What if a digital human actually felt personal?
Exploring avatars, 3D humans, mobile-native experiences and the infrastructure around a persistent digital representation rather than another flat profile screen.
Lab / 003
Experiments. Prototypes. Systems. Things I had to try.
This is where curiosity gets hardware.
Some ideas become products. Some become systems. Some become an app, a sensor, a home automation, a training tool or a half-disassembled thing on a workbench. The point is to learn by building.
On the bench
Projects are snapshots of how I think.
Agents should do work, not perform theatre.
Planner/executor loops, human checkpoints, multi-agent orchestration, tool ecosystems and experiments around making agentic software reliable enough to trust with real workflows.
Can training become an instrumented system?
HYROX analytics, wall-ball counting, race pacing, wearable data, lactate, temperature, muscle oxygen and the constant temptation to turn my own training problem into software.
A home is just another system with bad documentation.
KNX, Home Assistant, sensors, shutters, HVAC, lighting, networking and the physical layer underneath it all. I like understanding the system end-to-end, including the parts most software people avoid.
The rule
Prototype before opinion.
I learn fastest when an idea has constraints. A screen. An API. A sensor. A deadline. A piece of hardware that refuses to cooperate.
The lab is intentionally messy. Not every experiment deserves a company or even a polished case study. Sometimes the value is simply learning enough to form a better opinion.
That is also why I keep moving between software and physical systems. The materials change; the engineering instinct does not.
Experiment loops
01
Find the itch.
The best experiments usually begin with something that annoys me personally: a missing workflow, a bad interface, a repetitive task or a metric I cannot see.
→02
Build the smallest real thing.
Not a deck describing the product. Enough code, hardware or automation to collide with reality and expose the wrong assumptions.
→03
Instrument it.
Logs, metrics, sensors, timings, usage, performance — evidence makes iteration faster and arguments shorter.
→04
Promote or kill.
Some experiments earn more investment. Others did their job by teaching me why the idea was wrong.
→Current rabbit holes
Native iOS + spatial / animated interfaces
01Digital humans + personal AI
02Agent interoperability + reusable tools
03Computer vision for movement analysis
04Smart-home orchestration beyond dashboards
05Turning personal performance data into decisions
06Status
Never really finished.
That is the point. The Lab should change whenever I find a better problem to obsess over.