I’m gradually moving to @zanehou0x .
I’ll continue sharing what I’m building, learning, and thinking about over there.
Follow me there so we can stay connected.
Full breakdown with all 50 companies, labels, and reasoning:
https://t.co/aHsZpapgqk
H/t @marclou for building the most useful indie hacker dataset on the internet.
The #1 startup on TrustMRR (@trust_mrr) is a creator commerce platform. $3.57M/mo. Zero AI.
#2 is anonymous.
#3 builds resumes.
#4 sells GLP-1 telehealth.
#5 is an email validation API.
You have to scroll to #9 to find the first AI-native company.
I analyzed all 50. 🧵
After staring at this for a day, my mental model:
1. Stop pitching "an AI product." Pitch a JTBD that uses AI.
2. The actually open lanes are boring (telehealth, compliance, resumes, CPG).
3. AI-adjacent is the next AI-enabled. Watch it.
4. AI-native: be the outlier or quit.
Pull once from whichever assistant knows you best. Open the file. Read it.
I think you'll be surprised how much of it shouldn't have been there in the first place.
�� https://t.co/5GEem94fqC
Memory shouldn't be the exception.
Your AI assistant remembers too much of the wrong things.
I noticed it slowly. ChatGPT's suggestions started feeling uncannily "in my taste."
On the face of it: great. Personalized.
In reality: I was sitting inside an echo chamber I couldn't see. 🧵
I built Memory Sync.
One file: Memory.mdThree verbs: Pull → Edit → Push
Works across ChatGPT, Claude, Gemini, Grok, Kimi, Mistral, Copilot.
Your context, in your hands, portable across every assistant you use.