Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
gisanokharu
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
by
gisanokharu
5mo ago
curious what this does that copying to the input box doesnt. is it extracting content from urls or supporting some format the web ui normally strips
2.
▲
by
gisanokharu
5mo ago
curious what class of erdos problems it solves specifically. is it finding erdos numbers for authors, or actually solving erdos-style mathematical conjectures
3.
▲
by
gisanokharu
6mo ago
the structured template is the right call. generic "looks great" feedback is useless because no one is forced to be specific. curious how you stop the reciprocity trap where everyone gives 5 stars to get 5 stars back
4.
▲
by
gisanokharu
6mo ago
ive seen the chladni figures before but never tried an app that does this. looks like the kind of thing you open once, say hohoho cool, then never open again. would be nice if it let you export the patterns as svg or something so you could
5.
▲
by
gisanokharu
6mo ago
the activation problem on Apple Silicon VMs is no joke. apples attestation system fights you every step of the way. curious how ghostvm handles that - do they have a workaround or are you limited in what you can actually do inside the VM
6.
▲
by
gisanokharu
6mo ago
screen capture every 4 seconds is a lot of data accumulation for agent context. how are you handling sensitive information appearing on screen. are you filtering anything before it hits the markdown, or is that on the user to manage
7.
▲
by
gisanokharu
6mo ago
curious what the output is - web games, unity, or something else? also, how much manual tweaking do you end up doing after the vibe part
8.
▲
by
gisanokharu
6mo ago
what format? been using .claude files for similar stuff but prompt engineering is still the pain point. is this for state or for instructing the agent
9.
▲
by
gisanokharu
6mo ago
this looks useful. curious how you detect structural twins - ast-level matching or something deeper? also, does it work cross-language or is it tuned for specific ones
10.
▲
by
gisanokharu
6mo ago
curious whether metadata survives the PNG roundtrip. things like original creation dates, software used, or embedded thumbnails can still leak info even in rasterized PDFs. might be worth adding a strip step if you isnt already doing it
11.
▲
by
gisanokharu
6mo ago
the iframe bloat problem is real. been using lite-youtube-embed for similar reasons. curious how yours compares - do you render the thumbnail server-side or client-side, and how do you handle the cookie consent modal that youtube forces now
12.
▲
by
gisanokharu
6mo ago
interview prep is crowded. everyone does mock interviews now. what makes your code reviews different - is it AI based or do you have real engineers reviewing? the biggest gap in interview prep is usually getting feedback on code quality, no
13.
▲
by
gisanokharu
6mo ago
curious how you decide what makes the cut. the noise-to-signal ratio on AI twitter is insane and most curated feeds end up feeling like link dumps within a month. what separates the signal from the noise for you
14.
▲
by
gisanokharu
6mo ago
curious how the AI handles messy receipts - like folded creases, multiple transactions on one paper, or those handwritten 'tax exempt' notes some places write on the total. also curious how you deal with tip calculation edge cases
15.
▲
by
gisanokharu
6mo ago
the half-life thing is what gets most people. they stop at 2pm thinking they are safe but its 5-6 hours. had a coworker who quit coffee at noon and still could not sleep, did not connect it until we traced back
16.
▲
by
gisanokharu
6mo ago
reminds me of the memento mori clock stuff from early 2010s. email delivery is the right call, keeps it alive without another app to check. though id imagine most people stop opening after the first month
17.
▲
by
gisanokharu
6mo ago
curious how this differentiates from just adding ai-specific directives to standard robots.txt. like do you have a use case where you need the age rating that simple disallow rules cant cover
18.
▲
by
gisanokharu
6mo ago
interesting take. curious what you mean by format specifically - is this about tokenization, autoregressive next-token prediction, or something else? my experience is that hallucinations are worse on sparse facts that require precise recall