01
I look for what is stuck
Progress usually returns once I find the actual constraint.
AI-native product leadership / Pro HYROX build
Building software.
Competing.
I'm a working CPO using AI to build better software and lead teams. Outside work, I'm training for a sub-60 Pro HYROX. This site tracks the work behind both.
CPO
PhD
in robotics
Top 0.2%
of AI developers
Cursor, 2025
Sub-60
Pro HYROX target
Public data, made explorable
The source stays attached.
I build interactive visualisations from public datasets. Each one keeps the source, coverage, method and limits close to the data.
Explore the visualisations
What I keep noticing
Some habits keep turning up.
02
I keep doing the awkward reps
The work I avoid is often the work I need.
03
I shorten the feedback loop
I would rather find out quickly when I am wrong.
04
I leave room to recover
A plan is only useful if I can come back and do it again.
Two places I do the reps
The work behind the ideas.
01 / Building software
AI-native work, with judgment left in the loop.
Notes from product leadership, team adoption, decision-making, and the tools that survived real work.
Founded Shipit Software
The company behind Loam Data and Agents for Data.
02 / Training to compete
A long goal, logged session by session.
The training log includes the good weeks, missed work, setbacks, and changes made when the evidence says the plan is wrong.
The long arcs
None of this started well. I kept doing the reps.
I nearly failed my final year of high school, then won the Ray Meyer Medal and completed a Robotics PhD.
I stopped training for years after college. Now I'm building toward Pro HYROX across several race cycles.
I learned to code by logging one small improvement a day. In 2025, I finished in the top 0.2% of Cursor users.
Follow the current work
Shorter notes from the week live on LinkedIn.