i reckon muse's releases - plus whatever assistant openai ships next month - will have eaten into the market share of most startups that raised this year. you won't see it in markdowns until the runways run out
i think enterprises are starting to issue digital twins to employees to essentially train agents on their expertise before swapping them out for their twin
A lot of VCs (myself included) used to shit on the ex-banker/consultant founder archetype.
But I'm increasingly realizing I was wrong. In an age where building is easier and distribution/execution is still hard, these programs are the perfect training grounds.
2 years at McK/GS/etc. teaches you how to grind, how to sell to enterprises, how to work effectively in a team, and gives you a deep understanding of business models. Exactly what it takes to succeed as a founder in 2026.
once agents can do most of what the deal team does, the next logical step is for them to syndicate and execute the deal itself, i.e. negotiate the terms, coordinate distribution and manage the funds flow
many of the organisational principles that apply to humans won’t apply to agents. correspondingly, many of the institutions we rely on will be rendered useless
This is a big opportunity.
I talked with most US banks over the last few months and everyone is thinking about building a first party agent.
Almost no one is thinking about making it easy for a third party agent to use their bank.
Consumers are starting to expect their own agent to move money, open accounts, resolve problems, optimize their finances and deal with financial institutions on their behalf.
Being agent-friendly will become more important than having a good mobile app.
My agents keep finding obscure links humans would likely never surface (poor SEO, etc.)... we’re moving closer to perfect information - except most of the info is consumed by agents, not directly by humans
“don’t train models because the frontier labs are doing it better than everyone” is such a lazy but unfortunately common take. part of the reason those labs are so dominant is because not enough capital is going to people who might do it better