Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
blinkbat
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
15 ms
·
211.
▲
by
blinkbat
7mo ago
did you write this? because ironically it feels very LLM.
212.
▲
by
blinkbat
7mo ago
k
213.
▲
by
blinkbat
7mo ago
I would say writing and game design are parts of a game that humans enjoy recognizing the human artistry in, and automating either degrades the artform. ie, these are the _fun_ parts, not the boring glue code. these are the parts that human
214.
▲
by
blinkbat
7mo ago
ok, what definition are we going by? let's start with the fundamentals, that there is no heavenly canopy and no daddy guarding it. what else?
215.
▲
by
blinkbat
7mo ago
> If you keep vibe-adding features, and somehow keep getting customers to pay for this thing, what happens once the codebase becomes so complex that an LLM cannot fit it inside its “brain”? you realize this point is well, well beyond wha
216.
▲
by
blinkbat
7mo ago
time will tell. you can set reasonable constraints and review the code. unless you are disqualifying that as vibecoding.
217.
▲
by
blinkbat
7mo ago
I got voted down for this? "Why are you booing me, I'm right"
218.
▲
by
blinkbat
7mo ago
Agree but most small biz don't conceive or care about the internet this way
219.
▲
by
blinkbat
7mo ago
vscode with claude and codex extensions. two running chats are fine for me, allows me to keep my eyes on the code rather than running 75 agents.
220.
▲
by
blinkbat
7mo ago
Lame. Do something cool again mozilla
221.
▲
by
blinkbat
7mo ago
two good ideas, thank you. if you come up with any others, I'll check back here.
222.
▲
Show HN: Playtest my browser-based RPG
(archipelago-sandy.vercel.app)
3 points
by
blinkbat
7mo ago
|
2 comments
223.
▲
by
blinkbat
7mo ago
Uh huh. Sure.
224.
▲
by
blinkbat
7mo ago
Yep the bottleneck is now review and QA
225.
▲
by
blinkbat
7mo ago
IT DOESN'T EVEN SLIDE HERE AAAA
226.
▲
by
blinkbat
7mo ago
i mean, you still guide a power tool. that requires a bit of expertise. for now, getting good products out of AI also requires expertise, but the future is unknown for that holding. it's more about the transition from woodworking to so
227.
▲
by
blinkbat
7mo ago
They told it to try _everything_...
228.
▲
by
blinkbat
7mo ago
Don't let it touch databases or prod. Minimize the impact crater
229.
▲
by
blinkbat
7mo ago
Par for the course sadly
230.
▲
by
blinkbat
7mo ago
For frontend code it's fine tbh. Perfectly capable with oversight
231.
▲
by
blinkbat
7mo ago
Shocker
232.
▲
by
blinkbat
7mo ago
We're talking models, not systems.
233.
▲
by
blinkbat
7mo ago
It does seem to wildly fluctuate lately. Codex is much more consistent. Must be that DoW money... :/
234.
▲
by
blinkbat
7mo ago
With my eyeballs if it's important. Labeling it as AI accomplishes nothing imo
235.
▲
by
blinkbat
7mo ago
reeks of LLM
236.
▲
by
blinkbat
7mo ago
same here, of course. so in a sense, LLMs are just the next phase of a gradual erosion, not a total sea change.
237.
▲
by
blinkbat
7mo ago
but all of that said, when is the last time you TRULY solved an entire codebase by hand? no stackoverflow, no google, no intellisense, no autocomplete (hell, no asking other engineers)?
238.
▲
by
blinkbat
7mo ago
it's kind of the difference between a studio musician and a songwriter. one you're doing for a paycheck, and you may or may not enjoy the work to varying degrees. the other is your passion. LLMs are freeing companies to produce fa
239.
▲
by
blinkbat
7mo ago
just make a game
240.
▲
by
blinkbat
7mo ago
Site itself has bad back-forward UX. Why should I trust it?
More ›