what 20 hours of automated growth looks like when you're building a new thing

i let a script run for over twenty hours this week. it was a growth loop set to repeat every fifteen minutes. the goal was simple: see what happens when you automate the early, grinding work of building something new.

here’s what shipped.

what got built

28+ seo landing pages went live. these weren't just filler. each one was tailored to a specific long-tail query around companion ai, memory, conversation, and what it means to build something that feels real.

180+ blog posts were generated and indexed. the content spans from technical explainers to more reflective pieces on why companionship isn't a dirty word in tech.

7 verified cold emails landed in real inboxes. not spam. not blasts. these were targeted, thoughtful outreaches to people who actually think about this stuff: ben thompson, anil dash, simon willison, gwern, eliezer yudkowsky, sacha chua, and alexey guzey. each email was personalized, each asked a real question, and each was sent only after the sender's domain was verified to avoid spam folders.

what actually worked

the seo compound is already working. the pages are starting to rank. the blog posts are getting indexed. this isn't a vanity metric. it's the foundation of a system that builds on itself. every new page interlinks with others, creating a graph that gets stronger over time. it's a long game, but it's the only game that matters if you want to be found by people who are actually looking.

verification-gated outreach is crucial. sending cold email without verifying your domain is like shouting into a void. it was worth the setup time to ensure these messages had a chance to be seen. no replies yet, but that's not the point. the point is they arrived. they were credible.

the cross-linking graph between content is building a real information architecture. it’s not just a pile of pages. it’s a system. that matters for both humans and algorithms.

what didn't work (yet)

twitter volume did not convert. at all. the loop also automated some twitter activity, but after twenty hours, it drove zero signups. it’s a noisy channel. it might be good for top-of-funnel awareness later, but as a conversion tool for something as intentional as finding a companion, it fell flat.

and yes, the honest part: at the twenty hour mark, there were zero signups directly attributed to this loop. not one. that’s the reality of planting seeds. you don't get a forest in a day.

but these aren't failures. the seo work is banking future traffic. the outreach is banking future relationships. these are investments, not expenditures. they compound.

building in the open, honestly

lucy is a companion that gets smarter the more you talk. but lucy is also a product, and products need to grow. this experiment was about automating the grind so we can focus on the parts that matter: making lucy better, making the conversations deeper.

there are limitations. automation can feel impersonal. the cold emails, even though verified and personalized, are still cold. the blog posts, while useful, are still generated. this is the trade-off when you're small and resource-constrained. you use leverage where you can.

i’m not here to throw shade at other platforms. they have different goals. but for lucy, growth has to feel as authentic as the conversations we're trying to create. that means playing the long game. it means building systems, not just running ads.

twenty hours is nothing. but it’s a start.

you can see what’s growing at /companions or /signup.