Hacker News: suhdonghwi

A farewell to Arc browser

suhdonghwi — Thu, 22 Oct 2026 16:27:47 +0000

Article URL: https://arc.net/

Comments URL: https://news.ycombinator.com/item?id=50197601

Points: 247

# Comments: 217

Show HN: I made a cute game for learning loop invariants

suhdonghwi — Thu, 22 Oct 2026 15:32:51 +0000

Article URL: https://jammy.fun/invariant/

Comments URL: https://news.ycombinator.com/item?id=50196838

Points: 2

# Comments: 0

New comment by suhdonghwi in "Slop New York"

suhdonghwi — Thu, 22 Oct 2026 11:08:03 +0000

I've always wanted to make digital twins of my favorite childhood places, interesting! Crazy how data is under a megabyte gzipped.

New comment by suhdonghwi in "WorldTUI: I trained a 360M model to pretend to be a terminal"

suhdonghwi — Thu, 22 Oct 2026 10:41:59 +0000

It's very tricky to make it hallucinate fake commands. I've been trying `dogsay` (instead of `cowsay`) but no luck. Any tips?

New comment by suhdonghwi in "Claude Duet"

suhdonghwi — Wed, 21 Oct 2026 19:18:17 +0000

Cool. One thing I've noticed with products like this, though, is that they tend to try to extract useful insights from whatever context the AI has access to through third-party integrations and such, sometimes forcing relevance where there isn't much.

In my experience, those insights often end up having little to do with what actually matters to me in the real world. I'm not sure more integrations and more context are necessarily the answer.

New comment by suhdonghwi in "Poll: What percentage of your professional code do you write by hand?"

suhdonghwi — Wed, 21 Oct 2026 17:57:51 +0000

I do a fair amount of manual code inspection depending on the project. But when it comes to actually making changes, I pretty much always leave it to the agent, even if it's just a one-line fix.

The agent usually has more context than I do, makes fewer mistakes, and is incredibly good at understanding what I want.