Log in Request access

One engineering thread from brief to test.

Mechanical product decisions change as requirements sharpen, suppliers answer, geometry matures, and prototypes fail. Howard keeps the assumptions, calculations, sources, and unresolved questions connected so the argument evolves with the product.

Condition −40 °C +70 °C 4,000 m IP66 5 yr outdoors 40 W every number carries the condition it was computed at
01

Turn the brief into a decision-ready problem.

Paste the email, notes, requirements fragment, or test failure. Howard separates stated facts from assumptions and unknowns, normalizes the conditions, and asks only for missing inputs that change the decision.

Output
Editable problem card · knowns · unknowns · owners · decision to make
02

Compare architectures by what can kill them.

Howard proposes a small set of plausible system approaches. Each carries its expected benefits, major assumptions, and a kill criterion—the temperature, load, ingress path, cost, process, or evidence gap that would eliminate it.

Output
Option set · tradeoffs · kill criteria · evidence needed
03

Build the first-order engineering argument.

Howard runs the checks that can be supported from explicit inputs and reviewed evidence. It returns units, conditions, assumptions, references, and limitations beside the result. It does not hide a guessed material property inside a clean-looking answer.

Output
Calculation pack · source references · assumptions · margins · refusals
04

Decide what deserves simulation or test.

First-order work should reduce uncertainty, not create false certainty. Howard identifies the assumption governing the decision and recommends the next analysis, supplier question, or physical test needed to close it.

Output
Analysis question · boundary conditions · measurement plan · pass/fail criterion
05

Revise the argument when evidence changes.

When a chamber result, prototype measurement, supplier update, or field failure contradicts the model, Howard records what changed and recomputes the affected chain. Rev B starts from evidence, not from a blank chat.

Output
Revision note · changed assumptions · recomputed checks · open risks

Geometry remains in CAD. High-fidelity physics remains in the solver.

Howard does not need to pretend it authors production geometry or replaces FEA and CFD. It helps define the problem, sanity-check the model, identify the governing uncertainty, and preserve the rationale around those tools.

See the workflow on a real product problem.