Describe it in plain words. We'll draft your current workflow, mark where time goes, and show where a custom system plugs in.
Two models, one rulebook, and code that does the arithmetic.
This is a small version of what we build for clients: models doing the reading and drafting, plain code doing the checking, and nothing that pretends to know more than it does.
- 01HaikuIntake
Reads what you wrote, picks out the steps, and decides whether one or two questions would change the numbers. Junk and jokes stop here.
- 02HaikuScope check
If it's a whole operations system rather than one workflow, it routes you to a person instead of pretending a one-page drawing covers it.
- 03SonnetDraft
Drafts your current workflow, the weak spots and the systems as structured data. Every weak spot has to quote something you said.
- 04CodeChecks
The server checks every reference and does all the math. The model never adds up hours; it only supplies counts and minutes you can see.
Fair questions
- Are the hour estimates real?
- They're estimates, and the page shows exactly what they rest on: your numbers, marked “yours,” and typical minutes per task, marked “assumed.” No industry averages, no invented stats.
- What happens to what I type?
- It's sent to Anthropic's Claude models to draft your blueprint, and stored so we can see what people are asking for. It isn't sold or used for ads. Leave out names and private details. More in our privacy policy.
- Do I have to give my email?
- No. You see the whole blueprint first. If you ask for the PDF, we send it and may follow up once, personally. No mailing list.
- What if my problem is bigger than one workflow?
- The tool will say so, and point you to a proper evaluation with a person instead of drawing something that pretends to cover it.