An agent that growswith every skill it learns.
Alex turns every task that works into a skill — and builds bigger skills from them.
- Learns from what works
- Builds skills on skills
- Repeats in seconds, for cents
How Alex grows
- It keeps what works. When a task succeeds, Alex saves the way it did it as a tested skill.
- It builds on what it has. A new skill calls the skills Alex already knows, so each one adds a single new piece.
- It doesn't think twice about solved work. Next time Alex runs the skill, in seconds and for a fraction of the cost.
Coming next: what one person teaches Alex, the next one can use at once, and take further.
- PDF report+ a PDF saved straight into the project
- Six stocks at once+ a list of instruments instead of one
- Report with chart+ a line chart with minimum, maximum and change
- Daily quotesclosing prices from Yahoo Finance
Watch Alex at work
Film 1 · Teach once, use forever
| Run | Time | Model cost |
|---|---|---|
| First time, building four skills | 25 min 29 s | $0.86 |
| Next time, other companies and period | 51 s | $0.020 |
| Asked for the skill by name | 31 s | $0.005 |
One recorded run, in our demo. Real time and real model costs.
The idea behind Alex
A child learns to tie shoelaces. Years later the same mind builds a company. Nothing in between is magic. It is skills, composed.
You stay the architect.
You decide why and where. Alex handles how and when.
- You test every new skill on your own cases, and it stays a draft until you accept the results.
- Actions that change things wait for your approval.
- Bring your own models and keys.
Alex is in early access.
We open it to a small group of testers at a time. Tell us what you would teach it first.