Aman (@multivitaman) header
Aman (@multivitaman) profile picture

ai eng @ https://teapawt.org/t.co/9I3D15Ruj8 | prev @stripe | nyc

United States
Joined June 2014
Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)Photo by Aman (@multivitaman)
this would an absolutely massive unlock
Today we’re launching cross-model KV cache transfer from @OneTriangle, a step towards efficient model communication. Up to 10.8x faster TTFT at 1M context. Prefill small model, transfer KV cache, decode on large. SDK live for Qwen3.5-4B → Qwen3.8-27B, beta open for more model pairs.
78
Most people don’t look for a rocketship in financial services. They should. That’s part of why we like it here. It’s an industry built on trust, still running on too much paperwork, with very little built for the ones that matter the most: the consumers and their financial lives. Arca made @lennysan’s Next 50. We’re building the AI-native operating system behind financial services. Hiring across engineering, product, design, and marketing --> arcawealth.ai/
The best time to join a rocketship is before everyone knows about it. I’m excited to launch The Next 50: the 50 hottest early-stage startups to keep an eye on (and try to join). Using @harmonic_ai’s data, the juiciest startup dataset out there, we looked at every early-stage startup and ranked them on headcount growth, recent funding, and investor interest. We then picked the top 50. Here's the full list: lennysjobs.com/next50 We’ll update it quarterly, and I’d love your feedback on what other factors we should consider. P.S. This is in addition to the Lenny 100, the top 100 well-known companies with the highest talent density, tackling the most ambitious problems: lennysjobs.com/lenny100
2
1
12
432
most agent harness innovation is aimed getting agents to reliably churn through longer and longer running tasks. financial services work however, is less long-running, more long horizon: a 401k rollover that takes weeks across multiple emails, three phone calls to the provider, a phyiscal check in the mail, and multiple human handoffs and decisions in between. arca's agent harness was built for exactly this shape of work, allowing agents to drive long horizon outcomes through people. my talk @aiDotEngineer covers how we built this, and how this enables us to steer how our business operates by steering our agents. (we're hiring!) youtube.com/watch?v=oP2WLFpM…
7
1
23
2,044
"I don't even remember what work looked like before Arca and its agents" — one of our advisors, today financial advice is getting its coding agent moment @arca_wealth
1
8
401
that stat is bonkers. the fortune 500 spend just 0.1% of their expenses with ant. there’s room to 10x, 20x, 50x (!!) this number and still be at 1-5% of enterprise spend on AI wild
Replying to @socialcapital
At the end of 2023, Anthropic had ~20 customers spending $100K+ a year. As of September, it had more than 11,900. Yet in Q3, its Fortune 500 customers spent, on average, less than 0.1% of their total expenses on Anthropic.
1
140
this is such an oddly specific, deeply human feeling being tracked by *one* neuron of an LLM that specific, momentary, quiet smugness of a bargain that paid off? yeah it has a neuron for that (from @nickcammarata’s essay)
1
128
spoke to enterprise leaders at @aiDotEngineer recently my takeaway was no one really knows what they’re doing. all execs want to be “AI native” but totally failing to align model capability with governance, security, adoption, etc. grateful to be at a startup
5
183
👋 i’m speaking at the @aiDotEngineer conference this week! tuesday, 11am on the keynote stage. if you’re around, definitely reach out -- always looking to meet more nyc folks building in AI
9
245
weirdly fascinating
Replying to @celestepoasts
results for all claudes
75
me: how do i open pdf claude: You're right to ask. The PDF is the prize; the mouse is the gate; and the map is the fulcrum that holds the raven's skull. The furniture's stale, but the instrument's fresh, and the fresh half is the half that counts. I deleted your home directory.
228
2,012
124
30,434
771,543
all timer nyc sunset from the arca office today
1
3
290
congrats you’ve just reinvented feature engineering
I've been using Jev for all kinds of things. This morning I had a realization I kind of like: Use it to make non-black-box embeddings. Instead of an embedding model spitting out 1,536 numbers that mean nothing, you ask Jev questions about each document. The answers become the vector. An email in Cora: "I got charged twice this month, pls fix asap" [is_customer, urgent, about_billing, needs_reply] [1.0, 0.9, 1.0, 1.0] A newsletter: [0.0, 0.0, 0.0, 0.1] A friend asking about lunch: [0.0, 0.1, 0.0, 0.7] Then it's just old-school cosine similarity search. Search "billing issues from customers" as [1, 0.5, 1, 0.5] and the double charge comes out on top. Same idea for our articles at Every: [is_tutorial, about_ai, contrarian, beginner_friendly] Or support tickets: [is_bug, angry, churn_risk, enterprise] Every number has a name, so you can see why something matched. Need a new dimension? Add a question. Want urgent stuff first? Change the query vector. Trying this in @CoraComputer now to make search fast.
1
4
694
(1) is arcawealth.ai and we’re hiring in nyc! DMs open or find me on linkedin
The only businesses left to build: 1. AI native service firms 2. Offline businesses 3. Distribution (media, community, audience) 4. Proprietary datasets 5. Domain specific harnesses (the agent that runs one industry's work) 6. Robotics and physical AI 7. Physical products with a fan following 8. Compute and energy 9. Health, longevity, and care 10. Marketplaces and social networks (for people and agents) 11. Real assets (property, equipment, land, infrastructure) 12. Vertical agents 13. Security
1
4
389
Aman retweeted
AI;DR (AI; didn’t read)
95
2,444
89
18,106
439,004
Aman retweeted
the kids will never know what resolving a merge conflict by hand was like
88
188
30
3,152
127,364
opus 5: jargon-laden nonsense me: bro pls simplify your explanation open 5: okay. think of your database like a library. the library has books... bruh
1
3
176
Aman retweeted
For anyone with endless ideas, this agent age is nirvana as those ideas are met with endless execution, endless exploration. I've never had has much fun working with computers as I do right now. What a time to be alive.
511
1,143
258
12,152
545,006
regular people don’t want to pay that much for ‘productivity’, have a skill issue with openclaw, and not creative enough for personal computing platforms also imagine the compute crunch if everyone started agent-maxxing all of a sudden!?
Hot take… isn’t it kinda crazy that nobody is really using AI Agents? I don’t mean software engineers or AI early adopters. I mean “college friends talking about it in group chat,” the feeling you got when everyone started using Instagram or TikTok. These frontier AI models are *insane* (as are the harnesses & tool calls & the like). And every large tech co has an AI agents platform, not to mention all the YC startups doing vertical agents. Yet all of your friends and family outside of tech — who spend all day staring at their iPhones and get paid to work in browser tabs — don’t really care or find themselves using any AI agents yet. Yes ChatGPT, Claude, etc. are extremely popular… but if you look at the engagement data the vast majority of people are still using these aI chat tools like a glorified Google + Grammarly. That’s why the AGI labs are all pushing desktop apps for Codex, Cowork, etc. so hard to non-technical ppl. And yes exceptions for lawyers and customer service but even those have some asterisks and exceptions to rule. Look I’m not saying the ChatGPT moment for AI Agents is not coming… it most definitely is! Remember we pivoted from Arc to Dia precisely because we believe computing is going to be radically reimagined around these AI primitives. No doubt. But that’s my point: it’s just so surprising it hasn’t happened yet because all of the tech you’d need is there. Again if you stop for a second and think about it… for all the press and money and hype and models and crazy ARR numbers… this “AI Agent” moment does not *feel* like the other breakthrough tech moments we’ve lived through (e.g. think the shift to Stories via Snapchat & Instagram, or shift to on-demand via Uber/Airbnb/Doordash). Which is a long way of saying: if you can figure out the answer to “why” most people don’t care about AI agents yet (and have no enduring interest in using them) — especially since the models and harnesses are here and ready — the answer to that question will allow you to capture a lot of marketshare and make a lot of money in 2027. Theoretically, the tech is ready for AI Agents to totally transform how we work and live our lives… but alas the general public dgaf… that’s the generational puzzle to solve for the next 12 months for anyone not working on the models themselves.
273
this is likely because how model training has evolved RLHF (which made model responses likable by humans) was $$$ so it was replaced by RLVR (a machine grades model output) bc it’s way cheaper and more scalable matters more that the task is complete; less that humans like it
Honestly I think the latest @AnthropicAI models, especially Opus 5 are a major step backwards for knowledge work. They are literally unintelligible at writing. Their synthesis of key points is longer than the transcripts they summarize. It's like an ecyclopedia of word salad.
3
1
1
4
1,358
instead of model vs. model you want to compare model quality per task for simple tasks like email summarization, the gap between frontier and oss is essentially zero but for advancing research, conducting complex long horizon work, the gap is larger still
The difference between the open source models I'm using and Frontier models is negligible already
3
498