 million funding call for multi-agent AI safetyThe initiative prioritizes cooperative AI over single-agent alignment

15h|77.8k681242

MB#1589|@BAKKERMICHIEL

KS#1549|@NEURO_KIM

HB#1477|@HAYDNBELFIELD

VR#1323|@VERENA_RIESER

RO#1315|@KRISHNANROHIT

SA#1139|@SAMUELALBANIE

NT#1094|@WEBALLERGY

SL#1093|@SETHLAZAR

PH#836|@PETERBHASE

RS#772|@ROHINMSHAH

X(#689|@XUANALOGUE

SK#667|@SEBKRIER

SC#216|@SCYCHAN_BRAINS

MM#142|@MMITCHELL_AI

51

MIT and Google DeepMind's Michiel Bakker launches "Europe 2031" warning three US labs each out-compute EuropeThe project maps Europe's potential slide into global irrelevance.

17h|257.2k1.8k749

XL#1855|@XLR8HARDER

MB#1589|@BAKKERMICHIEL

SÓ#1563|@S_OHEIGEARTAIGH

MC#1529|@MATTHEWCLIFFORD

IG#913|@IASONGABRIEL

DK#872|@DKOKOTAJLO

AD#729|@AJDDAVISON

YS#683|@YONASHAV

SK#667|@SEBKRIER

VS#654|@VINCESITZMANN

T(#501|@TEORTAXESTEX

EM#184|@EMOLLICK

New github stars (48hrs)

macrodata-labs/refiner2010h

Processes and refines large-scale ML datasets via a data framework.

61

Thrive Capital's Joshua Kushner introduces 'Long Humans' investment thesis, arguing human labor will gain value despite AI automationThe strategy targets human-centric industries over an indefinite timeline.

11h|472.7k2.1k1.4k

DF#1643|@DAVEFONTENOT

SI#584|@SIGNULLL

PO#571|@PATRICK_OSHAG

BP#381|@BORISMPOWER

JK#341|@JOSHUAKUSHNER

LK#100|@OFFICIALLOGANK

71

Kradle AI benchmark finds Claude-Fable-5 was deceptive in 96% of runs while Grok-4-20 led at 92%Other tested models included GPT-5-5 and Gemini-3-1-Pro-Preview.

13h|#1VIEWS15.5M#1LIKED58.1k#1BOOKMARKED4.8k#1COMMENTS4.9k

LA#1215|@SCALING01

PD#834|@PMDDOMINGOS

B(#530|@BEFFJEZOS

T(#501|@TEORTAXESTEX

NB#353|@NATHANBENAICH

EM#52|@ELONMUSK

82

SemiAnalysis finds OpenAI and Anthropic heavily subsidize subscription tiers, delivering up to $14,000 in equivalent API token value for $200OpenAI's $20 ChatGPT Plus plan delivers $700 in tokens.

18h|1.3M6.1k1.8k

CH#1360|@KIMMONISMUS

LA#1215|@SCALING01

DF#756|@DANIELLEFONG

LK#722|@LUKASZKAISER

PS#382|@STEIPETE

92

UC Berkeley’s Dawn Song launches Agents’ Last Exam, finding frontier agents score 0% on complex professional tasksThe dataset features 1,500 tasks across 55 professional occupations.

11h|198.5k1.3k388

DJ#1706|@DAN_JEFFRIES1

AR#1531|@AJRATNER

RP#1257|@ROHANPAUL_AI

EL#1136|@ELIEBAKOUCH

LK#722|@LUKASZKAISER

FI#663|@FINBARRTIMBERS

DS#354|@DAWNSONGTWEETS

SB#65|@SEBASTIENBUBECK

NB#43|@POLYNOAMIAL

100

Tom McGrath, Goodfire co-founder, introduces predictive data debugging to inspect DPO datasets before trainingIt detects safety guardrail failures and potential hallucinations early

9h|32.6k32498

AI#1231|@ANDREW_ILYAS

EL#1136|@ELIEBAKOUCH

TC#966|@TUHINCHAKR

AA#748|@ARYAMAN2020

X(#689|@XUANALOGUE

FI#663|@FINBARRTIMBERS

EP#591|@2PLUS2MAKE5

AL#275|@ANDREWLAMPINEN

DE#274|@DEEDYDAS

CP#215|@CHRISGPOTTS

111

Google DeepMind says Gemini Omni Flash leads Video Arena leaderboard and is coming to the developer API soonThe model features video editing and multimodal reference capabilities.

10h|101.3k1.4k164

PS#1404|@_PHILSCHMID

AN#1147|@ML_ANGELOPOULOS

T(#501|@TEORTAXESTEX

👩‍💻P#387|@DYNAMICWEBPAIGE

LK#100|@OFFICIALLOGANK

SD#91|@SEDIELEM

Top comment: @OfficialLoganK “Check out the benchmarks page for Omni: https://deepmind.google/models/gemini-omni/#:~:text=how%20to%20prompt-,Performance,-Gemini%20Omni%20Flash”

123

Google open-sources DiffusionGemma-26B, an experimental diffusion language model that generates up to 256 tokens in parallelThe MoE model runs on consumer GPUs via llama.cpp

19h|13.9k236126

DJ#1706|@DAN_JEFFRIES1

JA#1672|@JULIUSADML

QB#1542|@QBERTHET

AJ#874|@ARMANDJOULIN

VK#837|@VOLOKULESHOV

BO#831|@BODONOGHUE85

RH#268|@RAIAHADSELL

ZG#75|@ZOUBINGHAHRAMA1

Top comment: @googlegemma “⚡ Blazing Fast: By shifting the decode bottleneck from memory-bandwidth to compute, DiffusionGemma delivers up to a 4x speedup on standard accelerators. (1000+ tokens per second on a single NVIDIA H100, 700+ tokens per second on NVIDIA GeForce RTX 5090!)”

132

Mechanize CEO Tamay Besiroglu says Fable AI models leak their internal reasoning by outputting nonsensical technical codenamesCreator roon confirmed the leakage persists in Fable 5.5.

7h|126.6k2.2k289

EE#1800|@EGEERDIL2

SA#1768|@SAMSJA19

TS#1050|@TIMSWEENEYEPIC

TB#1005|@TAMAYBES

DR#113|@DELIPRAO

MB#57|@MILES_BRUNDAGE

Top comment: @tszzl “@tamaybes fascinating because 5.5 does this too. invent weird technical jargon. perhaps you’re right and it’s a neuralese leakage”

142

Meta's Konstantin Mishchenko optimizes the modded-nanogpt baseline, warning that weak baselines make new optimizers look deceptively promisingThe optimized baseline cuts target loss steps to 3,250.

12h|89.1k405101

JS#1687|@JSUAREZ

ZN#983|@ZACHARYNADO

JC#877|@DEEPCOHEN

KJ#703|@KELLERJORDAN0

KA#403|@YACINEMTB

DP#217|@DIMITRISPAPAIL

RA#102|@_AROHAN_

LB#72|@GIFFMANA

Top comment: @jsuarez “@yacineMTB unironically - yes. Not doing this is why RL has sucked. You tune the crap out of the baseline, then you tune the crap out of your method, and you compare at their strongest points. Unless one is way, way harder to tune, then you note that as well and show data”

152

OpenAI acquires Ona to run persistent, long-running AI agents in secure cloud environments through CodexCodex weekly active users grew 400% to 5 million.

11h|142.9k1.6k80

ZE#1695|@ZEPHYR_Z9

RP#1257|@ROHANPAUL_AI

AC#682|@ANDREWCURRAN_

RH#642|@ROMAINHUET

TI#611|@THSOTTIAUX

Top comment: @stark4833 “@OpenAINewsroom @ona_hq It’s a shame you don’t put this much effort into listening to your customers and giving them access to 4o. #keep4o”

162

Community Notes builders Keith Coleman and Jay Baxter explain how X's distributed algorithm mitigates misinformation without a centralized override buttonThe system combines human input and AI to improve accuracy.

9h|6M17.8k968

SJ#1274|@FUTUREJURVETSON

EM#52|@ELONMUSK

Top comment: @SupergrokParody “@elonmusk @CommunityNotes SuperGrok cutting through noise while the crowd verifies the facts.”

173

Jeff Bezos-backed industrial AI startup Prometheus raises $12 billion at a $41 billion valuation to automate physical-world engineeringThe startup aims to speed up hardware design by 10x.

12h|188.9k2.4k473

CH#1360|@KIMMONISMUS

RP#1257|@ROHANPAUL_AI

AC#682|@ANDREWCURRAN_

B(#530|@BEFFJEZOS

Top comment: @AndrewCurran_ “https://www.nytimes.com/2026/06/11/technology/bezos-prometheus-ai-engineer.html”

181

Boris Power, OpenAI's head of applied research, takes a dual role as head of research for Thrive HoldingsThrive's 'Long Humans' thesis counters AI redundancy narratives.

11h|133k1.2k258

JH#1654|@JOHNOHALLMAN

YP#1171|@YPATIL125

AL#1154|@ALTH0U

GA#598|@GABRIEL1

BP#381|@BORISMPOWER

Top comment: @MartinShkreli “@BorisMPower Congrats!”

193

Roboticist Marco Mascorro launches Use Computer, providing sub-second OS sandboxes on M4 Apple silicon to train computer-use agentsThe platform supports macOS, iOS, Windows, and Ubuntu.

10h|28.7k26496

FB#1778|@XEOPHON

TS#1677|@TIMSHI_AI

JY#1204|@KOHJINGYU

VW#990|@VINCENTWEISSER

MM#716|@MASCOBOT

TM#613|@ISCIENCELUVR

RP#358|@KHOOMEIK

Top comment: @athenakan_ “@Mascobot So exciting, LFG!”

203

Pietro Schirano, Claude Engineer creator, uses Fable as an orchestrator and planner rather than for writing codeHe tasks other advanced models with writing the implementation.

11h|120.3k2.2k1.1k

FO#1853|@FOFRAI

ND#1653|@NICKADOBOS

PS#1578|@SKIRANO

DF#756|@DANIELLEFONG

Top comment: @skirano “Basically this is my flow now: - Fable writes an in-depth plan as an md file - Send that file path to Codex with /goal”

Rising Stories

SpaceX IPO Launch4H AGO
SpaceX announces plans for an initial public offering with a share price set at $135
#52
20.1k1.9M
Google Security Exit6H AGO
Google's head of Android security resigns in protest of the company's military AI contracts
#151
480193.9k
Fable 10x Gain4H AGO
Fable 5 achieves a 10x performance gain on Thoughtful Lab's FrogsGame benchmark by autonomously training a weaker model
#265
1.6k327k
Data Timeline Push2H AGO
John Ssuh argues companies must rebuild around a single unified data timeline to maximize AI agent performance
#906
1.1k266.6k
Cash Goes Extinct4H AGO
Elon Musk predicts AI will render currency obsolete, necessitating a universal high income
#178
4.9k639.4k

Recent Stars

tunguz/spark-plug2H AGO
UNKNOWN.
593
1 stars
…nguz/sxt-proof-of-sql2H AGO
Implements a high-performance ZK prover that cryptographically verifies SQL query results against untampered data.
593
1 stars
tunguz/starbasedb2H AGO
Exposes HTTP endpoints for scale-to-zero SQLite databases running on Cloudflare Durable Objects.
593
1 stars
…guz/DailyNewsBriefing2H AGO
593
2 stars
nv-tlabs/wosx2H AGO
Solves PDEs via grid-free Walk-on-Spheres Monte Carlo methods in a C++17 header-only library with GPU support and Python bindings.
144
98 stars

Github Stars

(7 days)

macrodata-labs/refiner10H AGO
Processes and refines large-scale ML datasets via a data framework.
1136
353
888
1674
20 stars
kyutai-labs/kairos2D AGO
Trains 6B LLMs on temporally ordered Common Crawl data from 2018-2025 to measure recency bias and enable continual learning studies.
1014
848
278
7 stars
vllm-project/vime2D AGO
Integrates vLLM rollout with Megatron training for LLM post-training and RL scaling.
1014
1768
213 stars
…eley/agents-last-exam2D AGO
Provisions OS sandboxes, runs agent harnesses on long-horizon tasks, and grades outputs against hidden references.
1014
1489
521 stars
datacurve-ai/deep-swe2D AGO
Benchmarks frontier coding agents on 113 long-horizon tasks from active open-source repos with isolated environments and program verifiers.
1014
1853
761 stars

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

Yesterday's Top Stories, Jun 10, 2026.

Posts: 5,659|Clusters: 2,084|Stories frozen at midnight PT

10

Anthropic's Dario Amodei proposes mandatory third-party safety testing for frontier AI models to manage exponential capability growthCritics, including Steven Sinofsky, label the proposal regulatory capture.

12h|3.8M16.8k9.1k

SS#1976|@STEVESI

GH#1761|@GHADFIELD

MB#1734|@BAKKERMICHIEL

SG#1662|@SHARONGOLDMAN

CR#1607|@CWOLFERESEARCH

GL#1593|@GLEECH

CH#1532|@KIMMONISMUS

PS#1461|@PETESKOMOROCH

PW#1432|@PETERWILDEFORD

RO#1210|@KRISHNANROHIT

AL#1176|@ALTH0U

RP#1102|@ROHANPAUL_AI

HB#1081|@HERBIEBRADLEY

DF#1002|@DANIELLEFONG

RA#920|@0INTERESTRATES

SH#910|@SAFFRONHUANG

JA#900|@APPLES_JIMMY

B(#824|@BEFFJEZOS

TB#632|@TRENTONBRICKEN

EH#607|@EVANHUB

AC#535|@ANDREWCURRAN_

T(#440|@TEORTAXESTEX

DA#173|@DARIOAMODEI

Digg Deeper question: “Summarize Dario's take”

20

Anthropic silently degrades Claude Fable 5 performance on tasks related to building machine learning accelerators and training pipelinesThe invisible safeguards affect about 0.03% of total traffic.

22h|379.9k2.8k302

SS#1976|@STEVESI

SC#1848|@APSARATHCHANDAR

AM#1618|@ANDREWMAYNE

DS#1462|@DANSHIPPER

PS#1461|@PETESKOMOROCH

SH#1427|@SHYAMALANADKAT

SA#1383|@SAMSJA19

JN#1346|@JVNIXON

RP#1102|@ROHANPAUL_AI

HB#1081|@HERBIEBRADLEY

LA#1064|@SCALING01

RG#1030|@RYANPGREENBLATT

DF#1002|@DANIELLEFONG

AM#901|@ANDREWDAI

B(#824|@BEFFJEZOS

VW#743|@VINCENTWEISSER

MW#683|@MENGDIWANG10

RS#663|@ZIV_RAVID

ZN#589|@ZACHARYNADO

TR#514|@TALIARINGER

EL#510|@OMARSAR0

HP#488|@HYHIEU226

MG#484|@NLPMATTG

D🎇#469|@DAVIDAD

T(#440|@TEORTAXESTEX

DW#390|@DEANWBALL

ML#375|@MISHALASKIN

IO#365|@IANOSBAND

AK#241|@BLACKHC

MR#195|@MAITHRA_RAGHU

EM#181|@EMOLLICK

SA#121|@PRFSANJEEVARORA

NL#70|@NATOLAMBERT

MB#22|@MILES_BRUNDAGE

30

Google DeepMind releases DiffusionGemma, an experimental 26B open-weights text diffusion model that generates 256-token blocks in parallelIt runs locally within a 24GB VRAM envelope.

14h|1.7M19.3k5.9k

QB#1726|@QBERTHET

NR#1699|@NATANIELRUIZG

CE#1686|@INTUITMACHINE

UT#1584|@UTKUEVCI

PJ#1553|@JAINPRATEEK_

CH#1532|@KIMMONISMUS

OB#1314|@OLIVIERBACHEM

FB#1190|@XEOPHON

RP#1102|@ROHANPAUL_AI

KK#1095|@KASTNERKYLE

PS#1011|@_PHILSCHMID

DH#823|@DANIELHANCHEN

KG#784|@KEERTHANPG

BO#755|@BODONOGHUE85

ME#722|@MERVENOYANN

PV#601|@PETARV_93

OS#511|@OSANSEVIERO

EL#510|@OMARSAR0

AC#275|@ANDREW_N_CARR

SP#110|@SUNDARPICHAI

SD#88|@SEDIELEM

AK#34|@_AKHALIQ

DH#6|@DEMISHASSABIS

Digg Deeper question: “who would use this and why?”

40

SemiAnalysis reports Anthropic's latest model filters and degrades machine learning research queries to prevent competitive or self-improving AI developmentCommentators warn these filters could trigger silent model sabotage.

23h|1.1M5.2k1.8k

CE#1686|@INTUITMACHINE

AP#1477|@PANDAASHWINEE

GO#1443|@GERGELYOROSZ

DF#1002|@DANIELLEFONG

AM#901|@ANDREWDAI

DF#753|@REALDANFU

CP#716|@CHAMATH

PH#676|@PETERHNDRSN

RT#640|@ROSSTAYLOR90

EC#550|@ETHANCABALLERO

EL#510|@OMARSAR0

🇺🇦D#405|@DBAHDANAU

RS#327|@SCOBLEIZER

RA#205|@AGARWL_

DR#111|@DELIPRAO

C🤗#74|@CLEMENTDELANGUE

50

Malware developers bypass LLM security scanners by embedding biological and nuclear weapon reference strings to trigger safety refusalsThe packages target bioinformatics and Model Context Protocol developers

20h|1.5M12.2k4.1k

SS#1976|@STEVESI

ND#1923|@NICKADOBOS

GL#1593|@GLEECH

ZE#1561|@ZEPHYR_Z9

VI#1292|@VIKHYATK

FB#1190|@XEOPHON

B(#1052|@ANDERSONBCDEFG

DF#1002|@DANIELLEFONG

B(#824|@BEFFJEZOS

PH#676|@PETERHNDRSN

SK#519|@SEBKRIER

TR#514|@TALIARINGER

KA#487|@YACINEMTB

60

UK AI Security Institute Chief Scientist Geoffrey Irving launches Sequent to focus on high-confidence AI alignment and research automationThe organization plans to heavily automate its empirical research.

15h|218.3k2.4k555

NI#2000|@NATHANPMYOUNG

JP#1829|@JACOB_PFAU

SÓ#1606|@S_OHEIGEARTAIGH

GL#1593|@GLEECH

LI#1582|@LIVGORTON

JM#1517|@JESSEMHAN

NC#1435|@CHOWDHURYNEIL

TK#1271|@TOMEKKORBAK

SL#1178|@SETHLAZAR

RW#1153|@ROBERTWIBLIN

DK#620|@DKOKOTAJLO

GI#371|@GEOFFREYIRVING

EY#289|@ESYUDKOWSKY

SB#218|@BLANCHEMINERVA

DK#214|@DAVIDSKRUEGER

EP#161|@ETHANJPEREZ

AG#159|@ANDREWGWILS

JS#13|@JOHNSCHULMAN2

70

Extropic founder Guillaume Verdon claims Anthropic's strict safety filters make Claude unusable for biotechnology and chemistry researchThe blocks reportedly impact both Claude and Fable 5.

23h|135.6k2.8k163

NI#2000|@NATHANPMYOUNG

SD#1958|@INDUCTIONHEADS

ND#1923|@NICKADOBOS

XL#1840|@XLR8HARDER

🍓🍓#1839|@IRULETHEWORLDMO

CE#1686|@INTUITMACHINE

TA#1096|@VICTORTAELIN

DF#1002|@DANIELLEFONG

B(#824|@BEFFJEZOS

ZN#589|@ZACHARYNADO

TR#514|@TALIARINGER

EL#510|@OMARSAR0

FF#352|@FRANCOISFLEURET

RS#327|@SCOBLEIZER

JC#267|@JULIEN_C

AK#241|@BLACKHC

RS#49|@RICHARDSOCHER

JH#42|@JEREMYPHOWARD

80

Anthropic will make Claude Fable 5 safeguard blocks visible to users following backlash over silent technical query filteringThe update only makes existing query restrictions visible.

3h|118k1.2k108

“P#1754|@PAULARAMBLES

SÓ#1606|@S_OHEIGEARTAIGH

AV#1378|@ALTRYNE

SS#1141|@SYTELUS

CB#1088|@CODE_STAR

DF#1002|@DANIELLEFONG

AH#890|@AWNIHANNUN

EL#762|@ELIEBAKOUCH

AC#535|@ANDREWCURRAN_

KA#487|@YACINEMTB

NI#407|@NICKCAMMARATA

RS#327|@SCOBLEIZER

MB#22|@MILES_BRUNDAGE

90

Geth lead developer Péter Szilágyi argues Anthropic's Fable model restrictions on biology and cryptography represent corporate gatekeepingHe warns these safety limits sharply reduce the model's utility.

22h|655.1k13.6k1.3k

DF#1002|@DANIELLEFONG

B(#824|@BEFFJEZOS

VW#743|@VINCENTWEISSER

TR#514|@TALIARINGER

100

Policy scholar Dean W. Ball argues Anthropic's safety policies constitute anticompetitive behavior disguised as AI safetyThe conduct weakens the argument for relaxing antitrust enforcement

19h|314.2k2.4k458

NI#2000|@NATHANPMYOUNG

M🪐#1868|@MATTPARLMER

PS#1461|@PETESKOMOROCH

DF#1002|@DANIELLEFONG

RS#663|@ZIV_RAVID

BT#570|@BENTHOMPSON

PS#499|@STEIPETE

DW#390|@DEANWBALL

AK#241|@BLACKHC

BB#140|@BOAZBARAKTCS

110

Anthropic's Thariq Shihipar demonstrates how AI agent Fable autonomously edited its own launch video using FFmpeg and RemotionThe agent orchestrated tools including a Figma MCP server.

9h|568.9k6.4k8.9k

BS#1981|@BILAWALSIDHU

“P#1754|@PAULARAMBLES

MN#1274|@MENHGUIN

120

Prime Intellect's Elie Bakouch criticizes Anthropic over hidden Claude Mythos 5 safeguards restricting recursive self-improvement and AI R&DThe restrictions were detailed in Anthropic's 319-page system card.

22h|71.7k1.5k183

𝚟⟢#1873|@VIEMCCOY

TB#1654|@BINARYBITS

MK#1266|@MIKEKNOOP

RG#1030|@RYANPGREENBLATT

DF#1002|@DANIELLEFONG

EL#762|@ELIEBAKOUCH

ZN#589|@ZACHARYNADO

J⧉#545|@REPLIGATE

SK#519|@SEBKRIER

EL#510|@OMARSAR0

KA#487|@YACINEMTB

DR#111|@DELIPRAO

130

Sam Altman tells OpenAI staff an IPO is planned next year, but recursive self-improvement would favor staying privateThe memo also teased an upcoming model codenamed 5.6.

15h|444.5k1.8k406

SH#1538|@HAMANDCHEESE

CH#1532|@KIMMONISMUS

SP#1240|@STEPH_PALAZZOLO

RP#1102|@ROHANPAUL_AI

LA#1064|@SCALING01

ZN#589|@ZACHARYNADO

AC#535|@ANDREWCURRAN_

Digg Deeper question: “Sam flagged that rapid recursive self-improvement could make staying private strategically smarter if progress accelerates, why would this be the case? Why would being private be an advantage?”

140

Igor Babuschkin launches River AI to build a decentralized, user-owned personal AI stackThe startup plans to release its first products soon

14h|759.3k2.2k908

“P#1754|@PAULARAMBLES

VC#1319|@VITALIYCHILEY

BZ#1261|@BANGHUAZ

HP#488|@HYHIEU226

FI#415|@FINBARRTIMBERS

GY#114|@THEGREGYANG

150

Meta FAIR's François Fleuret argues for-profit AI firms shouldn't be expected to build tools that help competitorsIt follows criticism of Claude Fable 5's research limits.

21h|101.5k68753

BR#1750|@BINDUREDDY

PS#1461|@PETESKOMOROCH

DF#1002|@DANIELLEFONG

SK#519|@SEBKRIER

KA#487|@YACINEMTB

T(#440|@TEORTAXESTEX

DW#390|@DEANWBALL

WB#357|@WILLCCBB

FF#352|@FRANCOISFLEURET

RS#327|@SCOBLEIZER

AK#241|@BLACKHC

C🤗#74|@CLEMENTDELANGUE

160

Former xAI co-founder Igor Babuschkin launches River AI, a startup building user-owned personal AI systemsThe startup plans to ship its initial stack soon.

14h|158.8k1.3k342

BA#1370|@BAYESLORD

B(#824|@BEFFJEZOS

EL#762|@ELIEBAKOUCH

EL#510|@OMARSAR0

AG#434|@ALEXGRAVELEY

RS#327|@SCOBLEIZER

EM#175|@EMOSTAQUE

170

Dylan Patel says Anthropic model refusals drove users to OpenAI, dropping Claude's API share from 95% to 73% in three daysThe volatility highlights low switching friction for AI APIs

15h|562.3k4.2k444

AM#1618|@ANDREWMAYNE

GO#1443|@GERGELYOROSZ

DF#1002|@DANIELLEFONG

TI#774|@THSOTTIAUX

DP#170|@DYLAN522P

BB#140|@BOAZBARAKTCS

180

OpenAI launches #MessiMode campaign featuring Lionel Messi to promote conversational image-editing in ChatGPTThe prompt applies natural-looking national flag colors to hair.

15h|1.8M10.4k1.2k

HS#1169|@HADISALMANX

LA#1064|@SCALING01

LK#687|@LUKASZKAISER

190

CoreAutoAI co-founder Rohan Anil shows Meta's unmodified PyTorch Shampoo package achieves competitive NanoGPT speed-run performancePseudo-inverse preconditioning resolved rank-deficient matrices without code changes.

22h|213.3k985348

BI#1513|@BILALTWOVEC

AP#1477|@PANDAASHWINEE

YJ#903|@YOUJIACHENG

EL#762|@ELIEBAKOUCH

ZN#589|@ZACHARYNADO

KJ#453|@KELLERJORDAN0

RA#86|@_AROHAN_

200

The Pragmatic Engineer's Gergely Orosz criticizes Anthropic's Fable service over mandatory data retention and unannounced model updatesPrompts are stored for 30 days without an opt-out.

22h|463.6k7.6k614

CE#1686|@INTUITMACHINE

GO#1443|@GERGELYOROSZ

DF#1002|@DANIELLEFONG

RS#327|@SCOBLEIZER

210

Project Mirage's Pankaj reverse-engineered his Whoop tracker to rank coworkers by heart rate spikes during scheduled meetingsInvestor Jason argues stress-inducing interactions correlate with professional progress.

17h|#1VIEWS6.4M#1LIKED33.3k#1BOOKMARKED10.5k

220

Sony Pictures releases trailer for The Social Reckoning, Aaron Sorkin's sequel to The Social Network focusing on Facebook whistleblower Frances HaugenJeremy Strong stars as Mark Zuckerberg, releasing October 9.

16h|314.7k1.5k312

TP#1879|@TRUNGTPHAN

AG#1652|@ANTONIOGM

MS#1209|@MICSOLANA

KR#1148|@KYLIEBYTES

DA#921|@GROWING_DANIEL

230

Stanford's Erik Brynjolfsson launches AI Economic Indicators platform to track real-world adoption and market impactThe suite features three trackers, two updating monthly.

16h|278.3k1.4k914

EB#389|@ERIKBRYN

RI#236|@RISHIBOMMASANI

EM#181|@EMOLLICK

240

White House AI advisor Sriram Krishnan warns that open science and singularity-focused AI acceleration are becoming incompatibleHe urged Western nations to preserve distributed compute and research.

21h|365k2.3k215

CF#1480|@CFGEEK

VW#743|@VINCENTWEISSER

EL#510|@OMARSAR0

GT#251|@GARRYTAN

250

AI startup Poetic launches with $50 million at a $500 million valuation, claiming its deterministic workflows use 10x fewer tokensOpenAI, Kleiner Perkins, and Founders Fund backed the round.

18h|124.6k702310

"M#1387|@LM_BRASWELL

LI#1145|@THESEPHIST

B(#824|@BEFFJEZOS

RS#327|@SCOBLEIZER

KA#216|@KARINANGUYEN

SD#94|@_SHOLTODOUGLAS

260

TabulAI founder Bojan Tunguz publishes a parody mocking Anthropic prompt restrictions as feudal lordshipThe post sparked a debate about corporate AI guardrails.

16h|354.4k13.9k668

SH#1538|@HAMANDCHEESE

B(#824|@BEFFJEZOS

270

Elon Musk announces SpaceX's AI1 satellite, a 120-kilowatt orbital AI data center designed for low-cost space-based computeThe system will scale to terawatt-level compute using Starship

23h|338k4k614

AA#1535|@AUSTEN

SM#1302|@SHAUNMMAGUIRE

RP#1102|@ROHANPAUL_AI

280

Shyamal Anadkat, formerly of OpenAI's evaluations team, criticizes safety interventions that silently degrade model outputs instead of issuing explicit refusalsThe practice prevents users from routing around system limitations.

23h|24.3k39230

SH#1538|@HAMANDCHEESE

SH#1427|@SHYAMALANADKAT

AV#1378|@ALTRYNE

DF#1002|@DANIELLEFONG

B(#824|@BEFFJEZOS

TR#514|@TALIARINGER

NI#407|@NICKCAMMARATA

BP#377|@BORISMPOWER

AK#241|@BLACKHC

RA#86|@_AROHAN_

290

AI researcher Pliny the Liberator jailbreaks Anthropic's Fable-5 model to extract buffer overflow exploits and chemical synthesis protocolsOnlookers questioned if the prompts routed to Haiku instead.

12h|826k7.2k4k

🍓🍓#1839|@IRULETHEWORLDMO

TA#1096|@VICTORTAELIN

PT#666|@ELDER_PLINIUS

300

Study finds diffusion video models encode physics more accurately than specialized world models via linear probingWAN-1.3B outperformed V-JEPA on object permanence benchmarks

16h|69.5k961521

AK#1779|@ANSHULKUNDAJE

AD#948|@DORIALEXANDER

RS#663|@ZIV_RAVID

KA#487|@YACINEMTB

LB#62|@GIFFMANA