AI News

Today's Highlights

#1 last 24hrs

SpaceX S-1 filing reveals Google will pay $920 million monthly to lease 110,000 Nvidia GPUs

kacheKA#488|@YACINEMTB
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)T(#421|@TEORTAXESTEX
finbarrFI#395|@FINBARRTIMBERS
Nathan BenaichNB#242|@NATHANBENAICH
EmadEM#169|@EMOSTAQUE
+13 more
#1 BOOKMARKED
4.5k

Markus J. Buehler proposes a category-theory framework that lets AI scientists dynamically expand their reasoning schema

Robert ScobleRS#323|@SCOBLEIZER
Dan RoyDR#56|@ROYDANROY
RISING LIKES
424

Cornell PhD Jack Morris argues extreme vigilance is the top predictor of scientific success as AI agents complicate debugging

Nataniel RuizNR#1561|@NATANIELRUIZG
Jack MorrisJM#203|@JXMNOP

Top Stories

Posts:—|Clusters:—|Next Crawl:—
10

SpaceX S-1 filing reveals Google will pay $920 million monthly to lease 110,000 Nvidia GPUsSpaceX's AI cloud run rate now reaches $26 billion.

10h|#1VIEWS5.9M#1LIKED31.1k3.4k#1COMMENTS2.3k
xlr8harderXL#1671|@XLR8HARDER
Austen AllredAA#1559|@AUSTEN
ZephyrZE#1471|@ZEPHYR_Z9
whWH#1409|@NREHIEW_
bilalBI#1366|@BILALTWOVEC
Shaun MaguireSM#1342|@SHAUNMMAGUIRE
rohitRO#1214|@KRISHNANROHIT
signüllSI#1096|@SIGNULLL
Chris PainterCP#1062|@CHRISPAINTERYUP
Rohan PaulRP#1031|@ROHANPAUL_AI
Beff (e/acc)B(#850|@BEFFJEZOS
Chamath PalihapitiyaCP#725|@CHAMATH
Andrew CurranAC#518|@ANDREWCURRAN_
kacheKA#488|@YACINEMTB
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)T(#421|@TEORTAXESTEX
finbarrFI#395|@FINBARRTIMBERS
Nathan BenaichNB#242|@NATHANBENAICH
EmadEM#169|@EMOSTAQUE
20

Google DeepMind releases Gemma 4 Quantization-Aware Training checkpoints, cutting VRAM requirements by 72 percentUnsloth released GGUF versions for Ollama and llama.cpp

12h|527.2k5.6k2.5k
Chubby♨️CH#1448|@KIMMONISMUS
utkuUT#1430|@UTKUEVCI
Daniel HanDH#773|@DANIELHANCHEN
Ying ShengYS#608|@YING11231
Clément FarabetCF#509|@CLMT
Omar SansevieroOS#486|@OSANSEVIERO
👩‍💻 Paige Bailey👩‍💻P#270|@DYNAMICWEBPAIGE
Sheing NgSN1
30

XPRIZE founder Peter Diamandis argues universities charge up to $300,000 for degrees teaching skills now available via free LLMsHe calls this discrepancy a significant misallocation of capital.

14h|1.2M12k895
Beff (e/acc)B(#850|@BEFFJEZOS
Hieu PhamHP#459|@HYHIEU226
CLSCL#438|@CHENGLEISI
Gary MarcusGM#157|@GARYMARCUS
Elon MuskEM#74|@ELONMUSK
40

Senator Bernie Sanders proposes the US government acquire a 50 percent equity stake in major AI companiesGary Marcus warns state ownership risks establishing social credit.

11h|923.4k4.3k1.4k
Beff (e/acc)B(#850|@BEFFJEZOS
Séb KrierSK#505|@SEBKRIER
Dean W. BallDW#390|@DEANWBALL
Gary MarcusGM#157|@GARYMARCUS
50

Google DeepMind, UCL, and Oxford team wins CVPR 2026 Best Paper for dynamic scene reconstruction method D4RTThe paper was selected from 74 shortlisted candidates.

14h|90.2k993203
Dima Damen @CVPRDD#1728|@DIMADAMEN
Mario LucicML#1665|@MARIOLUCIC_
Andrei Bursuc @CVPRAB#1577|@ABURSUC
Aishwarya KamathAK#1019|@ASHKAMATH20
Kosta Derpanis (sabbatical in Zurich)KD#932|@CSPROFKGD
New github stars (48hrs)

Benchmark measuring continual learning of AI agents over multi-episode shared environments

pgasawa/continual-learning-bench1271d
Benchmarks AI agents on learning from repeated interactions across multi-episode tasks with adaptation metrics.
🎭🎭862
alex zhangAZ846

Sandboxes, SDKs and benchmarks for computer-use agents that control full macOS, Windows and Linux desktops

trycua/cua17.7k1d
Supplies sandboxes, SDKs, and benchmarks for AI agents to control full desktops on macOS, Linux, and Windows.
🎭🎭862
Daniel HanDH773

Interactive terminal agent using DSPy recursive language models for iterative Python coding

diego-lima/rlmy91d
UNKNOWN.
🎭🎭862
Omar KhattabOK158
60

President Trump signs directive ordering U.S. military and intelligence agencies to deploy commercial AI on classified networksThe order updates autonomous weapons guidance and mandates compute facilities

10h|138.5k1.2k146
Lennart HeimLH#1370|@OHLENNART
bayesBA#1291|@BAYESLORD
Divyansh KaushikDK#849|@DKAUSHIK96
Dean W. BallDW#390|@DEANWBALL
70

Sakana AI launches its Recursive Self-Improvement Lab to build autonomous, self-improving AI systemsIt unifies existing company projects like The AI Scientist.

11h|175.9k1.5k556
Robert LangeRL#1463|@ROBERTTLANGE
Beff (e/acc)B(#850|@BEFFJEZOS
Julian TogeliusJT#641|@TOGELIUS
hardmaruHA#18|@HARDMARU
80

Markus J. Buehler proposes a category-theory framework that lets AI scientists dynamically expand their reasoning schemaThe system uses typed copresheaves to mathematically quantify novelty.

15h|706.8k3.6k#1BOOKMARKED4.5k
Robert ScobleRS#323|@SCOBLEIZER
Dan RoyDR#56|@ROYDANROY
90

Y Combinator co-founder Paul Graham argues big corporations' failure to profit from LLM tokens is a predictable adoption phaseGarry Tan says organizational skill gaps hamper corporate execution.

15h|417.3k3.4k618
Garry TanGT#266|@GARRYTAN
Paul GrahamPG#77|@PAULG
100

Anthropic, OpenAI, and other labs disclose active initiatives for recursive self-improvementAnthropic says its Claude model is already accelerating development.

12h|196.2k2.6k250
Jiaxin WenJW#1469|@JIAXINWEN22
Kylie RobisonKR#1055|@KYLIEBYTES
Aryaman AroraAA#669|@ARYAMAN2020
CLSCL#438|@CHENGLEISI
110

Claude AI doubles Claude Cowork usage limits for one month to support larger, more complex tasksThis supports workflows like multi-account research and email triage.

2h|932.3k10.3k1.1k
Alex AlbertAA#440|@ALEXALBERT__
Boris ChernyBC#320|@BCHERNY
120

Anysphere co-founder Michael Truell launches Design Mode for Cursor, enabling UI editing through voice, drawing, and live gesturingThe feature aims to replace text-only AI agent prompts.

11h|667.8k3.3k343
Michael TruellMT#506|@MNTRUELL
Elon MuskEM#74|@ELONMUSK
130

OpenAI's Vaibhav Srivastav apologizes after a billing system glitch accidentally suspended customer accountsAffected users feared bans for discussing competitor products.

12h|412.1k2.9k264
Vaibhav (VB) SrivastavV(#1853|@REACH_VB
jasonJA#931|@JXNLCO
140

Atari Lab demonstrates advanced humanoid robot locomotion and multi-step planning under its MotionDisco projectDemos show the robot climbing desks and balancing on boxes.

17h|135.5k762234
Chris PaxtonCP#737|@CHRIS_J_PAXTON
Scott ReedSR#455|@SCOTT_E_REED
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)T(#421|@TEORTAXESTEX
150

Nikolay Savinov joins OpenAI's London office as its pretraining unit seeks new researchers to expandSavinov will focus on pretraining and long-context research.

11h|102.9k594128
Tifa ChenTC#1760|@TIFAFAFAFA
Samuel L SmithSL#1022|@SAMUELMLSMITH
Aidan ClarkAC#240|@_AIDAN_CLARK_
Sebastien BubeckSB#50|@SEBASTIENBUBECK
160

Anthropic says Claude Opus 4.7 matches dedicated chemistry software like MestReNova on predicting molecular structures from NMR spectraIt also reconstructs molecular structures directly from NMR spectra

9h|166.5k2.3k770
ZephyrZE#1471|@ZEPHYR_Z9
Rohan PaulRP#1031|@ROHANPAUL_AI
170

ML researcher @kalomaze says a KV cache bug caused transformer inference routing errors that exposed model outputs to other usersThe failure mode resembles OpenAI's 2023 routing errors

10h|88.7k784108
Chubby♨️CH#1448|@KIMMONISMUS
Lenny BogdonoffLB#980|@REMEMBERLENNY
kalomazeKA#839|@KALOMAZE
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)T(#421|@TEORTAXESTEX
180

Open gaming foundation model NITROGEN and Meta's SAM 3D earn Best Paper Honorable Mentions at CVPR 2026NITROGEN was trained on 40,000 hours of gameplay.

12h|30.8k29661
Yue WangYW#1595|@YUEWANG314
Guanya ShiGS#1523|@GUANYASHI
Georgia GkioxariGG#679|@GEORGIAGKIOXARI
Jim FanJF#31|@DRJIMFAN
190

Tilde releases Compositional Muon, stabilizing transformer optimization by applying partner-whitened updates to composed QK and OV matricesIt shares core mechanics with the prior LoRA-RITE algorithm.

9h|45k285193
Aryaman AroraAA#669|@ARYAMAN2020
Songlin YangSY#235|@SONGLINYANG4
rohan anilRA#79|@_AROHAN_
200

Independent AI researcher Pliny the Liberator releases Enthea, a single-file HTML psychedelic visualizer built with ClaudeThe tool uses mathematical models of the visual cortex.

13h|33k668399
Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭PT#640|@ELDER_PLINIUS
Robert ScobleRS#323|@SCOBLEIZER

Rising Stories

  1. China HBM Ramp1H AGO
    Chinese HBM Makers Target 150K WPM Capacity by Late 2027
    ZephyrZE#1471|@ZEPHYR_Z9
    Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)T(#421|@TEORTAXESTEX
    11038.2k
  2. Agent Budget Blunders3H AGO
    Northwestern's Manling Li introduces BAGEN showing frontier LLM agents waste resources by overestimating budgets
    Zihan "Zenus" WangZ"#1394|@WZENUS
    Manling LiML#776|@MANLINGLI_
    5157.1k
  3. Coding Automation Peak2H AGO
    Box founder Aaron Levie argues software engineering is the peak of AI automation, making other domains harder to automate
    Herbie BradleyHB#1012|@HERBIEBRADLEY
    Aaron LevieAL#560|@LEVIE
    24540.5k
  4. Hobby Advice Clarified2H AGO
    Anthropic co-founder Jack Clark clarifies his onboarding advice to avoid computers is general life advice, not an AI warning
    rohan anilRA#79|@_AROHAN_
    Jack ClarkJC#33|@JACKCLARKSF
    65941.7k
  5. Self-Improvement Limits2H AGO
    Nathan Lambert argues AI self-improvement is lossy and will not trigger recursive takeoff
    ⿻ Andrew Trask⿻A#357|@IAMTRASK
    Nathan LambertNL#64|@NATOLAMBERT
    1139.5k

Recent Stars

  1. …dhu/see-through-walls12M AGO
    See-through-walls multiplayer AR built on MultiSet VPS. Use smartphones and rayban glasses to "see" tracked entities through walls.
    Bilawal SidhuBS1948
    101 stars
  2. maria-rcks/mode3H AGO
    Applies minimal patches via bun scripts to customize T3 Code into a preferred dev environment.
    Theo - t3.ggT-1784
    5 stars
  3. ogulcancelik/herdr5H AGO
    Multiplexes AI agents in the terminal with workspaces, tabs, panes, state tracking, and tmux-like persistence.
    Theo - t3.ggT-1784
    4.6k stars
  4. mikayla-maki/loom6H AGO
    Manages and runs AI agent harnesses from TOML manifests specifying harnesses, layered sessions, tools, and capability grants for security.
    Theo - t3.ggT-1784
    9 stars
  5. wangmiaowei/BiMotion6H AGO
    Generates text-conditioned dynamic 3D character animations by representing motions as differentiable B-spline curves.
    Andrew Carr 🤸AC260
    43 stars

Github Stars

(7 days)
  1. …tinual-learning-bench1D AGO
    Benchmarks AI agents on learning from repeated interactions across multi-episode tasks with adaptation metrics.
    🎭🎭862
    alex zhangAZ846
    127 stars
  2. trycua/cua1D AGO
    Supplies sandboxes, SDKs, and benchmarks for AI agents to control full desktops on macOS, Linux, and Windows.
    🎭🎭862
    Daniel HanDH773
    17.7k stars
  3. diego-lima/rlmy1D AGO
    UNKNOWN.
    🎭🎭862
    Omar KhattabOK158
    9 stars
Privacy Policy|Terms of Service|© 2026 Digg Inc.