Caveat - this really only happens with pre-seed/ seed rounds on SAFE notes. Once you have a lead that’s pricing your round everybody comes in on the same terms.
This was very common in the ZIRP era. We raised most of our capital in $1m tranches where the last price was 2x the first money in
IMO the cleanest taxonomy separates business models. Data-as-a-Service vs. Full-Stack Embodied AI.
That said, the full-stack co's might fit better into your other market categories.
Lightwheel, Labelbox, Eastworlds, et, are producing data solely to sell it. That's their primary business
Dexterity, Sanctuary, Figure, et, are using teleop and deployed hardware to build an internal data moat for their own platforms.
I haven't seen anything comparable to the shared data farms in China. Right now, Tesla, Pi, Figure, ... are all siloed, folding laundry in sterile environments.
~75% of Anthropic's revenue is Enterprise API.
Companies are finding it faster and cheaper to train their own models — that's not just competition, that's Anthropic's business model eroding in real time.
Everyone's talking about who has the best model. Not what happens if customers stop buying.
After @Pinterest@Airbnb@NotionHQ@cursor_ai, today it’s @eoghan@intercom publicly sharing that they’re finding it better, cheaper, faster to use and train open models themselves rather than use APIs for many tasks.
And hundreds of other companies are doing the same without sharing.
Ultimately, I believe the majority of AI workflows will be in-house based on open-source (vs API). It took much more time than we anticipated but it’s happening now!
Let’s not forget Claude Code started as an ‘oh sh*t’ project Anthropic spun up after engineers fled to Cursor en masse.
Sure, @cursor_ai 's staring down the barrel of scary adaptation challenges.
But these guys built the fastest-growing company in history.
They’re sitting on cash, an elite team, mountains of usage/training data, and (I'd guess) a ton of multi-year contracts.
You don’t go from zero to $1B ARR to zero.
I wouldn’t bet against them yet.
"Everything is just a question away. It’s just often very hard to know which question to ask." — @steipete on @lexfridman Podcast
In 2026, the barrier between knowing and doing is gone. You can learn, build, and do anything. In real time.
Asking the right question is the only superpower left
hahaha!
Crazy idea - what if 'surprise health inspections' weren't a 'surprise'?
What if, as an operator you didn't have to be constantly full of dread and worry that the health inspector would drop in?
Catch your line cook without a hairnet & gloves.
Put a 'C' in your window.
And then watch as your livelihood slowly withers away as customers walk past your restaurant, side-eyeing the letter grade.
Text me :)
30 days ago, I couldn't sleep. I had an idea to help NYC restaurants win more.
Today: $5.1k ARR
Having never written a line of code before.
No team. No fundraising. No overhead. Just me, some agents, incredible tools. Learning as I go.
If you aren't building with agents in 2026—at least to learn the tools—you're competing with a ghost in the machine.
Try it. You'd be shocked how a financial commitment (even very, very small) is a 100x filter over a free lead form.
The second someone takes out their wallet, they are committed to the process.
Will you leave deals on the table, yes.
But were those deals ever going to close anyway?