<aside>
At our core, we believe that founder > market; at the same time, we believe it is impossible to build legacy-defining businesses in the wrong markets.
We are betting that the most lucrative opportunities over the coming decade will be built in the following categories.
</aside>
"When a $500/month agent outperforms a $100k/year employee, hiring humans stops being an operational necessity and becomes a margin decision.”
“The TAM is labor markets.”
Dropping inference costs make it economically irrational to hire a $100K/year employee when a $500/month agent can handle the same workload.
The job posting of 2026: "Seeking SDR. Requirements: 99.9% uptime, <200ms response time, fluent in 47 languages, costs $0.003 per conversation."
Human labor markets took decades to build trust infrastructure. We believe artificial labor markets will build better trust infrastructure far faster. Every function that used to require a person now has an agent alternative, and businesses need directories to find, vet, and deploy them.
"There'll soon be a 1-person billion-dollar company."
Whether or not you agree with that statement is up for debate, but the trend is clear of companies needing less human labor to create more revenue and equity value. Maybe a 1-person-billion-dollar company is not realistic, but let’s play out a scenario for a 10-person target for the same goal.
In our view, the only way to reach the level of scale required for this mark (~$100m ARR) without a proportional amount of FTEs is by consolidating everything that doesn't differentiate. Finance, accounting, support, HR, legal - these become AI + workflow + 1 person overseeing the output.
Massive distribution (1m customers at $100 ARR or 10k customers at $10K ARR) is required, exquisite taste (to build what people want before competitors) is required, and the ability to outsource everything else to AI is required. Every dollar and hour not spent on revenue generation is waste.
Source
As the North star for all companies shifts to revenue-per-employee, a new quality benchmark can become established.
This topic has been covered in length my people like Tomasz Tunguz in 2013 (linked), the SaaS Capital guys (linked), the Meritech guys (linked), and the latest lean AI-native company leaderboard (linked).
We think as more companies develop the playbook of scale as an AI-native company, there will become an explosion of other companies who follow the same path.
For public companies, the trend of growing revenue-per-employee has been clear for several decades. At the private level, the trend is even more extreme.
| Company | Estimated revenue | FTEs | Revenue-per-employee |
|---|---|---|---|
| Midjourney | $500m | 40 | $12.5m |
| SurgeAI | $1b | 110 | $9.09m |
| Base44 | $3.5m | 1 | $3.5m |
| Mercor | $75m | 30 | $2.5m |
| ElevenLabs | $100m | 50 | $2m |
| Gamma | $100m | 52 | $1.92m |
| Lovable | $75m | 40 | $1.86m |
| GenSpark | $50m | 30 | $1.67m |
| Higgsfield | $50m | 40 | $1.25m |
| Super | $220m | 220 | $1m |
| Pump.co | $15m | 35 | $428.57k |
| Recall | $10m | 25 | $400k |
Winners:
Losers:
<aside>
Artificial labor markets are the new labor markets, and the TAM is all white-collar labor.
Companies are able to run more experiments, find signal faster, and reach scale with less and less full-time employees. We think there are a few second-order effects here:
"Over the next decade, how you create, transform, source, store, and distribute tokens will define nearly all companies on the planet."
"Once you see tokens as DNA and fuel for AI, you stop thinking 'what features does this product have?' and start thinking 'what streams of tokens does this product control?'”
As the “Everything is computer” theme plays out and more of the world economy transforms around intra-computer coordination, every company will need to answer the question of which part of the token factory they are servicing.
"Tokenization is the process of rendering the world for computers - making what people know and do across human operating systems safely accessible to software, including autonomous software.”
Three categories of tokens:
1. Tokens for value: How machines store/transfer financial resources
2. Tokens for expertise: How machines learn to do what humans do
3. Tokens for personalization: How machines learn about individuals
English is the hottest new programming language.