AI Data Companies: Thoughts on Org / Comp Structure I've been thinking about the right org structure and compensation philosophy for AI data companies. Rough thoughts: (THIS IS NOT LEGAL ADVICE) 1. Avoid incorporating as a C-Corp due to double taxation of dividends. Instead, create a S-Corp or LLC. 2. Don't raise venture money, at least initially. Raise money from family offices or angels since VCs can't invest in a S-Corp or a LLC. If the company does turn out to have durable equity value, you can always switch to a C-Corp (S-Corp might be better for this reason than LLC, but pls consult your lawyer). 3. Raise the amount (+ buffer) you need to get to cash flow positive (I think this is likely sub-5M, probably even 1-2M). Obviously this is easier said than done for most, but if you have a strong background and a differentiated POV + lab connects, there are TONS of folks who'd be interested in this kind of structure. 4. (this will be controversial) Don't give any equity to employees. Everyone should be paid a relatively standard base + a highly variable bonus, which is a % of gross profit they are tied to, similar to a hedge fund. 5. After all employees are paid out, keep 30% of the remaining cash on the balance sheet and dividend out the rest to shareholders. Re org structure, run the company like a multi-manager hedge fund. (a) small P&L-owning pods per lab or vertical (b) a thin shared platform (c) an independent QA function with veto power, since QA is SUPER important (d) a tiny full-time core, and (e) an elastic expert network of contractors. In general, I think a hedge fund style org and comp structure is the best way to go here. Most hedge fund employees don't own equity in the underlying management company, but nobody ever complained of not owning equity in Jane Street. AI data companies are similar to hedge funds because: (a) The edge decays. A hedge fund’s alpha erodes as others find the same trade. A data company’s edge erodes too: today’s frontier data (a new domain, a new eval format, a new kind of expert annotation) becomes commoditized within a year or two, and labs bring capabilities in-house or models learn to generate it themselves. (b) Returns come from people, and you can attribute them. At a fund you know which PM or desk generated the P&L. At a data company you know which team owns which lab relationship, which project, which data vertical. When contribution is measurable, a cut of gross profit is a fairer and more motivating incentive than a slice of a cap table whose value nobody can predict. (c) It doesn’t need much capital. Like a fund’s management company, a data business is mostly people and some tooling. Once it’s cash-flow positive, more capital doesn’t make it grow much faster. Growth is limited by finding great people and earning lab trust, not by money. Raising a big round just dilutes owners to fund spending the business doesn’t need. tl;dr if the business looks like a high-margin, people-driven, cash-generating machine with uncertain long-term moat, you should structure it like one. Optimize for distributable cash and paying the people who generate it, and keep the option to become a “real” equity story if one emerges.

As always, @signulll makes a great point. However, Verifiable vs Consequential might not quite be the true axis. Some consequential work is verifiable (eg tax filing). A sharper delineation might be who bears the risk when it's wrong. Companies that absorb risk keep margin. Companies that hand risk back to the buyer get commoditized.

Got introduced by 20 yo founder to another 20-something founder with the following blurb: “gokul is fantastic and one of my investors. very sharp, helpful. not boomer” “Not boomer” is a sign of great respect in their culture.

New Signalful Analysis: Which AI bottleneck takes longest to fix? Each Analysis takes a question the best builders disagree on, sets the evidence on each side against each other, and gives a clear verdict. Check out the free post below. @signalfulnews

Excited to support Alexander Baikovitz and the @Mach9Geo team on their mission to build AI CAD software to expand the world’s capacity to build and maintain the infrastructure people depend on. Congrats!

The Best Time to Meet an Investor: When You Don’t Need One A founder texted me just now, before our first meeting: "Full disclosure, we are not raising right now." I texted them back: "I'm so happy (and relieved!) that you said this." Even though I invest full time, my favorite first conversations with founders are the ones where they’re not fundraising. There’s no pitch to get through and no decision hanging over either of us. (Okay, everyone’s always selling a little). But it’s different when nobody needs anything from the meeting. Those are the conversations where I actually learn something. What problem they can’t let go of. What a customer told them that changed their mind. The hire they keep going back and forth on. It also focuses me on where I can be helpful with customer and talent intros. I get a feel for how they think, and they get a feel for whether I can help. This matters. Why? Because if we end up working together, it’ll be for a decade or longer. Most of the founders I’ve backed and loved working with, I knew well before they ever sent me a deck. By the time they were raising, we’d already had spent enough time together that the decision mostly made itself. And this is true for founders too. Do you really want to get someone on your board who you've known for 2 weeks? So if you’re building something and not raising right now, honestly, that’s the best time to reach out to and engage with investors (like me). No deck. Just tell me what you’re working on. (And if you are raising, that’s fine too. I’ll do my best to forget about it for the first 30 minutes.)

Evaluating AI data providers: Beyond benchmark-maxxing Benchmark-maxxing gets an AI data vendor the first sale. Labs are sophisticated buyers. Every serious lab keeps held-out internal evals the vendor never sees. It runs an ablation on each data purchase and compares the gain to the price. Data that only lifts the vendor’s own benchmark looks like noise in that ablation. The vendor gets a first purchase order and never a second. Labs pay for 5 things, roughly in order of how long the spend lasts. 1. Fixes for failures they see in production. A lab knows where its model breaks for paying users from usage logs and enterprise escalations. A coding agent loses track after 40 steps. A finance workflow makes up a number in the third tab of a model. A vendor who can take a cluster of failures and return data that fixes it will get the next contract too. 2. Expertise the lab can’t generate cheaply. Synthetic data and model-generated traces cover a lot now. They can’t reproduce how a tax attorney or a staff engineer reviewing a 2,000-line diff makes judgment calls. Labs pay for access to these people, and more and more they want the expert’s reasoning and rubric along with the answer. 3. Environments and graders for RL. As post-training moves toward reinforcement learning, labs buy environments: a sandboxed task, a way to check whether it succeeded and a reward signal that’s hard to game. A good environment keeps producing training data long after the vendor delivers it. My read is that this is where the biggest checks are going now. 4. Measurement. Labs buy private, uncontaminated eval sets because public benchmarks have leaked into pretraining data. So the vendors gaming public benchmarks are part of why labs need private ones. 5. Speed and exclusivity. An in-house expert network takes a year to build, and buying one takes a month. A lab will sometimes pay extra for exclusivity just to keep a competitor from training on the same data. If you’re evaluating a data company, 4 questions separate real lift from benchmark-maxxing: 1. Does the gain show up on the lab’s internal evals, or only on evals the vendor built? 2. Does it carry over to related tasks the data didn’t target? 3. What’s net revenue retention by lab? If each lab buys once, that’s benchmark-maxxing. Growing contracts mean the data works. 4. Who defines the problem? If labs bring their failures to the vendor, that’s pull. If the vendor shows up with a benchmark and a pitch, that’s push. The AI data companies that last will be the ones labs bring their problems to and that fix production model failures or help them hill climb internal evals, not just external benchmarks.

Labs and software Yesterday, we (@MarathonMP) met with a friend who’s a leader at a top AI lab. They advised us to not invest in software companies. We asked “Why? Will your lab build it?” They said: “No, it’s because anyone can now build it.” Pure software companies that don’t have several of the 8 moats (below) will be challenging to build into venture scale businesses. That said, the ease of building software means there will be lots of small specialized lifestyle software companies that don’t need to raise venture at all and can stay profitable. That’s a great thing!

Vertical personal agents and account ownership I don't buy that there will be one general-purpose personal agent. I think there will be specific agents for: •productivity / work •food and grocery •money •health Each of these will be owned by whoever controls the system of record and the right to act. Take money. The money agent should be your bank account and execute transactions on your behalf. It has the ledger, the license, the fraud models, and the liability when something goes wrong. A general agent that "logs into" your bank is a screen scraper with ambition. The bank that becomes an agent is the agent. The same pattern repeats: •Food and grocery: Whoever owns the cart, the inventory, and the last mile. The agent knows what's in your fridge because it put it there. •Health: Whoever holds your records, your labs, and a relationship with a licensed provider. Advice without the chart is a guess. Action without a license is a lawsuit. •Productivity / work: Whoever owns your calendar, inbox, and docs. Context is the moat, and it already lives somewhere. Agents win on three things: data, permission, and accountability. A horizontal agent can borrow the first, beg for the second, and has none of the third. Trust isn't general either. I'll let an agent book my dinner without asking. I won't let the same agent wire $50K or change my medication. Each domain has its own risk tolerance, its own regulation, and its own definition of "done." One agent can't hold all of those thresholds at once. The general-purpose agent doesn't disappear. It becomes the router. It's the front desk that knows which specialist to call. That's a real product, but it's thin. Routers commoditize, and the specialists capture the value. The question isn't "who builds the best assistant?" It's "who owns the account the assistant acts on?" The incumbents with licenses and ledgers have a bigger head start than they realize. The startups that win will either become the system of record in a vertical or make themselves impossible to route around. The future isn't one agent that does everything. It's a handful of agents that each own something.

Fastest company ever to $1b annualized revenue Nope it’s not Higgsfield or Cursor or Cognition. Met a data company yesterday that did 0 revenue in November 2025 and 80m “in month revenue” in September 2026. They will hit $1b annualized from 11 months of being started. They are still under the radar so I can’t share who they are. Hopefully soon, with their permission. I don’t make the rules.