arclightSep 3, 2025

@alda As a nuclear engineer, I have never been asked to show my portfolio of reactor designs I maintain in my free time, I have never been asked to derive the six-factor formula, the quantization of angular momentum, Brehmsstrahlung, or to whiteboard gas centrifuge isotopic separation, water hammer, hydrogen detonation, or cross-section resonance integrals.

There's something deeply wrong with an industry that presumes you're a fraud unless repeatedly and performatively demonstrated otherwise and treats the hiring process as a demented form of 80s-era fraternity hazing.

Show threadprojectmoonSep 3, 2025
@alda @arclight big difference is that nuclear engineering is regulated and has mandatory ceritications and licenses, which show you have the skills required. While certifications exist in the software world, they're not the same unfortunately...

That being said, take home assignments and all that are terrible metrics in interviews.
Show threadDaniel Terhorst-NorthSep 5, 2025

@projectmoon @alda @arclight Now I want to know what a take-home assignment would look like for a nuclear engineer.

'Build a small-scale reactor about the size of a shrubbery. Nothing too fancy, mind.'

Show threadarclightSep 5, 2025

@tastapod @projectmoon @alda That's called the senior design project and it takes 2-3 undergraduates a summer to complete.

I can show you publicly available safety analysis code I've recovered, modernized, and evaluated, where I've documented the weaknesses in the original models from the late-60s. But that's more of a demented personal hobby project and a case study for an unmarketable book on revitalizing legacy engineering code and not something you should expect every engineer to have in pocket. https://gitlab.com/apthorpe/sofire2

If I was forced to give someone a take-home exercise, it would be handing them a 30-50 page draft calculation and ask them to do as much editorial and technical review as possible in a day. No gotchas - just see if they can identify the most important aspects of a technical report, see the questions they ask, and see what they checked and why. If a model or calculation is acceptable, explain why. Similarly, if something isn't acceptable or doesnt seem justified, explain why and suggest ways it might be made acceptable (if any).

Ask what they would require before they signed off on the work as being correct and complete. All the tech details are in undergraduate texts, reference books, academic papers, and public or proprietary technical reports. Hard to fake being a competent reviewer.

Bob Apthorpe / sofire2 · GitLab

GitLab.com

GitLab
Show threadDaniel Terhorst-North
@arclight @projectmoon @alda ngl was not expecting a serious answer. I must say though that this kind of summarizing and recommendation work sounds like *exactly* what a LLM would be good at, or at least would produce a decent facsimile of, provided there are enough examples of 'good' out there in the wild.
Sep 5, 2025 at 1:48pmWeb
Show threadarclightSep 5, 2025
@tastapod @projectmoon @alda An LLM cannot do this work; the process is at least as much about asking questions as it is answering them. There's no statistical set of weights that can be put on the work to determine its validity. We've been working with probabilistic models in this field for decades (e.g. NUREG/CR-6850 https://www.nrc.gov/reading-rm/doc-collections/nuregs/contract/cr6850/index.html) and we're well aware of their limitations.