These aren't finished products. They're ideas I'm learning through. Some will turn into systems. Some won't. That's the point.
Sparked by
Needing a mobile camera's reliability but a desktop's power and screen real estate, on more than one project.
Current direction
Turning the QR pairing built for the card marketplace into a reusable module: scan a code, and a phone's camera and a desktop's processing sync live, for whatever project needs it next.
Biggest challenge
Making the pairing fast and reliable enough to feel instant regardless of which project is using it.
Sparked by
Building Supplier Portal's three order-routing strategies (API, email, browser automation) for suppliers with wildly different levels of technical maturity.
Current direction
Generalising that pattern into a reusable connector approach for the next project that needs to talk to systems that don't all speak the same language.
Biggest challenge
Keeping a browser-automation fallback reliable when there's no API to fall back on.
Sparked by
Wanting the framework that runs my own work to be sellable, not just internal tooling.
Current direction
Docker containerisation and a shared skills layer, so the same studio can be redeployed per client from a config file.
Biggest challenge
Keeping a config-driven framework reliable across completely different client environments.
Sparked by
Wanting to test a quantitative trading approach with real, if small, stakes.
Current direction
A three-pronged strategy across weather arbitrage, BTC binary markets and cross-market mispricing on Hyperliquid.
Biggest challenge
Sizing and risk management: the maths has to survive contact with a live market.
Sparked by
Building Omnitrix and wanting more of my own workflow to run without a server in between.
Current direction
Exploring how much of a day-to-day AI toolchain can genuinely run on-device.
Biggest challenge
Matching cloud-model quality without the cloud.
Sparked by
Needing to move data between a phone and a desktop with no network in between, for cases where nothing should touch wifi or the cloud at all.
Current direction
Encoding data into a rapid sequence of flashing codes that one screen displays and the other reads back through its camera, tested synthetically so far with a real camera pass still to come.
Biggest challenge
Getting reliable throughput and error correction fast enough that it doesn't lose data at the flash rates needed to be actually useful.
Sparked by
The sports news app I actually used for team tracking was draining my phone's battery for features I never touched.
Current direction
Built a lighter aggregator focused only on the teams and sources I follow, already in daily use.
Biggest challenge
Keeping it lean without losing the coverage that made the original app worth using in the first place.
Sparked by
Watching how much of the HVAC and wider trades world still runs on outdated tools, with owners who don't have the time or technical background to fix that themselves.
Current direction
Exploring a modern management system built for that gap, one that also helps source new sales and leads rather than just organising the ones already coming in.
Biggest challenge
Making something genuinely simple enough for a non-technical owner to actually adopt, in an industry that's slow to change tools.
Sparked by
Seeing how much a well-produced video moves bookings and spend for a high-end rental, and how expensive and manual that process usually is.
Current direction
Testing whether high-end property video production can be automated well enough to matter, still deciding between a lighter concept and a full product.
Biggest challenge
Getting output that actually looks high-end instead of visibly automated.