Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
drusepth
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
22 ms
·
271.
▲
by
drusepth
3y ago
Imagine you have a smart, multi-purpose robot that has been trained to be a jack-of-all-trades. The robot knows a little about a lot of things, but the extent of its knowledge about any specific thing is generally limited. Fine-tuning (or j
272.
▲
by
drusepth
3y ago
Lightsail is a pretty good middleground between fully-managed and fully-self-hosted, though. One of my WP blogs is on Lightsail and it was <5 minutes to set up and be ready; and TBH I'm pretty darn devops-challenged.
273.
▲
by
drusepth
3y ago
I would trade GPU cycles for more time in my day.
274.
▲
by
drusepth
3y ago
Quick fix: the higher the temperature, the more varied the potential responses. For uniform responses, you want the lowest temperature.
275.
▲
by
drusepth
3y ago
>As far as I've heard, the EU isn't chasing folks who run a small website. But they could , which has already had a chilling effect on small businesses. Even though the intent (and current enforcement) is to punish large com
276.
▲
by
drusepth
3y ago
Part of the problem with the bar arrow keys is that the up/down are so small. I wouldn't want to "fix" the problem by making the left/right equally small.
277.
▲
by
drusepth
3y ago
Pry is amazing for understanding code. IMO the most useful method is absolutely `some_object.methods`, but you can also improve the output a ton by always doing: some_object.methods - {}.methods # or some other "parent" clas
278.
▲
by
drusepth
3y ago
Not everyone has access to GPT4. Those with access may also not want to waste one of their 25 quota messages on summarizing a random article.
279.
▲
by
drusepth
3y ago
This reminds me of a game I made a while back with gpt3.5 based on the AI Box thought experiment [0], where instead of social-engineering a password you're meant to social-engineer the AI into pressing a button. Plot-wise, the AI think
280.
▲
by
drusepth
3y ago
I think there are plenty of humans who wouldn't notice, though. And probably plenty of AI implementations that would notice.
281.
▲
by
drusepth
3y ago
Been using Alpine.js in a Ruby on Rails app for about 6 months now. It's a very cool, lightweight js library that I rarely see talked about.
282.
▲
by
drusepth
3y ago
Neither could/would most men on the street.
283.
▲
by
drusepth
3y ago
I think a lot of Lyft's early "we're like Uber, but better for women!" marketing hints that this is a killer feature that will absolutely drive growth. Especially now that Lyft, too, is saturated to the point of equally-
284.
▲
by
drusepth
3y ago
Does the fourth passenger sit in the "driver's" seat, or is it a three-packed-in-the-back kind of situation? Never seen a self-driving-car in person, but I assume it's the standard 4-seater sedan layout, right?
285.
▲
by
drusepth
3y ago
Legally and literally: yes. Knowingly using a private API without authorization can fall under CFAA, contract law, copyright law, trespass to chattel, etc -- and you can issue a C&D and/or sue for whatever is relevant.
286.
▲
by
drusepth
3y ago
Presumably, they're looking out for their paying users (see: they want to keep those paying users), who would have a terrible experience if and when they found out someone else had been using their APIs and/or API keys.
287.
▲
by
drusepth
3y ago
> If even 1 of those people decides to re-subscribe, Netflix has lost nothing. If 2 or more subscribe, they win. While extrapolating over a large dataset, it's important to get the initial numbers as correct as possible. Since the p
288.
▲
by
drusepth
3y ago
Pedantry activated: in my game dev circles, I've only ever heard .io pronounced "dot i-o" when talking about itch.io. The style, however, I usually hear as "i-o style" or an "i-o game".
289.
▲
by
drusepth
4y ago
>There's no point in creating/producing anything as almost anything can be produced by anyone, at incredible quality, at no cost and with little skill. Does this imply that some significant portion of art "value" is d
290.
▲
by
drusepth
4y ago
The late-night March 31 release of Worldcoin had the (unintended?) side-effect of making me think "a token to prove my personhood" was an April Fools Joke when I saw it the next morning and never thought about it again.
291.
▲
by
drusepth
4y ago
> The big problem is the variation. If you want a specific, complex scene it's basically impossible to get exactly what you want in my experience. The UI/UX of a proper workflow is still very janky (even in user-friendly UIs li
292.
▲
by
drusepth
4y ago
I use a standard disk icon to represent "saving" in a popular web app I run. I very often see users referring to it with "click the square thing to save".
293.
▲
by
drusepth
4y ago
Can gifski handle the initial recording (e.g. selecting an area of your monitor and recording)? Or is it just for converting other recordings into gif?
294.
▲
by
drusepth
4y ago
With so many sites designed to track activity (and per-user activity) on Twitter, I'd lean towards "no mass exodus" simply from the fact that I haven't seen any gotcha graphs/charts/data literally showing it. H
295.
▲
by
drusepth
4y ago
I wasn't able to see the Substack design until I hacked away at the obtrusive signup wall. Here's a side-by-side comparison for anyone that doesn't want to go through the same process: https://i.imgur.com/On0R
296.
▲
by
drusepth
4y ago
I don't have a resource for you (and will probably read whatever you get linked), but one intuitive way to think about it is that VCs/investors (and most of the startup ecosystem) are generally focused on "growth", not &
297.
▲
by
drusepth
4y ago
As someone who bootstrapped 8 years ago (and has taken nothing but rent+ramen since then), working _only_ double time sounds like a dream. Side-note: $130k being competitive obviously depends on your area/market, but that's far ab
298.
▲
by
drusepth
4y ago
What was the alleged zero-day used? The article fails to mention whether the security researcher used ChatGPT to implement a known zero-day or if it discovered one on its own. I don't think "AV doesn't detect this malware&quo
299.
▲
by
drusepth
4y ago
This looks interesting if it works. A side benefit of prompt hacking like this appears when you transition from OpenAI to local, cost-static models, too. Losslessly minimizing the number of tokens in a prompt should also (somewhat) reduce r
300.
▲
by
drusepth
4y ago
Historically, I don't think an absurd waste of power is gonna dissuade most people from throwing LLMs at everything they can. I've already seen functions like LessThanOnePage(text) that combine a long prompt + freeform text in a s
More ›