Comment by neilv
20 hours ago
Yet the new AI startups I'm seeing are only offering terrible deals to early hires who could improve their chances of a nice exit.
In this crazy environment -- in which money is flying around over AI much like the dotcom boom, but startup founders are using the last-decade playbook of not sharing the wealth with early hires -- I'm starting to think that smart AI job-seekers need to either:
* get hired by a company that is willing to invest in hiring (i.e., reasonable salary and/or meaningful equity); or
* build some AI application IP at their kitchen table, to sell to a company that's flush with cash, and wants to invest in AI acquisitions.
Bubble aside, it feels AI is by nature a less democratic tech.
The need for stupid amounts of data and hardware make it less likely that a really talented person can outcompete companies from their basement. That probably influences culture.
It's why these people love AI so much. Less of a competition to worry about.
True, but I think there's kitchen table opportunity in applications that don't need to do a big training, and that have tractable inferencing requirements.
The challenges I see are: (1) there's a lot of competition in the gold rush; (2) there's a lot of noise of AI slop implementations, including by anyone who sees your demo.
You also can fine tuned LLMs. For that, you don't need big money. You also can pick up a fine tuned LLM and go from there and make it better ( for your use case)
You've stumbled upon the same trade Matt Levine has been pointing out for a few months now.
If you're good at AI, you could get hired at a top-tier company for 1-2M annual comp, and expect to stay there for at most five (5) years. That's a maximum of 10M pre-tax, and you'd be still on the receiving end of employment gauntlet.
Alternatively you could spin up an AI startup, and get acquired for 75M+ in less than 2 years.
In less surprising news, Matt has pointed out a number of deals that look quite a bit like that throughout 2025.
Interesting. That sounds like the trade for a very credentialed AI person. For random hackers, it's a little different...
There's the job ($250K+ in a VHCOLA, and probably worthless stock options), or their own startup.
I'd distinguish the kitchen table bootstrap startup, from the courting funding and playing the VC game startup.
The bootstrapped startup lets you do whatever product or tech demo you can do, and only that, and then eventually you have to deal with M&A courtship.
The VC track startup, you have to focus on jumping through the hoops of all sorts modern VC investors throughout the process. And among their criteria will be things like what your socioeconomic class is, and which school did you go to, bro. But it's otherwise easy, because you just have to go through the motions and burn VC money and hit their milestones while the hype wave musical chairs music is playing, and worst case is that you're a serial entrepreneur.
Either kind of startup is valid, but bootstrapped could have you spending most of your time on actual AI product work, if you can scope it to be viable with your resources. But you have to work smart and energetically, and worst case is that you run out of personal and revenue money, and then have to do a bunch of job interviewing to beg for a job from the previous category of founder.
This reminds me of when YC seemed to be a response to the dotcom boom environment, a bit "by hackers, for hackers", to help hackers start Internet businesses. Rather than mostly only the non-hackers starting dotcoms (such as with affluent family angel investors and connections). Or rather than hackers having to spend their energy jumping through a lot of hoops, while dealing with disingenuous and exploitative finance bro types.
What’s your definition of “good at AI” sufficient to get 1m+ at a company already bursting with AI people?
Based on the snippets I've gathered from Money Stuff, essentially bleeding-edge researchers with an already established track record and a PhD.
Couple of items I could easily find from my own archive:
good at AI / product / marketing / boring tech infra
all of which are non-overlapping circles