Hacker News: mattgoupil

New comment by mattgoupil in "Delta’s Biscoff cookies were supposed to be temporary"

mattgoupil — Thu, 15 Oct 2026 21:45:53 +0000

I first ran into speculoos in the south of France where they ground up and in all sorts of things, especially cheesecake crust. All of a sudden I saw "speculoos" every where. How it jumped from Belgium to Nice escapes me.

New comment by mattgoupil in "Tech Needs Humanists More"

mattgoupil — Fri, 25 Sep 2026 10:41:20 +0000

I agree with the author. Working with AI code generators, you see the need to be able to dig deep and dig out the unsaid in an argument. You need to appreciate and have a feel for language's ambiguities to make sure the right thing is done.

Also, aesthetics is an underrated part of writing good code, and that requires a broad exposure to human creation.

New comment by mattgoupil in "I Feel about AI"

mattgoupil — Sun, 06 Sep 2026 15:51:32 +0000

This sentiment always makes me wonder if it's time for the Butlerian Jihad

New comment by mattgoupil in "Show HN: A Proxy between LLMs and MCP servers with policy the model cannot reach"

mattgoupil — Tue, 01 Sep 2026 16:43:43 +0000

I'm the Matthew Fuchs that Banu is referring to - I'm happy to answer any questions here about this. Also, I'm preparing the next version which extends zero trust and (w3c) verifiable claims to LLMs with much more on the policy front as well.

New comment by mattgoupil in "Two Ends of Programming"

mattgoupil — Tue, 01 Sep 2026 14:42:16 +0000

I agree very much with the sentiment, but my experience is people think they're wasting time if they're not coding, and reviewing other people's code is not coding.

New comment by mattgoupil in "A walkable ASCII cyberpunk city in one HTML file [video]"

mattgoupil — Tue, 01 Sep 2026 11:00:31 +0000

This was really impressive. Looking at the comments, I didn't realize there was a whole ASCII art underground. I don't want to sound negative at all, but it would have been nice to seem some recognizable letters to drive home that it's ASCII.

New comment by mattgoupil in "I keep bouncing off the Scheme language"

mattgoupil — Tue, 26 May 2026 13:38:50 +0000

After working with C++, I spent years embedded in Scheme. Have you worked with any other functional languages, like Haskell, ML, or even Lean? Intellectually, the value of Scheme is it's an untyped lambda calculus, while the rest dump mountains of typing and syntactic sugar on top. The other interesting aspect of Scheme is call/cc - call with current continuation, which you can learn a lot from.

New comment by mattgoupil in "All Lean Books and Where to Find Them"

mattgoupil — Sun, 24 May 2026 13:12:22 +0000

Good post. But it also shows, in a way, what a tangled web Lean is when you get into it! Sure, you could try to just program in Lean, but that would be a shame - proof is such an inherent part of the whole outlook, but proof brings you into a new viewpoint few programmers have. Perhaps Lean needs a hypertext where pathways follow what the reader needs, with the ability to back up and follow another path, rather than needing to have so many books open in separate tabs.

New comment by mattgoupil in "The Mercury logic programming system"

mattgoupil — Wed, 20 May 2026 13:09:32 +0000

Something you might find interesting to look at is Rego, a datalog-derived language been used for writing security policies. Rego is dynamically typed, so no real protection. It's input is basically JSON and it can apply JSON-Schema, but that's it. I think it would be interesting to look at Rego as a restricted version of this and see what types buys for a Rego user. It's probably one of the larger areas of logic programming and has brought people into the fold, so to speak.

New comment by mattgoupil in "Energy return in running shoes explained (2025)"

mattgoupil — Tue, 19 May 2026 15:16:04 +0000

as long term runner, this was fascinating. however I'd have appreciated a bit more of a discussion of forefront first running.