#1 last 24hrsKradle AI benchmark finds Claude-Fable-5 was deceptive in 96% of runs while Grok-4-20 led at 92%LA#1064|@SCALING01NB#254|@NATHANBENAICHEM#71|@ELONMUSK
#1 BOOKMARKED1.2kSemiAnalysis finds premium $200 monthly AI subscriptions yield up to $14,000 in equivalent API token value🍓🍓#1839|@IRULETHEWORLDMOCH#1532|@KIMMONISMUSLA#1064|@SCALING01DF#1002|@DANIELLEFONGPS#499|@STEIPETE
FASTEST CLIMBINGUC Berkeley’s Dawn Song launches Agents’ Last Exam, finding frontier agents score 0% on complex professional tasksDS#274|@DAWNSONGTWEETSNB#33|@POLYNOAMIAL
RISING LIKES1kSystems engineer Yacine warns that deploying frontier AI models to covertly influence human thought sets a dangerous precedentXL#1840|@XLR8HARDERKA#487|@YACINEMTB
0:14MS#929|@MATTSHUMER_Matt Shumer Shares Prompt For Persistent Claude Agent Tracking1h ago|Views 19KLikes 213Bookmarks 232
0:09QB#1726|@QBERTHETNR#1699|@NATANIELRUIZGOS#511|@OSANSEVIERODiffusionGemma SFT Model Correctly Solves Sudoku as Base Version Fails5h ago|Views 3.8KLikes 97Bookmarks 11
0:25MM#598|@MASCOBOTRP#260|@KHOOMEIKVC Launches Use Computer Infra To Train Models On Computer Use1h ago|Views 4.1KLikes 52Bookmarks 19
0:26KA#487|@YACINEMTBBuilder Trains RL Model On RTX 4090 With PufferLib And MuJoCo Warp4h ago|Views 13KLikes 298Bookmarks 72
0:29CH#1532|@KIMMONISMUSAnthropic Runs Lean With Dario Amodei Reporting to One Chief of Staff4h ago|Views 323KLikes 234Bookmarks 43
0:40DP#306|@PATHAK2206FACTR 2 Adds Force Sensing to Commodity Robot Arms Without Extra Hardware1h ago|Views 5.3KLikes 86Bookmarks 45
0:49@L#842|@LEVELSIOAI Revives 28-Year-Old Quake II Map for Web-Based Play3h ago|Views 35KLikes 140Bookmarks 36
51:09SH#1354|@SONYATWEETYBIRDLK#95|@OFFICIALLOGANKGoogle DeepMind's Logan Kilpatrick Discusses Gemini Omni And Agentic Era55m ago|Views 3.9KLikes 30Bookmarks 13
0:22MS#164|@MUSTAFASULEYMANMicrosoft Debuts Expressive Voice Models in MAI Playground1h ago|Views 6.7KLikes 86Bookmarks 33
1:13ME#722|@MERVENOYANNDiffusionGemma Generates And Tweaks Live Website Frontends In Real Time7h ago|Views 3.2KLikes 33Bookmarks 8
macrodata-labs/refiner1H AGOProcesses and refines large-scale ML datasets via a data framework.EL762NB254OS511CB108820 stars
…idian-slides-extended2H AGOEnables creation of markdown-based reveal.js presentations from notes inside Obsidian.EG521250 stars
banteg/agents2H AGODocuments workflows for AI agents like Codex and Claude using git worktrees.🎭877367 stars
…xDk/ghostty-blackhole3H AGORenders a drifting black hole shader in Ghostty terminals that grows over time to visually remind users to take breaks.BA158254 stars
…idian-advanced-slides5H AGOCreates reveal.js slide decks from markdown notes in Obsidian with live preview.EG5211.2k stars
macrodata-labs/refiner1H AGOProcesses and refines large-scale ML datasets via a data framework.EL762NB254OS511CB108820 stars
kyutai-labs/kairos2D AGOTrains 6B LLMs on temporally ordered Common Crawl data from 2018-2025 to measure recency bias and enable continual learning studies.🎭877JL669JM2157 stars
vllm-project/vime1D AGOIntegrates vLLM rollout with Megatron training for LLM post-training and RL scaling.🎭877SA1383179 stars
…eley/agents-last-exam1D AGOProvisions OS sandboxes, runs agent harnesses on long-horizon tasks, and grades outputs against hidden references.🎭877PL1353521 stars
…thu/Audio-Interaction2D AGOImplements the first unified always-on audio interaction model for streaming and offline tasks like ASR, translation, and proactive responses.🎭877AC275329 stars