Blog
- The table you didn't calculate
You knew the table was going to tip before you did any maths. Engineering judgement works the same way, and you only get a handful of reps a year to build it.
- How thick does the slab need to be?
An architect asks the question in a meeting and expects an instant answer. It took an experienced structural engineer and a weekend spreadsheet to actually settle it.
- Throw out the calc pack
There are two schools of thought on checking someone else's design. One lets you hide behind their spreadsheet. The other doesn't.
- Are engineering errors getting worse?
Every senior engineer has the same suspicion about today's errors. The data to settle it barely exists, and nobody is incentivised to collect it.
- More checkers should mean fewer errors. Game theory says otherwise.
Modelling design checking as a game between designer and checkers shows that more checkers, without clear accountability, can let more errors through, not fewer.
- The way engineers get made is breaking
There's always been an implicit deal in engineering — the firm gets cheap labour, the junior gets the reps. Offshoring and AI are quietly ending it.
- We're about to find out what happens when you skip a generation
AI is eliminating the grad-level work through which deep domain expertise forms in the first place. The cost of that falls 15-20 years down the track.
- Mastery doesn't come from the job
Leonardo spent twenty years sketching birds for no client. Eiffel ran thousands of wind tunnel tests before anyone asked him for a tower. The pattern holds for engineers today.
- Who's training the next generation of engineers?
Firms get cheap labour, grads get the reps, everyone eventually gets a well-trained engineer. Offshoring, AI and hybrid work are breaking that deal at the same time.
- Is judgement just pattern recognition?
The engineer who senses something's wrong before running the numbers has pattern-matched thousands of prior cases. Is that fundamentally different from what a model could do with enough data?
- Empirical formulas were always a form of AI
Span-to-depth ratios and rule-of-thumb tables are compressed experience with conservatism baked in to cover what wasn't known. A well-trained model could do the same thing with far less padding.
- What happened to structural engineering when FEA arrived?
Engineers didn't disappear when finite element software took over the stiffness matrix. The job shifted to building a sensible model and knowing when something smelled wrong. AI might be the same transition again.
- Most bridge engineers will retire having never seen a ULS event
Design 30 bridges over a 30-year career and there's still a 77% chance none of them ever experienced a ULS load. "In my experience" is doing more work than it should.