mgraczyk

14 ms·

211.
by mgraczyk 1y ago
It's not a tree though
212.
by mgraczyk 1y ago
Tree growing? And I don't follow, we've had vehicles capable of reaching the moon for over 55 years
213.
by mgraczyk 1y ago
This is meant to be some kind of Chinese room argument? Surely a 1e18 context window model running at 1e6 tokens per second could be AGI.
214.
by mgraczyk 1y ago
This is only ideally true. From the perspective of the user of a large closed LLM, this isn't quite right because of non-associativity, experiments, unversioned changes, etc. It's best to assume that the relationship between input
215.
by mgraczyk 1y ago
The total number of lottery winners who have won, in total, an amount of say 50,000 USD is at least 2 orders of magnitude smaller than the number of people who have received that amount or more from startup equity You are making an empirica
216.
by mgraczyk 1y ago
Pick a prize amount that is anywhere in the same order of magnitude as any engineers total compensation. For any such amount, there are more people receiving prizes above this amount from startups than lotteries
217.
by mgraczyk 1y ago
You are confusing expected value with probability of "winning". The purpose is to trade EV for a small probability of high payout. It's less extreme than the big ticket lotteries but more extreme than a scratcher. If that&#x2
218.
by mgraczyk 1y ago
You may want to try moving to the bay area or NYC for a few years if you're not already in one of those places. It's much easier to get a good job, then you can move away after you have the "right" experience. Also follo
219.
by mgraczyk 1y ago
What fraction of lottery participants win, vs startup employees? It's rare and there is luck, but the quantities matter. You are much more likely to make money from a startup and it's much more under your control as an employee
220.
by mgraczyk 1y ago
> all equity does is make your cap sheet look marginally worse But at the time the employee is negotiating this has already been decided. The company has some valuation and you are offering some known percentage of that scarce resource.
221.
by mgraczyk 1y ago
It's not really, because it's finite. You don't generally dilute the options pool when you hire new employees, you give them some slice of a scarce options pool
222.
by mgraczyk 1y ago
I don't think they misrepresented anything. The European Commission is wrong on the facts. Technology improved in unpredictable ways. Large public companies do not lie very often because it's incredibly easily for lies to be disco
223.
by mgraczyk 1y ago
If you have ~5+ years of experience and live in the US, things haven't changed. Regardless, negotiating matters more than ever. At startups you negotiate for equity instead of salary but a lot of the same advice applies.
224.
by mgraczyk 1y ago
Actually not even that. What happened was we added a mechanism to preload the camera to reduce startup time. And it was not gathering any data You're confusing the audio calls with secretly listening to microphone, which never happened
225.
by mgraczyk 1y ago
Where's the lie?
226.
by mgraczyk 1y ago
Dark patterns aren't lies though... And you're not even talking about Meta
227.
by mgraczyk 1y ago
I was around for a lot of these. In none of these cases did Meta lie. They are all either fake or honest mistakes that Meta never lied about. For the second one in particular, Meta never listened to anyone's mic. I would know, I worked
228.
by mgraczyk 1y ago
Is there any evidence Meta has ever intentionally lied about anything? Like do you have any examples?
229.
by mgraczyk 1y ago
And they are legally required to do this in most places
230.
by mgraczyk 1y ago
The trainium toolchain is not as mature as GPU. Your model may fail to compile out of the box, and even if it does it may be slow and require you to dig into details for reasonable training/inference performance
231.
by mgraczyk 1y ago
Maybe a dumb question, but how do you know how many frames to seek past? For example say you want to seek to 10MB into the uncompressed file. Do you need to store metadata separately to know how many frames to skip?
232.
by mgraczyk 1y ago
Not many, because cars are slow
233.
by mgraczyk 1y ago
Light could only go to Andromeda and back 1000 times before the sun burns out. That's not very many times IMO. On the scale of galaxies, light is slow relative to any timescale relevant to large objects.
234.
by mgraczyk 1y ago
I just tried in the app on Android and it would not allow me to take more photos unless I pay. When you say "you have access to everything as part of the free tier", does that apply to Android as well? Because what you are saying
235.
by mgraczyk 1y ago
When you write code, you have to spend time on ALL of the code, no matter how simple or obvious it is. When you read code, you can allocate your time to the parts that are more complex or important.
236.
by mgraczyk 1y ago
The paywall says I get extra stuff from premium, including "unlimited photo, voice, and barcode logging", "unlimited AI nutritionist chat", "unlimited Personalized tips from your AI nutritionist". It sounds lik
237.
by mgraczyk 1y ago
What do you mean by this? > it's completely free to use SnapCalorie, the $79.99 price is a donation to support our research I just downloaded the app and I don't see any way to use the premium plan without paying 79.99 or 6.67&
238.
by mgraczyk 1y ago
Good post, the only part I think I disagree with is > Never. Let. AI. Write. Your. Tests. AI writes all of my tests now, but I review them all carefully. Especially for new code, you have to let the AI write tests if you want it to work
239.
by mgraczyk 1y ago
Yeah I included quotes for the comment to delimit the quiet, you don't need them in the search
240.
by mgraczyk 1y ago
Try "topic site:chatgpt.com/share/*"

More ›