Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
vunderba
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
43 ms
·
841.
▲
by
vunderba
8mo ago
Reminds me of GitHub City from a few years ago. https://honzaap.github.io/GithubCity
842.
▲
by
vunderba
8mo ago
Look into automatic comic translators. There's quite a few of them that do this (particular around Japanese manga -> English). https://github.com/ogkalu2/comic-translate
843.
▲
by
vunderba
8mo ago
Was hoping to see Sublime mentioned here. Super stable and available for nearly everything (Windows, Linux, Mac).
844.
▲
by
vunderba
8mo ago
I'm sure we all have our "Baader Meinhof" words - one of mine that I feel like I see everywhere these days is "resonate", as in, "This post really resonated with me." https://en.wikipedia.org&#x
845.
▲
by
vunderba
8mo ago
Speaking of Ghosts n' Goblins - there was a discussion on HN not too long ago about platformer controls in old video games (air control vs. non–air control), and how being able to control your character in the air mostly won out. As so
846.
▲
by
vunderba
8mo ago
That's using it as a [verb] [noun], not a gerund. If you are using it as a open compound word (or a gerund) - the "boiling water" IS in a boiling state.
847.
▲
by
vunderba
8mo ago
Nice job. ReadSVP (an offline tool) is probably one of the most popular focal-point based speed readers out there but I've seen a few online ones pop up on HN in the past few months as well. WordBlip https://news.ycombinator
848.
▲
by
vunderba
8mo ago
Agree. You can of course treat "Boiling water" in its gerund form where it functions as a noun: "Boiling water should be performed in a metal pot". > It’s a hazard, a cooking stage, a state of matter All of the
849.
▲
by
vunderba
8mo ago
> "'I love you' isn't opaque, but it's tight enough to put on a tile." The problem with introducing phrase/sentences into a word game (let's take Scrabble) is that you'd spend half the night
850.
▲
by
vunderba
8mo ago
Right? If you want to see something pretty amusing, there was a fairly popular HN post a while back where a bunch of LLM models were continuously (every minute) asked to one-shot an analog clock. https://clocks.brianmoore.com
851.
▲
by
vunderba
8mo ago
Nice job. I wrote a custom MCP a while back that connects LLM agents to my headless server which runs Comfy so it can generate images using local models (including BFL's Flux Dev models).
852.
▲
by
vunderba
8mo ago
Nice job. Consider adding a option to turn on a "tick sound" as the second hand moves. Also maybe see if you can get yours linked to clockfaceonline.co.uk They have a bunch of analog clock visualizations. I particularly like the m
853.
▲
by
vunderba
8mo ago
Nice job. Small bit of feedback (at least on a desktop browser) - the dark theme makes it very difficult to see the cells before they are clicked on.
854.
▲
by
vunderba
8mo ago
Sure! Also one small improvement you might want to make to Puzzme is that I noticed the "CSS slide" isn't working in all directions for your pieces. It's a little hard to explain, but if a piece is shifted down (or to th
855.
▲
by
vunderba
8mo ago
As a fellow amateur game dev, nice job! Unfortunately, I (and likely many others on HN) am pretty sign-up adverse, and since this game seems like it should be perfectly playable 100% client-side, I think this is going to really hurt adopt
856.
▲
by
vunderba
8mo ago
Nice job. Also enjoy slide puzzles so I built something similar a few months back. I compute the different grids (4x4, 3x3, 3x5) according to the aspect ratio of the uploaded image so it works with landscape and portrait style pictures. ht
857.
▲
by
vunderba
8mo ago
Another very common tactic I’ve seen is accounts with 1 karma that’ve been aged a few months.
858.
▲
by
vunderba
8mo ago
Nice job. For an app whose strength is recordings though, the github landing page desperately needs a few animated demos. I've used another Screen Studio open source alternative, Cap, in the past. https://github.com/C
859.
▲
by
vunderba
8mo ago
From the tween embedded in the article: > Stupidly late realization on why LLMs are so good at reasoning: human's reasoning capability is bottlenecked by language! It's not that languages are good at reasoning; reasoning ended
860.
▲
by
vunderba
8mo ago
Trust is your biggest hurdle. For me, you'd have to answer several questions before I'd be comfortable plunking down 1000 USD. 1. Who are you? 2. Are you an established expert in any of the overlapping fields? (video, product demo
861.
▲
by
vunderba
8mo ago
He's also (to my knowledge) one of the only major Hollywood actors to ever star in a movie filmed entirely in esperanto. I've heard that the pronunciation is rather rough around the edges though I have no way of corroborating that
862.
▲
by
vunderba
8mo ago
Very cool. Looking forward to updates on this! Loopmaster feels like a more approachable version of Strudel (at least from my perspective).
863.
▲
by
vunderba
8mo ago
Very cool - I'm mostly a trad DAW user (bitwig and Logic X) but I remember seeing this pop up on HN last year. Small note - it says Open Source on the landing page. I'm assuming that's a mistake? I couldn't find an assoc
864.
▲
by
vunderba
8mo ago
A bit pedantic but: > Here’s the thing about Space Invaders, the game: it didn’t end when the aliens reached the bottom. It just started over, harder and faster. Author might be thinking of Centipede? The original Space Invaders from t
865.
▲
by
vunderba
8mo ago
Nice write-up! From the article: > Instead of asking the model to choose a move, I asked it to write the logic that decides what a good move looks like. Once I reframed Tetris as a coding problem, everything changed. Agreed. Nowadays t
866.
▲
by
vunderba
8mo ago
I used to have a bunch of bespoke node express server utilities that I liked to keep running in the background to have access to throughout the day but 40-50mb per process adds up quickly. I’ve been throwing codex at them and now they’ve al
867.
▲
by
vunderba
8mo ago
I’m glad I’m not the only one who’s noticed these seemingly arbitrary calls to write files using the cat command instead of the native file edit capabilities of the agent.
868.
▲
by
vunderba
8mo ago
At the risk of stating the obvious, DUH. There was a relatively comprehensive article around benching LLMs to play chess that measured even the SOTA models at around a mediocre 1000 ELO as compared to Carlsen who is rated at ~2850. https:&
869.
▲
by
vunderba
8mo ago
Nice job. There's another one I've used in the past where you can play host/join Open Arena games which is a Q3 fork. https://openarena.live
870.
▲
by
vunderba
8mo ago
Nice job. A small suggestion but it might be a good idea to rename Piano to something a bit more inclusive like Keyboard (which would include harpsichord, organ, pianoforte, etc) - since I initially did a double-take when I saw Buxtehude, a
More ›