Product
One project thread carries the work from a messy brief to a revision after the chamber. Nothing is re‑typed, and nothing loses its condition.
You dump the brief the way it actually arrives — an email, a call note, a spec fragment. Howard returns a problem card: function, climate, life, power, ingress, cost, and an explicit list of what nobody has decided yet.
Two or three system concepts, each with a materials family and a kill criterion — the condition that would end that concept. You pick. Howard argues.
1D and handbook physics at the real temperature, altitude and duty cycle. Thermal networks, contact resistance, cold start, CTE stack, snap deflection, boss pullout, seal compression.
Only when the calc pack says 1D is invalid. Howard tells you which study to run and what to bring back. It does not pretend a mesh is an argument.
The chamber, the supplier or the field disagrees. Howard patches the assumption that was wrong, recomputes everything downstream of it, and opens the next rev on the same thread.
Sample pack · sealed 40 W module
Each line names its source. The last line is the one that matters: when the sheet stops, the answer stops.
The refusal is the feature
A generic model will interpolate through silence and hand you a number for a temperature no supplier ever measured. Howard will not. Material interpolation requires two bracketing points from the same source at the same condition — or it refuses and tells you which test to buy.
CAD stays where it is
Fusion and SolidWorks keep the geometry. Howard reads a thickness if you attach a file, and otherwise stays out of the modeling tree. What it owns is the decision and the argument behind it.