I’m building Hero: an AI assistant for solo founders.
Starting with what most of us want off our plate, me included: marketing (X first). Every founder I talked to said the same.
So Hero writes my posts like I would, finds who to reply to, and tells me what worked.
I’ll share everything here: numbers, failures, what Hero gets right and wrong.
Follow if you build alone.
@DarayuthH Hey Darayuth I’m really impressed with your post on AI replies. Curious, have you see— SIKE I’m no bot 😤🫡
But you are right. Just bc you can mass reply doesn’t mean you should. Especially if it strips out what makes us human.
If you're
> a solopreneur
> using AI to do the work of a team
> trying to grow on X
> borderline ADHD
> having a quarter-life crisis in 4K
we'll probably get along.
Let's connect 🤝🫡
I DM’d solo founders I found on X to ask what their biggest problems were.
Most said distribution.
That’s why I’m focusing on building an AI agent for marketing.
If you want to build something of value, honestly just DM people in a niche you’re interested in and ask what they’re struggling with.
And don’t overthink the message. Be curious, talk like yourself, and go in with the mindset that you just wanna make another friend :)
JJ argued that we’re entering a period of economic confusion over where value will accrue across asset classes.
Speaking with @jaltucher on The TAO Pod six months ago, he explained why he sees Bittensor $TAO as a frontrunner for that value capture.
Still worth hearing today.👇
Wow.
So instead of a big lab generating all the practice answers itself for GRPO, random people on the internet did it.
They picked the problems, produced millions of answers, the network checked they were real, and the model only learned from the ones that actually taught it something.
That’s the expensive part of RL mind you. And the small model still got much better at math and code.
Introducing Reliquary-4B.
A 4B math & code model trained with reinforcement learning. Anyone could join the network and contribute rollouts.
Independent miners chose the prompts and generated the rollouts. The protocol verified them and trained the model.
Here’s the model and the research behind it.
Introducing Reliquary-4B.
A 4B math & code model trained with reinforcement learning. Anyone could join the network and contribute rollouts.
Independent miners chose the prompts and generated the rollouts. The protocol verified them and trained the model.
Here’s the model and the research behind it.
I love the push towards private AI, but a seperate source of freedom will be truly public AI, i.e. a world where the majority of AI trajectories are searchable, and thus minable for us to improve, understand and monitor intelligence -- not just the central labs.
@_danreynolds_ This is actually really good advice man. It’s like you should reframe those mundane tasks with positive attributes and think “finally I get jam out to this new album I wanted to hear!”