Shawn Wang

btw we list job postings for our AIE sponsors and @sentry has a very good one here see them all at http://ai.engineer/jobs very proud and grateful of the AIE Expo this year. this is the highest concentration of koding talent NYC will have had in basically ever. team did well.swyx🔜 @aidotEngineer CODE 🗽: whoaaaah Sentry gathering some serious js talent recently Link: https://x.com/swyx/status/1989173675736854636

seeing a lot of evidence that Gemini 3 is rolling out soon... i'm not saying anything specific here but i'd definitely tune in to the @aidotengineer CODE livestream next week if I were you (https:// ai .engineer / youtube) (also for @thekitze's banger first talk at AIE!)kitze: just tried gemini 3.0 jesus fucking christ Link: https://x.com/thekitze/status/1989111025271206053

whoaaaah Sentry gathering some serious js talent recently

as I talked to many many friends (thank you!) about the Priscilla x Zuck pod, I found that many of you non biologists feel intimidated by the science. Don't worry, the Biohub story is now primarily building a frontier AI lab, so I put this in more familiar terms to you in today's @latentspacepod writeup. also included an extract of my favorite quote from Priscilla, where she articulates why exactly they're aiming at the Immune System next after solving the virtual cell. Makes sense that if we want to keep people healthy, we should deeply understand nature's first party healthcare system in order to fix/supplement it!Latent.Space: Priscilla Chan and Mark Zuckerberg co-founded the Chan Zuckerberg Initiative (CZI) in 2015, committing 99% of their Meta shares to advance science, education, and opportunity. As a pediatrician and CEO of Meta respectively, they've built CZI into one of the most ambitious Link: https://x.com/latentspacepod/status/1986487301397176813

ok lots of the most cracked ai ux people at @sync_conf today but i’m going to be thinking about this demo for a long time heres @_adamwiggins_ , @threepointone , me et al grilling Steve about his new collaborative multiagents on Tldraw!!Steve Ruiz: An interesting realization as we’re working on multi-agent collaboration fairies project: we’re kind of rediscovering “old ai” as I understood it in videogames, ie entities with a programmatic state machine responding to events, except with “new ai” handling the rest Link: https://x.com/steveruizok/status/1986164318032044338

yes Aether is a biglab model, no its not a llama one thing i learned from the inside is that all code ide model strings are combinations of models combined with prompts and toolcalls for the model. so no you wont feel like it is “raw” matching what youd find in the normal chat. abandoned blogpost: The Model Selector is a Lieswyx: {highly anticipated model} is coming sooner than all of you anons have been speculating. can try all 3 sizes now in @windsurf Next and see if you can figure it out :) Link: https://x.com/swyx/status/1988425979216437653

## Lists outside, Detail inside. one fun discovery i've had using Cursor 2.0 is that the Agents view and the Editor view are not actually that different if you happen to be #rightsidebar gang I've been #rightsidebar since @shanselman first talked about it many moons ago and just realized my natural koding layout looks a bit like this now. I encourage people to think about AI UX in terms of literal screen real estate dedicated to agents. with Cursor 1.0 it was ~25%, with Cursor 2.0 it's now ~50-100% depending how you use it. I like being able to inspect my filesystem (which also doubles as logs) so I'm at 50%. but its interesting to see how LHS = Agent and RHS = Manual and have the karpathy "autonomy slider" visually manifest as the VSCode pane splitters.swyx: first try of the new Cursor Composer model (btw I'm still a DAU of Cursor! @smol_ai is entirely a Cursor vibecode) one impressive example - Composer 1 finished 2 rounds of human feedback and debugging with me and got me what I wanted, while Sonnet 4.5 was still working on its Link: https://x.com/swyx/status/1983585407368609909

one notable problem about this dream of "continuous learning" models / "self evolving" agents is that the business model of centralized ai right now is quite antithetical to this try telling your GTM people you can't hype customers about a new version of the product because the product versions itself try telling your agent engineers they need to improve the agent, but no they can't access real user data to do so the transition to Level 4 probably involves going from "global best" to local maxima and that is going to be a very hard reset for a lot of the way AI strategy is set up today for most agent labsprinz: Google has released a new "Introduction to Agents" guide, which discusses a "self-evolving" agentic system (Level 4). "At this level, an agentic system can identify gaps in its own capabilities and create new tools or even new agents to fill them." https://www.kaggle.com/whitepaper-introduction-to-agents Link: https://x.com/deredleritt3r/status/1988244609504334283

ok one of the things that i've always wanted an AIE is coming to pass, after the Great @dylan522p v @jefrankle debate of 2024: the Great MCP debate! @vtahowe and @ianlivingstone are taking on all challengers - if you are a knowledgeable MCP skeptic, come do a live debate next week at AIE CODE! sign up/let Allie know you are interested!Allie Howe: Some say MCP is >obsolete >a security nightmare >creates unnecessary abstraction Others say MCP provides >standardization >a security boundary at the network layer >federated data integration Which is it? Time to host a debate, live at @aidotengineer CODE next week! Link: https://x.com/vtahowe/status/1988050009233723890

RT Surge AI Everyone's acting like models are ready to replace humans in work settings. We put that to the test by creating an entire company and having 9 models act as a customer service agent handling 150 tickets and requests of increasing complexity. Verdict: without common sense, models are nowhere near ready. 👇 https://surgehq.ai/blog/rl-envs-real-world

RT Latent.Space "You know for the next period, we really wanna make science the main focus of what we're doing and specifically, the Biohub is really gonna be like the main focus of our philanthropy, and it's just something that we're very excited about. When we started 10 years ago, we had this idea like, 'Okay, I bring experience as a physician, Mark's an engineer and he builds things, and we have an opportunity to give back resources to make an impact on this world.'" Priscilla and Mark elaborate on their future mission of their philanthropy and how they can leverage their experiences to do something special. @czi @officialbiohubLatent.Space: Priscilla Chan and Mark Zuckerberg co-founded the Chan Zuckerberg Initiative (CZI) in 2015, committing 99% of their Meta shares to advance science, education, and opportunity. As a pediatrician and CEO of Meta respectively, they've built CZI into one of the most ambitious Link: https://x.com/latentspacepod/status/1986487301397176813

RT Allie Howe Some say MCP is >obsolete >a security nightmare >creates unnecessary abstraction Others say MCP provides >standardization >a security boundary at the network layer >federated data integration Which is it? Time to host a debate, live at @aidotengineer CODE next week!

RT SMB Attorney Wow... Warren Buffet says goodbye in his final annual letter today (full copy in the comments below). As he signed off, the following were final words of advice: "One perhaps self-serving observation. I’m happy to say I feel better about the second half of my life than the first. My advice: Don’t beat yourself up over past mistakes – learn at least a little from them and move on. It is never too late to improve. Get the right heroes and copy them. You can start with Tom Murphy; he was the best. Remember Alfred Nobel, later of Nobel Prize fame, who – reportedly – read his own obituary that was mistakenly printed when his brother died and a newspaper got mixed up. He was horrified at what he read and realized he should change his behavior. Don’t count on a newsroom mix-up: Decide what you would like your obituary to say and live the life to deserve it. Greatness does not come about through accumulating great amounts of money, great amounts of publicity or great power in government. When you help someone in any of thousands of ways, you help the world. Kindness is costless but also priceless. Whether you are religious or not, it’s hard to beat The Golden Rule as a guide to behavior. I write this as one who has been thoughtless countless times and made many mistakes but also became very lucky in learning from some wonderful friends how to behave better (still a long way from perfect, however). Keep in mind that the cleaning lady is as much a human being as the Chairman."

RT Akshay Kothari Notion’s Q3 was our finest yet. Growth is accelerating at scale. We used to sell software; now we sell work itself. That’s a profound shift, and the numbers show it. I ran 50+ demos this past quarter, refining our story for customers each time. I shared the latest version internally last week and thought: why not share it with all of you too? It’s an 18-minute watch; feedback welcome!

Ahead of AIE CODE presented for the first time by @GoogleDeepMind, I'm happy to release this special weekend chat with our Day 2 emcee @jedborovik, product lead of @Julesagent: https://www.youtube.com/watch?v=emWgP_fr04k&lc=Ugy2OGAfghkKIgO_bIl4AaABAg which was very fun because Jed turned the tables on me and challenged me on what I think needs to happen next for the hyper competitive landscape of Coding Agents. see you soon in New York!

i think this is beautiful - doing ViT from raw pixels means you need to jointly train everything - this poor model must independently solve MNIST, and THEN/ALSO learn to be a perfect calculator in its weights. then keep going.... only constrained by the data you give it. it's why @percyliang's concept of "foundation models" in 2021 was so disruptive/sacrilegious in the Google vs OpenAI sprint to the GPT: instead of 1000 different small models all specialized in their tasks, concentrate all that budget/data/resources in one supermodel that has the capacity to model 1000 tasks; along the way you get 1) transfer learning, 2) capabilities you never explicitly trained for, 3) emergent abilities that only unlock at a given param/depth/data exposure rate.varchasvi: @karpathy said we should delete tokenizers. So I did, and it worked! I read @karpathy 's post on OCR where he talks about models that read text straight from pixels instead of using a tokenizer (and how having a tokenizer is a problem), and I wanted to build one such model from Link: https://x.com/varchasvee_/status/1986811191474401773

RT Latent.Space Congratulations @alexgshaw @Mike_A_Merrill on the exciting and impressive launch! Shoutouts to @lschmidt3 , @andykonwinski, @LaudeInstitute for supporting this amazing work!Alex Shaw: Today, we’re announcing the next chapter of Terminal-Bench with two releases: 1. Harbor, a new package for running sandboxed agent rollouts at scale 2. Terminal-Bench 2.0, a harder version of Terminal-Bench with increased verification Link: https://x.com/alexgshaw/status/1986911106108211461

RT Alex Shaw Today, we’re announcing the next chapter of Terminal-Bench with two releases: 1. Harbor, a new package for running sandboxed agent rollouts at scale 2. Terminal-Bench 2.0, a harder version of Terminal-Bench with increased verification

"use brain"Aiden Bai: also who the hell made this diagram. so based Link: https://x.com/aidenybai/status/1985819403825664233

RT Gabe Greenberg we've been cooking @g2i_co very excited to party with @swyx on this! @MichelleBakels / @Beccalytics are organizing @ReactMiamiConf and @AIEMiami the same week!!!AI Engineer: Miami: The world's leading AI Engineering conference is coming to Miami! AI Engineer: Miami, April 20–21. Two days. One track. A community of engineers and founders on the frontier of AI technology. Organized by @g2i_co, supported by @aiDotEngineer. Join us! https://www.ai.engineer/miami Link: https://x.com/AIEMiami/status/1984305102920785977