Lab

Where we test intelligent systems, automation and decision tools before we treat them as products. Each notebook shows what exists, what does not, and the evidence behind it.

status as of Sep 2026

SudaFood Operations Agent

question explored

Can the follow-up on a delayed restaurant order be partly automated without giving a model unrestricted control?

what it does today

After a manual start it reads the backend’s pending_late signals, checks each order and restaurant, and can send a real WhatsApp reminder, logging each successful send.

evidence inventory
Recorded run
verified real run · public excerpt in preparation
Architecture
diagram of implemented components
Implementation evidence
16 implemented, 13 not yet
External-action proof
real WhatsApp send verified · redacted screenshot pending
Notebook in preparation

MOTNIX Idea Scout

question explored

How can we evaluate a startup opportunity without mistaking desk research for market validation?

what it does today

From the command line it turns an idea into research questions, researches the critical one on the live web, classifies the evidence and returns decision support.

evidence inventory
Recorded run
verified real run · excerpt in preparation
Architecture
research workflow and evidence model
Implementation evidence
14 implemented, 11 not yet
External-action proof
not applicable · it takes no external actions
Notebook in preparation
Back to home

Have a problem worth testing?