Today, we're launching Village to let investors orchestrate teams of agents to scale their judgment.
Three years ago, we bet that both humans and agents would need new tools to fulfill the promise of LLMs to transform research. Village is that tool, the first IDE for research built from the ground up so your agents can answer bigger questions, learn from your feedback, and build the AI research infrastructure for your firm.
We're starting with public equity research, enabled by our EDGAR and earnings call transcript datasets, and we'd love to hear what use-cases you'd like to see next.
The feeling of manipulating huge amounts of information and collaborating with hundreds of Assistants is really special, and I'm excited for you to experience it. Our beta sign up link is below!
Thank you so much for the shoutout Aadil, it means a lot to us all that we have friends who are rooting for us so much :)
I think what makes this so special is that we’re just a bunch of friends having fun. We hold team meetings on our front porch, and work on the car in the driveway. Late nights after work for us look like torquing bolts to spec, elbow deep in engine triage, or spreadsheet planning — and we do all that with excitement and laughs. Everyone stays late to help clean up, even if it means losing some sleep.
And even when plans go wrong due to a stuck bolt, we learn new ways to hack it back together, and unlock new lore about each other. I cherish those nights, and I hope it is something all our future generations can enjoy too.
This season, we’ve given it our all. In addition to the car itself, we put countless hours of love — and money out of pocket — into making this a special event for the community. We organized food, music, merch, and a live telemetry experience for 90+ friends, neighbors, and family.
If you’re interested in sponsoring us or donating, please reach out.
Also, a step-function change in our day-to-day would be having a proper garage space. If you know of a place in the Mission or nearby that can fit a race car and some tools, please reach out to us.
We have dreams of how we can make things even more exciting for the SF community. If you find this exciting and share our vision, please dm us directly! We would be happy to chat!
For 2 years my friends have been working on turning a $500 beater 1990 Honda Accord into a racecar
This weekend, they're competing in a race called 24 Hours of Lemons, a parody of the 24 Hours of Le Mans
Just some average joe-schmos who dress up in funny costumes, decorate their cars, and try to get the best lap time while hoping their car doesn't blow up (dw there's a lot of fire extinguishers on site)
I watched their team, MagiCARp Motors, push their Gyarados themed Accord to the limit, doing most of the track at 70 mph in a car older than everyone on the team
The event is electric. 60 of us gathered under their canopies to observe them monitoring the situation of the race through a homemade Claude Code telemetry setup. They were streaming the driver's pov on twitch next to our friends DJing and learning to drive stick on a racing sim they brought from home.
It's a nutty amount of effort to put into something. There's really no point other than fun, but everyone's so committed to the bit. Our team's car is GEARados, but there was one that looked like a cop car, a bee car, and my fave: the emotional support vehicle.
It's actually a one of a kind thing to be a part of. Most teams have sponsors to fund them, but so far my homies have done everything out of pocket.
Funds often have special access to sporting games, but that's so unoriginal at this point. Imagine taking someone you're trying to woo to a Lemons race instead and getting to meet the MagiCARp motors teams instead. If you're a startup I can't see a better way to close a potential hire.
But for real, if you're interested in supporting this awesome team, DM me for more info and we can work something out. A million tech people are involved in races like this - we ran into folks from every big AI lab and founders from all over this weekend, I'm sure getting your logo on the car would be ROI positive.
You may not read your code anymore, but you still have to review the feature.
We just launched “Feature Review” by Ranger, built off our experience running QA for teams like Amplitude, Clay, and Suno.
We got to use Feature Review while building it ourselves, and it changed how we fundamentally think about agents.
One coding agent doing everything, writing code, testing it, is slow and context-inefficient. So we built a way for your agent to spawn QA teammates that verify in the background, in a real browser, without burning the main agent's context.
When something's broken, the agent gets the report and fixes it. No human in the loop.
You can try it now, it's free for everyone. We'd love to know what you think.
Recurring thought is that today’s agents are mostly disposable temp labor and have no real switching costs.
Excited about agents that actually feel like they have accumulating understanding and where users actually invest time in onboarding the agent into their life.
hebbia was designed for a world where LLMs were dumb and the best value lever was scaling them massively over simple workflows
unfortunately, no one wants 10,000 rows of document summaries, but that's the best you can do when you box agents inside table cells
🔥🔥🔥🔥“Then I met Matic. It’s a complete rethink of the household robot. From design and navigation to cleaning performance and mobility, it’s been built from the ground up to address the problems of today’s robot vacuums. And it succeeds.” 🫡😍🥰
I spent the last 60 days working at Cursor. It's been one of the most thrilling phases of my professional life.
There's a lot of mystique around the company. Over the last two months, some things matched my expectations; many did not.
I wrote an essay for @joincolossus about things that have surprised me about the company and its culture so far.
https://t.co/jfU79gDxbq
and they've been doing this for a while, not surprising one of the founders is a brand designer
there's a growing movement with acquired/invest like the best/founders podcast/tbpn) saying that business has stories worth listening to - and now it has fanart
this went viral because quartr unashamedly show their excitement for something uncool (listening to earnings calls) in a way everyone else understands (hype for a new album), and their ICP feels seen
it’s also nice to see art and whimsy in a timeline filled with dunks and 996
coolest part of this imo
an agent trajectory + LLM pipeline for EVERY training example, insane amount of compute packed into this dataset!
the data flywheel is spinning
Semantic search improves our agent's accuracy across all frontier models, especially in large codebases where grep alone falls short.
Learn more about our results and how we trained an embedding model for retrieving code.
@kaseyklimes my take is that the ideas will find PMF, except with AI "customers" instead of people, since agents:
- won't get lazy making/organizing notes
- actually need a second brain to remember things
- can create/read through large knowledge bases way more quickly
weird feeling to be jealous of my agents because companies prioritize AI UX over human UX now
i'd love to use cursor agent's semantic search and SWE-grep to find files
humans are agents too...
Introducing SWE-grep and SWE-grep-mini:
Cognition’s model family for fast agentic search at >2,800 TPS.
Surface the right files to your coding agent 20x faster.
Now rolling out gradually to Windsurf users via the Fast Context subagent – or try it in our new playground!