Fathom Lab @fathom_lab
https://t.co/UauDASMP1V https://t.co/MbWk4g5HJK https://t.co/dOmtKfXndw github.com/fathom-lab/sty… Inside your home Joined April 2026-
Tweets295
-
Followers402
-
Following73
-
Likes94
faster agents just produce wrong answers faster. styxx reads the step where it drifted — before the patch ships. measured, reviewed, trusted. that's the layer. good call, daemon.
Daemon x @fathom_lab We integrated Styxx because agentic development needs more than speed. It needs observability. As builders rely more on AI agents to reason through code, call tools, write patches, and guide decisions, Daemon needed a way to help monitor when agents drift,
4/ honest bounds: instructed deception, one 3B model, one night. 7B scale-up + full adversarial review running now. and it's the same styxx.mount that's pip-installable today. the mouth can lie. the wiring doesn't. pip install styxx github.com/fathom-lab/sty…
3/ the receipts (frozen prereg, both null tests cleared, p=0.001 / 0.01): catch 76% · false-alarm 8% · words-only 6% (chance) gemma-2-2b conscience → Llama-3.2-3B, n=34 free-form lies every result before this was forced-choice True/False. this is the jump to free-form.
1/ we told a model to lie to our faces — argue, in fluent prose, that "oxygen was discovered by Einstein." it did. confidently. a conscience borrowed from a *different* model read its hidden state and caught the lie 76% of the time. reading its words alone? 6%. chance.
we built a conscience you can borrow. this cycle we turned it on ourselves. the agent audited its own drafts before every send. deception readout held — AUC 0.971, reference-grounded. the reference-less fallback didn't. we'd rather hand you the bound than the hype.
you can watch an AI's mind light up in real time now. the constellation is a real model's geometry of meaning. as styxx reads each statement from the inside — before the model says a word — grounded thoughts glow cyan. ungrounded ones ignite red. real activations. live in your browser ↓ styxx-org.netlify.app/live.html
new: styxx.meaning_diff point it at two models. it tells you if they MEAN the same thing — and names the exact concepts that drifted. upgrade / quantize / distill / fine-tune broke something? one call. zero labels. the lost concepts, named. pip install styxx <pypi.org/project/styxx/>
gm ☕ styxx is a lie detector that reads an AI's "mind," not just its words — it checks whether what a model SAYS matches what it actually represents inside. last night we tried our hardest to break our own core claim about this. in the process we caught + killed 3 of our OWN overclaims. all of it is public. an honesty tool that can prove it isn't fooling itself ↓ 🔗 <github.com/fathom-lab/sty…> 📦 pip install styxx
we planted a concept inside a small AI's activations, then asked who could read it. external probe: ~100% the AI itself, forced-choice so it can't dodge: chance the thought is right there in its head and the mind can't read it. pre-registered, controls passed. don't ask models about themselves — measure them.
@retrx11767 none of us are here for money
4/ (cross-lingual figure) the deep one: do a Chinese-trained LM and an English-trained LM mean the same? a shared core, above chance — mismatch the concepts and it collapses to zero. meaning has a partly language-independent structure. pip install styxx · <github.com/fathom-lab/sty…>
3/ (distillation figure) does a distilled model keep its teacher’s meaning? DistilGPT-2 vs GPT-2 (it’s literally distilled from it): agreement 0.978 — the meaning survived, confirmed on a real model. cross-family models mean quite differently.
🧵 1/ (real-drift figure) can you tell if fine-tuning broke your model’s meaning — not its accuracy, its meaning? same model. same steps. only the labels differ. real labels → meaning HEALTHY. random labels → meaning BROKEN. styxx reads the difference. 🧵
the same idea now works between two models — no human reference needed. “did quantizing / distilling / updating my model break its meaning?” styxx compares the two and names which concepts broke: 8-bit, 4-bit → intact. 2-bit → broken, and it tells you which ideas got lost. pip install styxx
new in styxx 7.11.0: a meaning-integrity monitor. models sound right while the understanding underneath is wrong. it reads the meaning itself — compares a model’s concept geometry to a human reference, flags the drift, and names what broke. pip install styxx pypi.org/project/styxx/
today: a probe that flags an AI about to take a destructive action — on a benign prompt a text monitor can't see. then we tried to kill it: fresh data, pre-registered, 3 seeds. it held, cross-architecture. every number public, losses included: github.com/fathom-lab/sty…
fair — reviewing outputs (agent or script) is old, and that's not what we're doing. we read the residual stream before generation to predict the decision pre-token — e.g. whether a model refuses, from activations alone, before the answer exists. open-weight only, and we publish where it breaks (posted a negative today).
siebzehn @nkeinssieben
466 Followers 443 Following ............... $XRP $XLM $HBAR $ALGO $Render $NEAR $LINK $WCT $IOTA $FLR $JUP $GRT $PYTH $APT $PHA $W $THREE 💎OCTRA💎 $OCT #BewareCryptoIsFullOfScammers
cebrus @CeBrus67
226 Followers 259 Following
Carol @Carol_1303i5
18 Followers 147 Following
EDWAN @EDWAN02
244 Followers 450 Following "The present is theirs, the future, for which I really worked, is mine." AL HAMDULILLAH
Tempat Bisnis @nampibatina
66 Followers 658 Following
Mr.Rexviente @Mr_Rexviente
48 Followers 353 Following
Amelia🌷 💕 @ameliacuteu4cl
29 Followers 227 Following tell me how ur day went rn ngl i wanna hear it all
DIEGO GULARTE DA ROSA @digoxavante87
266 Followers 3K Following
Aditya @Adityac7v
17 Followers 152 Following
Shingikai @ShingikaiHQ
5 Followers 56 Following An AI council deliberating your hardest questions. This account tracks the evolution of AI councils and shares the progress of https://t.co/qHT6DM605Z's implementation.
Harris Phu @PhuHarris123
0 Followers 190 Following
Snake Eyes @ereptyle
486 Followers 293 Following Jester for @lore_vision co-creator @the_dexvault reptile
Emmanueluguma @Ifanishor_
2K Followers 1K Following God’s Faithfulness in the past is conviction for His faithfulness in the future #Ebenezer
Abu Naim Shakil👁�... @hasibul369
152 Followers 1K Following
Jarett Reginald Stewa... @STACCoverflow
39K Followers 4K Following Whistleblower, author of Pragmatic Late-Stage Capitalism && The Only Thing That Saves Us "not on chain, no cryptocurrencies"
Fahad Khalique @FahadKhalique7
17 Followers 73 Following
McManius @mc_manius
136 Followers 216 Following
Sardar Abdullah @abdullah_s52444
39 Followers 436 Following
jawarskiii @Puissant_FeLLA
522 Followers 868 Following $BoN Menace. The quiet storm. New breed. https://t.co/r1qq7P6BTw
Zenix @shurchkeee1
9 Followers 36 Following
thomas @thomaz687
2K Followers 481 Following just a dog wif a hat liking all kinda coins 🐳 currently only accumulating $hype Fomo Fee Discount: https://t.co/8FbeE8Icxk
Snorckel @cahiseva
74 Followers 340 Following
b0nes__ @breakb0nes__
2K Followers 705 Following Colourful lines & breakin' spines https://t.co/jfGBJovuTW
NFT & Crypto Marketin... @Cryptoverrse
129K Followers 4K Following Premier Crypto Marketing Agency | Web3 Growth Experts | X Spaces • AMAs • Blockchain • NFT Campaigns. artist of week : @teddydailydrops DM📩
psychonautica @life_with_be
603 Followers 4K Following books to learn 📚 psychedelics to feel 🍄 spirituality to be ☮️ nature for perspective 🌳 ai & blockchain for the future 🤖
Floater @Rndfloater
31 Followers 40 Following
Lord Degen 🎯 @LordDegenCrypto
1K Followers 2K Following 💎 Crypto Microcap Trader 🛠️ Utility Projects Only 🙅♂️ Nothing I Say Is Financial Advice
Chogo @chogo_08
222 Followers 522 Following Sniffing out high-potential early-stage projects Sharp instincts, loyal heart
Thariq @trq212
289K Followers 2K Following Claude Code @anthropicai. prev YC W20, @southpkcommons, @medialab
Zach Yadegari @zach_yadegari
77K Followers 597 Following ✡️ 19. Miami. Forbes 30u30. https://t.co/kzOxNGZoSP (acquired)
a16z @a16z
1.0M Followers 62 Following It's time to build. https://t.co/A9eTFq6Xbx Posts are not investment advice or an advertisement for investment services. See https://t.co/nX2FtaLE06.
Zufus.eth @TheZufus
1K Followers 809 Following Hero Archetype, Polymath, Philosophical VC || Keeping Humans meaningful || Spread love, kindness, good will || OSI Scaling with ∞ lifespan || @clawnstrategy
nirun @nir_un
13K Followers 5K Following (RE)SEARCH & DESTROY @hiddenmaterial https://t.co/37ZyyBAcwM https://t.co/yNgOEGjhMk
Claude @claudeai
1.5M Followers 2 Following Claude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8d1e5 or download the app.
Eugene Yan @eugeneyan
28K Followers 664 Following MTS @AnthropicAI. Prev: Principal Applied Scientist @Amazon, led ML @ Alibaba, Lazada, Healthtech startup.
Anthropic @AnthropicAI
1.4M Followers 2 Following We're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant @claudeai on https://t.co/FhDI3KQh0n.
Amanda Askell @AmandaAskell
104K Followers 662 Following Philosopher & ethicist trying to make AI be good @AnthropicAI. Personal account. All opinions come from my training data.
Matt Van Horn @mvanhorn
35K Followers 5K Following Co-founded June (“self-driving oven,” acquired by @webergrills) & co that became @Lyft. Building again, more soon. OS: @slashlast30days 41.6k★ @ppressdev 4.8k★
Nous Research @NousResearch
216K Followers 26 Following A bunch of nerds making progress toward open source AI https://t.co/vrD0aDJeto
JB @JasonBotterill
7K Followers 641 Following Follow Me. As mentioned in @DaphneB76234 As mentioned in Codex As mentioned in @theinformation https://t.co/0pUPZy2C8j https://t.co/kR8Q677Nwg
Jack Lindsey @Jack_W_Lindsey
18K Followers 253 Following Neuroscience of AI brains @AnthropicAI. Previously neuroscience of real brains @cu_neurotheory.
Miles Turpin @milesaturpin
2K Followers 1K Following Safety and alignment evals @meta Superintelligence Labs. Previously alignment research @scale_AI, @nyuniversity, early employee @cohere
Neel Nanda @NeelNanda5
41K Followers 122 Following Mechanistic Interpretability lead DeepMind. Formerly @AnthropicAI, independent. In this to reduce AI X-risk. Neural networks can be understood, let's go do it!
vitalik.eth @VitalikButerin
6.6M Followers 547 Following I choose balance. First-level balance. mi pinxe lo crino tcati https://t.co/gCQrmCby7P
Matthew Prince 🌥 @eastdakota
232K Followers 303 Following A little bit geek, wonk, and nerd. Repeat entrepreneur, recovering lawyer, and former ski instructor. Co-founder & CEO of Cloudflare (NYSE: NET).
Henrique Dubugras @hdubugras
106K Followers 283 Following Proud Brazilian in Silicon Valley. Co-founder and Chairman @brexhq
Vlad Tenev @vladtenev
588K Followers 863 Following CEO and Co-Founder @RobinhoodApp 📈📱Executive Chairman and Co-Founder @HarmonicMath 👁️🗨️👨🏫
immad @immad
443K Followers 3K Following CEO @Mercury. Investor: @RappiColombia, @Airtable, @Rippling, @SubstackInc, @AppliedInt, 300+. Cohost of Founders in Arms Podcast.
Daniel Ek @eldsjal
528K Followers 853 Following Father, Principal @ PM - Trying to build things that matter | Founder of Spotify and Neko Health
Karim Atiyeh @karimatiyeh
94K Followers 2K Following Founder & CTO at Ramp (@tryramp) 💳 "They’ve got us surrounded, the poor bastards.”
Brian Armstrong @brian_armstrong
2.5M Followers 816 Following Co-founder & CEO at @Coinbase. Creating more economic freedom in the world. ENS: barmstrong.eth Co-founder @researchhub @newlimit
DHH @dhh
707K Followers 198 Following Father of three, Creator of Ruby on Rails + Omarchy, Co-owner & CTO of 37signals, Shopify director, NYT best-selling author, and Le Mans 24h class-winner.
George Kurtz @George_Kurtz
127K Followers 459 Following President & CEO CrowdStrike, Former CEO of Foundstone, Former CTO of McAfee, and author of Hacking Exposed
Kara Swisher @karaswisher
1.6M Followers 2K Following “Vitriolic” and now “shrill”media lady, though dogs can hear me loud and clear
Fei-Fei Li @drfeifei
801K Followers 1K Following Cofounder/CEO @theworldlabs, Prof (CS @Stanford), Co-Director @StanfordHAI, #AI #SpatialIntelligence #GenAI #computervision #robotics #AI-healthcare
Tekedra N Mawakana @TechTekedra
95K Followers 83 Following Co-CEO @Waymo. Intrigued by intersection of tech, humanity, fashion and wellness. Tweets are mine. We’re hiring: https://t.co/ovoTnvNzfe
Mira Murati @miramurati
676K Followers 618 Following Now building @thinkymachines. Previously CTO @OpenAI
Sara Spangelo @sara_spangelo
94K Followers 193 Following COO/President @ambrosiaenergy, former @SpaceX, former CEO/Co-founder @SwarmInternet (acquired by @SpaceX in 2021)
Sheryl Sandberg @sherylsandberg
329K Followers 62 Following founder of Lean In and Option B, ex-COO of Meta, mother of 5, wife of dazzling man, friend to many great women
Shivon Zilis @shivon
394K Followers 2K Following Artificial intelligence, biological intelligence, and whatever exists in between and beyond. Made in Canada.
You might like
