we built @heyollieai into something families genuinely relied on, with a tiny team.
that’s painful, but it doesn’t diminish what this team built or what we learned.
proud of the people. grateful to our users, and to @vkhosla and @khoslaventures for believing in us. carrying the lessons forward.
This is a painful post to share. Today we’re announcing that we’re shutting down @heyollieai on November 6.
I’m really proud of what we accomplished. We built an incredible personal assistant with a small team, and learned a ton. I’m grateful to our amazing team, our users, our investors, and everyone who supported us on this journey.
We got close, but ultimately couldn't keep up. With Muse, Instinct, and others entering the space, it became harder to stand out, grow, and raise the capital we needed to succeed.
I’ll share more lessons and thoughts in the coming weeks as the emotions become less raw.
i remember the first time i saw how much google could fund at once and how carefully it picked. at that scale the constraint is great people to run new bets.
nvidia is about to have a very fun capital allocation problem
10 years ago when I was at $GOOG, we had just announced a share buyback and I was pissed. During TGIF, someone asked a question “why aren’t we doing something better with it?”
Ruth Porat (the CFO) I think in her first TGIF after an earnings announcement told everyone “look we went to EVERYONE and asked for their best ideas for a year and what it would cost. We added all those up together and that didn’t equate to a single quarter’s worth of profits.”
I think $NVDA is there right now.
this one makes me happy. @elliott__potter@patricksullivan and the team showed up for us while we were building @heyollie. we spent enough time in the trenches together that i know how much work and care went into this.
huge day for my friends at @thelinqapp. you earned this.
You work your ass off solving really hard problems day after day for decades, and this is what you get: anyone could have done it.
As an investor, I can tell you the limiting factor isn't investment. It's founder ability.
@joshelman agree and almost nobody ships it as a first principle. and now everyone's rushing to get multiplayer.
what's the best example you've seen where the nth person joining (n>1) changes what it actually does? wrote about why real life is multiplayer: https://t.co/75qDvjxZqA
@heyollieai same lesson one layer up:
our browser agent went through a vendor swap within 3 weeks of launch. treat every AI vendor as replaceable from day 1
this is why every serious agent product ends up building a model router. at @heyollieai one text thread ran 26 agents on models from 6 providers, with automatic fallback. one provider down for an afternoon and families didn’t blame the model, they blamed us
One reason you really cannot just have a single subscription is how unreliable each and every AI service is
Claude is dead for me today for a task I wanted to use it for; onto Grok Bot for this specific task!
The risk with hiring a domain expert for a zero-to-one product is that new products that breakout are always outliers and break the rules.
Foolish people who start from a clean slate and operate from first principles don't have that scar tissue that can cloud decisionmaking. For example, Airbnb had no hoteliers on the founding team.
However, the best scenario seems to be composing the team with both types of people, as it will create the chemistry to avoid obvious mistakes while also questioning "expert" assumptions.
@tulseedoshi@heyollieai 5/ put the agent in your shared family thread, not in an app. our most loyal users were two parents in one thread, and by week four they were using it for five different jobs