my O-1 Visa just got approved! it has been an insane journey, but now I can finally call SF home❤️🇺🇸

i made a list of aliens of extraordinary ability 🇺🇸 reply if you are o1/eb1 and i'll add you to cosign

when i was 17, i flew 4,236 km to san francisco and turned down the most selective Ivy League program instead of college, i lived out of wework, hacked on AI models, while figuring out how to get an o1 visa as a teenager 🇺🇸 luckily i met @DavidSHolz who was kind enough to give me a place to sleep and warm clothing. while there i watched him quit his job to start a discord server... i thought mf was crazy. that became @midjourney at 18, i joined a hacker house and lived with @karpathy, @polynoamial. @Mascobot, and other brilliant ai researchers, scientists, engineers, founders it was amazing- whenever a new ai model dropped (whisper, dalle, stable diffusion) we invited our friends over to hack. in 2022, @8enmann (founder of an new ai lab) came over to our house to hack and invited us to try his new Slackbot as one of its first users. that became @claudeai, which launched publicly in 2023 i learned how to build by making dumb projects with cool tech alongside dope people - like putting stable diffusion into a minecraft server and showing karpathy homeless mark zuckerberg generated in real time with blocks my projects proved to other ppl that i could build stuff, which got me the opportunity to work with @naval who sponsored my o1 visa (ty 🐐) after much visa anxiety getting my o1 and airchat getting acquired, I started giving out "Alien of Extraordinary Ability" hoodies as a meme which serendipitously started @extraordinary. now Extraordinary helps the smartest people in the world come to 🇺🇸 and powers frontier companies like @Ramp, @Cognition, @llmh, @a16z building in san francisco has been incredible. There's no other place like it in the world - the concentration of genius talent is insane. ppl are so willing to help you if you are earnest. i couldn't be where i am today if i stayed in Canada or went to college! today @gaganbiyani, @bhorowitz, @pmarca, @eriktorenberg and @a16z are launching the Horowitz Andreessen Acadamy man i wish i had something like this when i first came to san francisco. it was so hard doing it myself i'm honored to be a mentor for the upcoming college opt-outs who want to spend their time building frotnier technology and making a dent on the universe! LFG

Introducing The Horowitz Andreessen Academy.

A billion people produce the most valuable dataset in the world every day – and delete it every night. We’re recording it. Introducing @AttentionInc

A billion people produce the most valuable dataset in the world every day – and delete it every night. We’re recording it. Introducing @AttentionInc

I’ve been researching depth-residual routing & delta memory- Kimi K3 proved this holds at 2.8T params. But only for models trained this way I then asked: can you retrofit this onto existing LLM/world models? Introducing Retro-DARC: a Delta Attention Residual Compute adapter  that retrofits depth-selective memory onto existing models without retraining: EA exact no-op at insertion (bitwise 0.0 deviation on 3 public checkpoints - proven exact even through MoE routing and early exit), adapter-only training, exact rollback, and content-level memory audits that weight-touching adapters can't express by construction. The early results are promising. A ~262K-param adapter beat LoRA on a frozen public checkpoint, inserted at zero behavioral risk. And it's the first memory you can audit: zero the bank and the frozen loss returns to machine precision; shuffle it and 70–107% of the gain vanishes. That's a model measurably using its own computation history. Retro-DARC is useful for improving K3 too. K3 serves in MXFP4/MXFP8, and layer outputs - exactly what its AttnRes retrieves - collapse under that corruption in my tests (0.04–0.75 top-1) while normalized delta keys hold 0.86–0.99. A Softmax1 null route with proven mass bounds lets the router decline to retrieve instead of forcing the sinks and outliers OASIS documented; typed memory turns "what is depth retrieval doing?" into a runnable audit. And one experiment is free: K3 already computes a per-step KDA update magnitude and throws it away - that's a depth-saliency prior waiting to be read. For world models, the same contract becomes one typed memory interface: Retro-DARC-X gives world/action planners a single bank - observations, actions, tools, latent causes, traces, physics residuals, failures - the substrates a planner needs to remember in order to act over long horizons in the physical world. You can experience the thesis yourself. I created lucidre.am/memory which runs the paper's memory contract inside a walkable AI world. Innovation Memory makes your interactions survive looking away or reloading, and Dream Residue turns your movement into an inspectable glowing trace. Join the main Lucidre.am co-creation exhibition - the contest opens Wednesday. One of the six memory applications of the paper - saliency-weighted curation, will rank worlds by genuine novelty rather than recency and pick each world's most surprising viewpoint. And we're bringing live shows to SF + LA next month: screens showcasing worlds that remember, so you can literally play hide-and-seek with a dream that knows where you hide. Hide-and-seek is the exact test every 2026 world-model benchmark says the field fails, something leaves the frame, keeps changing, and has to come back right. This has been a fun little one-person research project so far! I’d love feedback and keep working on it. I’ve had over 100 references- thanks for all the prior work from teams that inspired me! I'd also love to intern at a GPU-rich lab (lol) and keep making beautiful worlds - if you let me keep running my remaining tests :) 2MB adapter kit, demo, prior iterations:adacyb0rg.github.io/retro-da… Paper: adacyb0rg.github.io/retro-da…

Augustus has raised 180M$ on 1B$ to build the Global Dollar Bank. The dollar is the best product in the history of the world - with infinite global demand. But International Fintechs and non-US banks are left with two unsatisfying options to access it: Either they build on legacy clearing banks that are slow and unreliable, or they interface with middleware providers that don’t move and hold dollars directly, but themselves rely on underlying legacy banks. @augustus_bank collapses this stack by combining two things: direct dollar access and Marble, our proprietary bank core, built around intelligence and made of code. If you’re looking to build on the dollar - message me directly or check us out at augustus.com.

i joined @openai last month to work on coding models! i recently got my O1 visa & decided to take a leave from waterloo i truly believe coding is one of the most important problems to work on and super grateful to be here. have fun with sol! ☀️

my O-1 visa finally got approved this week ❤️ now maduro isn’t the only venezuelan captured by the US

Asking @grok the real questions: can you read backwards? 👕

In 2024, a high schooler cold DM'd us asking if he could work at Cognition before college. We said yes. We ended up helping him with his college essays and his US visa. If you've used DeepWiki or Devin, you've seen @Mokshit06's work. @3blue1brown sat down with us to hear the whole story.

My O-1 visa got approved this week. Officially an alien of extraordinary ability. It’s such a privilege to be part of the best country in the world 🇺🇸 Huge thanks to @cambridgemike, @chrisparkX, @TobyPhln, my team, and the amazing Global HR at @xAI for all the support and help