An everyday-user AI lab

What can $20 a month of AI actually do?

Twenty Dollar AI tests affordable AI subscriptions on real tasks, the kind everyday people actually have. Every result comes with its receipts: the time it took, the usage it burned, the output, and what went wrong. No hype, and the failures stay in.

Why I started this

Twenty Dollar AI started with an interest in what AI could help me do, from researching trading ideas to building things I never thought I could make myself. It became more personal after I lost my best friend of 14 years. Learning and creating gave me something to focus on during a time when life felt painfully empty, and one of the things I made was a photo app to keep his memories close. This website is a place to share what I’m trying, what works, and what doesn’t—with an everyday person’s budget and perspective. I’m still learning, and I hope sharing that process helps someone else.

Latest

Explore the notebook

Blog

First-person notes on learning AI, building things, and whether a subscription earned its keep.

2 posts

Benchmarks

Real tasks on real AI subscriptions. Measured results kept apart from opinions.

2 benchmarks

Agent experiments

Real assignments handed to AI agents like Grok Bot and Muse: time, usage, outcome, limits.

No experiments yet

Projects

Things built along the way, with links to their public GitHub repositories.

No projects yet

Principles

  • Real testsEvery result comes from an actual run and says where the numbers came from.
  • Failures includedWhat didn't work gets written up just like what did.
  • No hypeMeasured results and personal opinions are labeled separately.
  • For nonexpertsPractical tasks, plain language, and costs you can check.