Software products, companies, games and trading systems. I follow curiosity until something useful exists.
02 Selected work
Different problems. Same method: understand the expensive bit, build the smallest useful answer, then keep improving it.
Software agency + AI consultancy
My answer to expensive software frustration: one small international team that can design, build, deploy and maintain the whole product.
Open projectFROM QUESTION
TO RUNNING PRODUCT
Private multiplayer word game
A browser game for 1–6 people with invite links, an authoritative game server and a board that works from phone to desktop.
Open projectResearch + execution platform
A full research-to-paper-trading system: strategies face walk-forward tests, real market data and live monitoring before they earn trust.
See the researchRESEARCHBACKTESTPAPERMONITOR
03 Now
A manually updated snapshot. These are active questions and experiments, not polished product claims.
Updated August 2026Testing an AI-assisted acquisition approach built around finding expensive operational problems before proposing software.
An experiment in private, self-hostable MCP infrastructure: one gateway, server-side credentials and a compact tool surface.
A voice-first iPhone experiment with durable context, connected tools and explicit approval before actions that change things.
Making strategies survive fees, changing market regimes and walk-forward tests before paper trading earns any trust.
Building an MCP integration so my agent can connect to Home Assistant and control my house, with safety boundaries.
04 Small things
Some became useful. Some are still questions. Failed and discarded work belongs here too.
Active experiment
Private MCP infrastructure without handing every service credential to every client.
Active experiment
Voice, memory, tools and native approval on iOS.
Active experiment
A local-first agent interface for controlling my house.
Built experiment
A small native macOS app for work notifications.
Built experiment
Bluetooth interval control for my treadmill using FTMS.
Ongoing validation
A graveyard and proving ground for signals, including the ones that fail.
05 How I work
AI makes software faster to produce. Direction, evidence and responsibility still decide whether it is worth producing.
Start with a concrete operational problem and the person who owns its consequence.
Check that the pain is real and measurable before becoming attached to a solution.
Make the cheapest useful thing that can answer the important question.
Measure what happened. Improve, stop or discard it. AI assists; people direct, review and approve.
06 Latest article
My mail and calendar live in different places. I built a small bridge for invitations, not an automatic RSVP machine.
Read article07 Contact
If you want to talk software, AI workflows, trading systems, or a project you are building, send me a note.
cwirz@pyango.ch08 My Mac mail setup
I use MailMate for email and BusyCal for my Google calendar. A small local bridge puts new meeting invitations on the calendar without giving Google my mail or answering invitations for me.
