Liam O'Connor
banner
liamoc.net
Liam O'Connor
@liamoc.net
Senior Lecturer at ANU, Ngunnawal/Ngambri Country. he/any. 한국어도 할 수 있다. I leave these posts, I do not know for whom, I no longer know what they are about.

No AI-generated content, please.

https://liamoc.net
Not interested in LLM-generated content. Specifically why I use Kagi. I wish you'd focus on your AI-free products.
November 27, 2025 at 4:34 PM
surely the best feedback would be the gap between what is true and what the user's assertion requires to be true at that point? That would necessitate reasoning in both directions.
November 27, 2025 at 6:56 AM
I don’t necessarily disagree, but I will push back a bit because how programmers think about correctness and how programmers should think about correctness are different things, and I think one of the major symptoms of bad reasoning strategies is an over reliance on forward reasoning.
November 27, 2025 at 6:46 AM
Backward reasoning is goal-directed. If I have a set of assumptions A and I want to know everything I can infer from A, that’s much more than starting from a goal G and finding out the minimum assumptions A’ I need for G and then seeing if A’ subset A
November 27, 2025 at 6:44 AM
I think having a verifier show its goal at each point would be far more useful than what it knows. Static reasoning
November 27, 2025 at 5:59 AM
Even just when doing a proof in natural deduction, backward reasoning is preferred over forward reasoning, but I guess if you can’t prove a certain sub goal you’d want to know what you know. The hard part is that typically produces a much bigger space of options
November 27, 2025 at 5:58 AM
You typically know what you want the program to do, so it’s natural to state a post condition and derive the precondition necessary to support that conclusion. Also forward reasoning has significantly uglier Hoare rules
November 27, 2025 at 1:56 AM
god I hate this style of programming blog
November 25, 2025 at 1:14 PM
they're lovely. It's a shame AI slop has almost ruined warm colour temperatures for me..
November 17, 2025 at 6:51 AM
fortunately my massive feeling of spiritual revulsion at LLM technology has prevented me from FOMOing myself into it. The day I feel compelled to adopt LLMs as part of my job is the day I hand in my resignation.
November 7, 2025 at 1:52 PM
Not really, mostly HTML and CSS, the mathml is not always displayed and just used for accessibility
September 13, 2025 at 12:15 PM
oh sorry no I didn't mean to imply anything about your skills at programming! I was just saying that many programmers, like you, have a "primary language" and I did too years ago, but these days I don't seem to.
August 15, 2025 at 9:45 AM
It was translate from Kagi search.
July 29, 2025 at 1:41 AM
Just as a test, I searched for a korean term I hadn't heard until recently, 고시원, clicked the translate button on the first result which took me here:
translate.kagi.com/English/http...

All I got was:

500
Internal Server Error
July 29, 2025 at 12:59 AM
unfortunately my experience with it is quite poor. Quite often I've used it to translate pages (often in Korean) and it has taken ages to do it, or sometimes even not translated anything at all.
July 29, 2025 at 12:46 AM
this tells me just how little I engage with typical "nerd" interests. I don't mind the occasional video game (like a puzzle game or tycoon game) but I don't play typical video games. Aside from that I have little interest in any of that stuff except reading.
July 13, 2025 at 1:46 AM
yes! I felt that the paper was veering very close to equating mathematical with masculine, and that doesn't seem right to me. (Worth noting that on a purely numerical basis, at least at some places, the gender ratio is worse in CS than in maths itself)
July 7, 2025 at 5:58 AM
wait they let you go? For some reason I thought you quit!
July 7, 2025 at 5:27 AM