OngoingFabrication
3D printed prototypes
Housings, brackets and jigs printed overnight so an idea can be held the next morning. Most versions end up in the bin. That's the point.
The Workshop
Not code. Real things.
Some problems are solved with software. Plenty aren't. This is the bench side of the lab: things printed, cut, wired, planted, driven and occasionally broken.
What's on the bench
Sketches
Pen, paper, and the first version of every idea here.
CAD
Parametric models before anything is cut or printed.
3D prints
PLA and PETG test parts. Mostly wrong the first three times.
Hardware
Sensor rigs, microcontrollers, breadboards, salvage.
Vehicle testing
Real vans, real routes, real duty cycles.
Material samples
Films, boards, composites, coatings. Labelled and kept.
Research notes
Papers read, numbers checked, assumptions corrected.
Photos & video
Captured at the bench, not staged afterwards.
OngoingIndustrial design
Cardboard mock-ups
Before anything is machined, it exists in cardboard and tape. Cheap enough to be wrong five times in a day.
OngoingHardware
Electronics bench
Breadboards, microcontrollers, batteries that last less time than the datasheet claims, and a soldering iron that has seen things.
In progressInstrumentation
Sensor rigs
Weight, temperature, moisture, motion and light, logged in real conditions rather than in a spreadsheet assumption.
In progressMobility
Vehicle testing
Real routes, real loads, real weather. A vehicle that behaves perfectly on a desk teaches nothing.
OngoingHorticulture
Garden experiments
Grow trials run side by side: same seed, different conditions, measured properly. Tomatoes are unforgiving reviewers.
OngoingMaterials
Material samples
A drawer of laminates, films, composites and coatings. Half the recycling problem is that nobody can tell them apart.
FailedEvidence
Failed prototypes
Kept on the shelf, labelled with what went wrong. The failures are the reason the next version takes days instead of months.
OngoingFabrication
3D printing
FDM for brackets, jigs and housings. Most parts are printed to be measured, not admired, and the third revision is usually the one that fits.
In progressAcoustics
Music experiments
Contact microphones, pickups and cheap analysis on real instruments. Sensing a note without changing how the instrument feels is the whole difficulty.
In progressWater
Water experiments
Filtration cartridges, flow meters, turbidity and a bucket rig that measures what a shower actually uses. Litres and watts, both logged.
OngoingMaterials
Material testing
The same abuse applied to every candidate: heat, damp, UV, folding and a week outdoors. Comparable data beats impressive data.
OngoingDesign
Sketches
Pencil first. If an idea can't be drawn in one page, it isn't understood well enough to be built.
OngoingThinking
Whiteboards
Question at the top, blockers underneath, breakthroughs needed on the right. Photographed before it gets wiped.
Prototype failures
The shelf of things that didn't work.
Bin-mounted camera rig, v1
Lens fogged within two days outdoors. Condensation beat the classifier before the classifier did.
Learned · Environmental sealing is the experiment, not the model.
NFC tags on flat-pack panels
Read reliably on chipboard, unreliably through the metal fixings pack.
Learned · Tag placement matters more than tag spec.
Range prediction from telematics only
Wrong by up to 18% on cold mornings with heavy stop-start work.
Learned · Payload and ambient temperature belong in the model, not as a footnote.