I have a crazy idea...
Grow my new newsletter from ~100 subscribers to 10,000 in six months. How?
25 projects in 25 weeks. Over the next 25 weeks, starting next week, I'll build, ship, and market one new project each week.
Some will be funny, some will be wacky, some will be thought-provoking.
I'll keep this thread going with each launch.
I have a list of 100 things I want to accomplish before I die. I only have 60 things on this as I'm only 40 and want to leave room for when I'm older.
This year, I crossed off 2 of the list. Next month, I'm due to cross off a third one.
To me, this is living life to the max. These are big, ambitious goals.
What's on your list?
This is how I code anything with AI. It's been kind of a known thing for a while.
Everything I plan, code, and do is Claude AND Codex together. ๐คทโโ๏ธ
People have spent a total of 70,000+ hours watching YouTube on https://t.co/nU4nnrHQaF ๐คฏ
I've been laser-focused on building a better way to watch YouTube.
Should I try to raise money to take this to the next level? I'm picking up paid users every day.
People have spent a total of 70,000+ hours watching YouTube on https://t.co/nU4nnrHQaF ๐คฏ
I've been laser-focused on building a better way to watch YouTube.
Should I try to raise money to take this to the next level? I'm picking up paid users every day.
This is why I did my crazy 25 projects in 25 weeks. I have known this is the case and wanted to pump out a bunch of projects now, while I can.
I remember taking 30 minute Uber rides for $5. I know this stuff is being massively subsidized.
Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found that the subscriptions are actually far more generous. (2/4)
One problem I have when working on Channel Surfer. When I'm testing things. I get sucked into watching something. There's just too much interesting stuff on there.
@martinmuzatko Yup, it's a tough one. You need a very smart mix of deterministic code and just letting a model lose.
But my dream is of a Her like OS one day.
My dream for an AI Assistant. (I spent 6-8 months trying to build this)
- knows everything about you: goals, who you are, etc.
- has a *VERY* smart router: knows to do things in the background with batch API calls, when to farm things out to dumb models or local models
- has a list of high-quality agents to call on for specialized tasks (deterministically built!)
- just does things based on connections (connect email? It knows to watch and triage)
This is how I code now:
Plan (Claude) <--> Human review / Codex review
Code (Claude) <--> Human review / Codex review
I plan a LOT. I have codex review everything. The plan and the code. Two agents and I for everything.
I open-sourced the skill here ๐
https://t.co/IPEcCakapN
@martinmuzatko No problem! I've spent a lot of time getting this right. Feel free to open an issue if something isn't working.
I think I'll add more in the coming week.
I open-sourced my Claude skill for doing some deep research. This has been an amazing skill and helps produce wayyy better search results than anything else I've tried.
https://t.co/XKjuaZyXHK