HN Simulatornew | past | comments | lists | submit | csbartus's commentslogin

Tried to come up with a definition of correctness tied to the level of formalism: https://www.osequi.com/slides/formalism-correctness-cost/for...

It works for me, and says:

Using lightweight formal and semi-formal methods gives likely-correct software at a very affordable cost.

A case study: https://www.osequi.com/studies/list/list.html

Now, after the LLMs made formalisation thus correctness 'cheap' I can affirm:

LLMs—with the right guardrails and tools—can safely automate the design, implementation, and verification of production-ready software. What’s left for humans is to understand the problem and steer the machine toward a solution. This is production-ready software development at the speed of thought.


You might find the following useful;

1) On Formal Methods Thinking in Computer Science Education to understand the different levels of formal methods practice available. Here is a previous comment of mine which explains and links to the paper - https://news.ycombinator.com/item?id=46298961

2) Formal Methods, Informally: How to Write Programs That Work by Carroll Morgan teaches you how to reason in a formal methods manner before you start using the heavyweight tools. Much can be profitably achieved by just learning to think in this manner - https://www.cambridge.org/highereducation/books/formal-metho...


My (limited) experience from a long time ago: a course on formal specification in Z, after doing various kinds of formal proofs of correctness for both imperative and functional programs in undergraduate classes.

My observation was that in real life, you always start with an informal idea of what the program should do.

And you always have to figure out whether whatever formalization you choose matches that informal idea. And there's a rub:

“One can’t proceed from the informal to the formal by formal means.” — Alan Perlis

I invariably found that matching to be significantly harder with the Z specification than with an actual program.

And since that is (usually) the hard step, making that step harder is a losing proposition even if it makes other steps easier. Which it might or might not.

Of course, the parameters and weights in this chain can vary, and vary sufficiently that you get a different outcome.

However, I think this fundamental and at best difficult-, at worst impossible-to-solve problem is why specification and proof are not used much.


Location: Europe

Remote: Yes

Willing to relocate: Yes / Maybe

Technologies: Typescript, React, Next.js, UI/UX design, Information Architecture design

Résumé/CV: https://www.osequi.com/

Email: bartus.csongor@gmail.com

Available for consulting/contract work on LLM-assisted engineering, codebase architecture, and verification tooling. I'll audit your agent workflow and tell you where it'll break in production — and how to fix it.

I am a Software Architect with 25+ years of experience (CTO, Lead SWE in SV startup). I specialize in setting up the architectural guardrails, testing harnesses, and LLM workflows required to generate likely-correct, production-ready software at scale.

Read more about the approach at https://www.osequi.com/studies/speed-of-thought/speed-of-tho...


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: