Tech News

Today's Highlights

In case you missed it

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

EM#171|@EMOSTAQUE

JL#111|@JASONDEANLEE

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

Top Stories

Posts:—|Clusters:—|Next Crawl:—

10

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

2h|9k603

ZE#1695|@ZEPHYR_Z9

KA#1213|@KALOMAZE

EL#1136|@ELIEBAKOUCH

T(#501|@TEORTAXESTEX

Top comment: @zephyr_z9 “@eliebakouch @teortaxesTex @antirez @_xjdr @norpadon Mostly the performance degradation”

20

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.

58m|24.7k79424

B(#530|@BEFFJEZOS

Top post: @beffjezos “Analogizing AI models with nukes had backfired. It was always a dumb analogy that was going to have consequences”

30

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.

58m|11.5k5117

ET#816|@ERICTOPOL

TM#613|@ISCIENCELUVR

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

Watchlist

Video Signals

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

2h ago|Views 21KLikes 446Bookmarks 71

ME#861|@MERVENOYANN

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

2h ago|Views 17KLikes 0Bookmarks 0

DE#274|@DEEDYDAS

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

2h ago|Views 3.4KLikes 10Bookmarks 3

RV|@RONALD_VANLOON

New Architecture Adds Trust Layer to AI Agents for Safer Execution

54m ago|Views 1.2KLikes 4Bookmarks 0

RV|@RONALD_VANLOON

Humanoid Robots Learn to Climb and Vault Obstacles Using AI

1h ago|Views 762Likes 1Bookmarks 0

40

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

2h|4.5k3410

RP#1257|@ROHANPAUL_AI

WI#1049|@WSISAAC

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”

50

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.

49m|25k952

ZE#1695|@ZEPHYR_Z9

T(#501|@TEORTAXESTEX

Top post: @zephyr_z9 “YES”

60

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.

25m|2.8k326

FB#1778|@XEOPHON

T(#501|@TEORTAXESTEX

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

Recent Stars

deepseek-ai/DeepSpec3H AGO
Provides full-stack code for training/evaluating speculative decoding draft models with data prep and eval scripts.
1136
55 stars
…/terrain-diffusion-mc4H AGO
Generates Minecraft worlds with diffusion-model procedural terrain as a Fabric mod.
1014
580 stars
…gos/terrain-diffusion4H AGO
Generates infinite deterministic real-time terrain via diffusion models as a Perlin noise successor.
1014
621 stars
THUDM/slime7H AGO
Integrates Megatron training with SGLang rollouts for LLM post-training and RL scaling.
458
1014
484
6.8k stars
…You/AutoResearchBench7H AGO
Implements inference scripts, search tools, and evaluation pipelines for the AutoResearchBench dataset of deep and wide scientific literature tasks.
1590
45 stars

Github Stars

(7 days)

THUDM/slime7H AGO
Integrates Megatron training with SGLang rollouts for LLM post-training and RL scaling.
458
1014
484
6.8k stars
xlang-ai/OSWorld-V216H AGO
Benchmarks computer-use agents on long-horizon real-world desktop tasks via VM environments and gated task sets.
397
843
16 stars
…nd-models-of-ai-tools1D AGO
Aggregates leaked and open-sourced system prompts plus tool definitions from AI coding agents like Cursor, Claude, and Devin.
1014
827
141.2k stars
anthropics/skills1D AGO
Hosts collections of self-contained skills, specs, and templates for dynamically extending Claude's capabilities on specialized tasks.
1014
1467
155.5k stars
…entX/OpenCaptchaWorld1D AGO
Hosts a Flask-based web platform with 20 CAPTCHA types to benchmark MLLM agents on visual reasoning and interaction tasks.
1014
939
79 stars

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

Yesterday's Top Stories, Jun 26, 2026.

Posts: 4,492|Clusters: 1,471|Stories frozen at midnight PT

10

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

13h|#1VIEWS5.4M#1LIKED61.8k#1BOOKMARKED4.8k

BH#1762|@BENHYLAK

ZE#1695|@ZEPHYR_Z9

JH#1667|@JEFFREYHUBER

JH#1654|@JOHNOHALLMAN

CM#1436|@CHARLIERMARSH

AV#1419|@ALTRYNE

AM#1396|@ANDREWMAYNE

CP#1392|@CHETANP

EV#1387|@ERICVISHRIA

CH#1360|@KIMMONISMUS

RP#1257|@ROHANPAUL_AI

LA#1215|@SCALING01

TA#1140|@VICTORTAELIN

SH#1031|@STEVENHEIDEL

DR#981|@DANINTHEORY

SR#957|@CLONEOFSIMO

PT#821|@ELDER_PLINIUS

JM#792|@VENTURETWINS

DF#756|@DANIELLEFONG

RS#741|@ZIV_RAVID

MS#691|@MATTSHUMER_

EL#684|@OMARSAR0

AC#682|@ANDREWCURRAN_

AP#672|@THEANDIPENGUIN

RH#642|@ROMAINHUET

DA#619|@GROWING_DANIEL

TM#613|@ISCIENCELUVR

TI#611|@THSOTTIAUX

GA#598|@GABRIEL1

SI#584|@SIGNULLL

SM#545|@SANGMICHAELXIE

B(#530|@BEFFJEZOS

YB#494|@YUBAI01

GB#449|@GAVINSBAKER

KA#403|@YACINEMTB

BA#398|@BRIAN_ARMSTRONG

BP#381|@BORISMPOWER

YD#357|@YANNDUBS

BT#292|@BTAYLOR

KA#265|@KARINANGUYEN

S🔜#225|@SWYX

EM#171|@EMOSTAQUE

JL#111|@JASONDEANLEE

SN

Digg Deeper question: “how to access it?”

answerview answer

20

US Commerce Department lifts restrictions on Anthropic's Mythos 5 model for 100 domestic critical infrastructure organizationsAnthropic's Fable 5 model remains under global licensing restrictions.

16h|618k5k445

MB#1780|@MATTHEWBERMAN

CB#1428|@CHRISJBAKKE

CH#1360|@KIMMONISMUS

RP#1257|@ROHANPAUL_AI

LA#1215|@SCALING01

DF#756|@DANIELLEFONG

AC#682|@ANDREWCURRAN_

J⧉#674|@REPLIGATE

B(#530|@BEFFJEZOS

T🪽#331|@TEKNIUM

GM#178|@GARYMARCUS

GT#121|@GARRYTAN

MB#57|@MILES_BRUNDAGE

30

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.

1d|271.2k3k280

XL#1855|@XLR8HARDER

“P#1846|@PAULARAMBLES

AS#1809|@ALEXSTAMOS

LM#1737|@LUKEPROG

AD#1537|@DORIALEXANDER

CH#1360|@KIMMONISMUS

EN#1301|@ERICNEWCOMER

LA#1215|@SCALING01

🎭#1014|@DEEPFATES

DF#756|@DANIELLEFONG

RS#741|@ZIV_RAVID

B(#530|@BEFFJEZOS

T(#501|@TEORTAXESTEX

RS#447|@SCOBLEIZER

KA#403|@YACINEMTB

GM#178|@GARYMARCUS

40

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.

1d|149k994403

TB#1863|@BINARYBITS

AS#1809|@ALEXSTAMOS

SA#1714|@SJGADLER

DJ#1706|@DAN_JEFFRIES1

PS#1594|@RECURSEPARADOX

SÓ#1563|@S_OHEIGEARTAIGH

EN#1301|@ERICNEWCOMER

DF#756|@DANIELLEFONG

YS#683|@YONASHAV

NL#80|@NATOLAMBERT

MB#57|@MILES_BRUNDAGE

AD

Digg Deeper question: “What are the current second-order effects of what's happening right now with the labs seemingly being nationalized?”

answerview answer

50

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.

18h|394.8k5.2k509

“P#1846|@PAULARAMBLES

LA#1783|@LIORONAI

MB#1780|@MATTHEWBERMAN

PS#1594|@RECURSEPARADOX

MA#315|@MARTIN_CASADO

BG#259|@BGURLEY

GM#178|@GARYMARCUS

EM#52|@ELONMUSK

60

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.

1d|986.4k5.1k1.3k

DJ#1706|@DAN_JEFFRIES1

RP#1257|@ROHANPAUL_AI

BG#259|@BGURLEY

GM#178|@GARYMARCUS

C🤗#109|@CLEMENTDELANGUE

70

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.

13h|191.8k1.6k220

TK#1513|@TOMEKKORBAK

CP#1472|@CHRISPAINTERYUP

RP#1257|@ROHANPAUL_AI

LA#1215|@SCALING01

DR#953|@IDAVIDREIN

AE#946|@ADRIENLE

DF#756|@DANIELLEFONG

EL#684|@OMARSAR0

AC#682|@ANDREWCURRAN_

KA#403|@YACINEMTB

80

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.

13h|262.2k1.8k1.4k

GO#1155|@GERGELYOROSZ

FC#865|@ALTIMOR

HC#698|@HWCHASE17

BA#398|@BRIAN_ARMSTRONG

GM#178|@GARYMARCUS

C🤗#109|@CLEMENTDELANGUE

90

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

1d|176.1k1.4k121

DM#1766|@DAVEMCCLURE

NT#1699|@NITASHATIKU

MM#1452|@PHOTOMATT

EN#1301|@ERICNEWCOMER

DM#1198|@DAVEMORIN

JL#755|@JESSLIVINGSTON

JW#710|@WOLFEJOSH

MB#512|@BENIOFF

RS#447|@SCOBLEIZER

BT#238|@BENTHOMPSON

100

OpenAI plans a July launch for GPT-5.6 Sol on Cerebras hardware at 750 tokens per secondGPT-5.6 Sol Ultra scored 91.9% on TerminalBench 2.1.

13h|157.3k2k165

V(#1939|@REACH_VB

AG#1848|@ITSANDREWGAO

ZE#1695|@ZEPHYR_Z9

AM#1396|@ANDREWMAYNE

RP#1257|@ROHANPAUL_AI

🎭#1014|@DEEPFATES

LK#722|@LUKASZKAISER

YJ#586|@YUCHENJ_UW

B(#530|@BEFFJEZOS

110

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.

13h|231.8k3.1k174

WH#1824|@NREHIEW_

RO#1315|@KRISHNANROHIT

B(#530|@BEFFJEZOS

T(#501|@TEORTAXESTEX

KA#403|@YACINEMTB

NB#43|@POLYNOAMIAL

120

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.

11h|685.5k11.3k523

PS#1358|@PETESKOMOROCH

SM#1225|@PITDESI

PB#771|@PAULTOO

130

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.

15h|56.9k752196

FB#1778|@XEOPHON

CP#1472|@CHRISPAINTERYUP

AC#1278|@ARTHURCONMY

LA#1215|@SCALING01

DR#953|@IDAVIDREIN

DK#872|@DKOKOTAJLO

SK#667|@SEBKRIER

EM#184|@EMOLLICK

LB#72|@GIFFMANA

140

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.

15h|351.2k3.7k348

XL#1855|@XLR8HARDER

ZE#1695|@ZEPHYR_Z9

CH#1360|@KIMMONISMUS

AC#682|@ANDREWCURRAN_

150

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.

12h|143.7k1.1k651

MB#1780|@MATTHEWBERMAN

GK#1219|@GREGKAMRADT

GO#1155|@GERGELYOROSZ

LR#543|@LENNYSAN

DE#274|@DEEDYDAS

160

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.

15h|60.4k38589

AR#1531|@AJRATNER

LA#1215|@SCALING01

JY#1204|@KOHJINGYU

PQ#769|@QI2PENG2

YS#397|@YSU_NLP

XE#393|@XWANG_LK

WC#356|@WENHUCHEN

SR#336|@SIVAREDDYG

JL#326|@JUSTINLIN610

170

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

14h|143.8k741497

EV#1387|@ERICVISHRIA

YP#1171|@YPATIL125

"M#866|@LM_BRASWELL

MM#716|@MASCOBOT

LC#219|@LULUMESERVEY

SN#153|@SATYANADELLA

180

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.

21h|169.5k1.2k468

MH#1119|@MAXHODAK_

YI#1023|@YISHAN

ET#816|@ERICTOPOL

TM#613|@ISCIENCELUVR

JM#433|@JAYELMNOP

190

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.

14h|170.4k1.4k1k

ZE#1695|@ZEPHYR_Z9

MJ#1433|@MIMANSAJ

KE#1262|@EHSANIK

KD#904|@CSPROFKGD

SH#12|@SARAHOOKR

200

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.

1d|1.7M5.7k144

RP#1257|@ROHANPAUL_AI

EM#52|@ELONMUSK

210

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

13h|771.9k6.9k392

CA#1059|@CHRISALBON

TI#611|@THSOTTIAUX

220

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

20h|44.5k422118

KD#904|@CSPROFKGD

DF#756|@DANIELLEFONG

AC#682|@ANDREWCURRAN_

B(#530|@BEFFJEZOS

RS#447|@SCOBLEIZER

KA#403|@YACINEMTB

GT#121|@GARRYTAN

230

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.

1d|288.1k2.3k941

BR#1861|@BRIANROEMMELE

XL#1855|@XLR8HARDER

CH#1360|@KIMMONISMUS

240

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

9h|149.9k2.1k101

MB#1780|@MATTHEWBERMAN

AV#1419|@ALTRYNE

LA#1215|@SCALING01

B(#530|@BEFFJEZOS

250

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.

11h|286.8k11.8k377

MB#1780|@MATTHEWBERMAN

LT#1719|@_LEWTUN

SS#819|@STEVESI

260

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.

18h|152.8k1.6k192

RM#1770|@ROBERTSKMILES

SS#819|@STEVESI

PM#444|@PATIO11

270

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

1d|181.6k2k331

BH#1762|@BENHYLAK

🎭#1014|@DEEPFATES

DF#756|@DANIELLEFONG

J⧉#674|@REPLIGATE

GA#598|@GABRIEL1

280

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.

12h|185.1k1.3k392

HH#1682|@HAMELHUSAIN

🎭#1014|@DEEPFATES

290

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.

13h|68.4k37495

RO#1315|@KRISHNANROHIT

AE#946|@ADRIENLE

YS#683|@YONASHAV

B(#530|@BEFFJEZOS

JA#173|@JACHIAM0

300

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

20h|65.2k784170

CL#1076|@NPCOLLAPSE

AK#982|@ANDREY_KURENKOV

SK#667|@SEBKRIER