semicolon_storm

It’s truly strange that people keep citing the quality of Claude code’s leaked source as if it’s proof vibe coding doesn’t work.

If anything, it’s the exact opposite. It shows that you can build a crazy popular & successful product while violating all the traditional rules about “good” code.

Everything's more costly in Hawaii, including maintaining infrastructure

Sorry you got scammed, FIFA awarded me their 100m sprint gold medal so I'm the real fastest 100m runner in the world

It seems unlikely, athletes are consuming liters of oxygen per minute. Plus, the impact is at least partially offset by needing to carry the extra weight with you.

Different story for apnea sports like freediving where a little bit of extra oxygen goes a long way.

Don't hold your breath for that, Azure still has spotty IPv6 support

A proper training plan makes you more injury resistant and makes it easier to bounce back from injuries if you do happen to get hurt.

For example, just look at the data for sarcopenia: https://pmc.ncbi.nlm.nih.gov/articles/PMC9654071

It requires mercury-198 as an input. The only quote I could find for mercury-198 is about $15K per milligram.

So now we just need to figure out how to make mercury-198 cheaply.

You can also have the most flexible system ever designed, but if the rest of your team doesn't understand it then good luck implementing that required use cases

And treating politics like team sports is the root of the problem

Adding on to what others have said, LastPass stored vault "metadata" unencrypted. Metadata included things the url. This allowed the attackers to prioritize cracking vaults of higher value.

See a vault with just a facebook.com and google.com login? Skip it. See a vault with coinbase and 10 other crypto sites in it? Spend a few thousand trying to crack it.

Source: https://github.com/cfbao/lastpass-vault-parser/wiki/LastPass...

Not sure about that. A hash function suitable for security sensitive work, used properly, should make a collision so unlikely that you can basically forget it that it's even possible.

Think about it, that's what hashing passwords relies on. We don't store a plaintext password for a final check if the password hash matches, we count on a collision being basically impossible.

A hashmap is different, because it's using a much weaker hash function with far fewer security guarantees.

Plus, you're assuming the original values are even kept around for comparison. The cache key likely just mapped to something simple like a boolean or status flag.

The censorship is not present in the distilled models which you can run locally

Are you referring to the distilled models?

Perhaps because that's a strawman argument. "Scaling" doesn't mean double the investment and get double the performance. Even OpenAI's own scaling laws paper doesn't argue that, in the graphs compute increases exponentially. What LLM scaling means is that there hasn't been a wall found where the loss stops decreasing. Increase model size/data/compute and loss will decrease -- so far.

In security, industry standard seems to be about the same as military grade: the cheapest possible option that still checks all the boxes for SOC.

Phi-3 small is

No mention or comparison with phi-3 seems odd. Isn't phi-3 leading the other models by a bit?

Water on the leaves isn’t bad if you’re watering in the morning where it’ll soon evaporate in the sunlight.

No paywall: https://web.archive.org/web/20240410092221/https://www.nytim...

And install all their dependencies from non, which Microsoft also owns

Liberland had a government. Apparently no one ever told them they didn't need it.

Why? Apple is a “premium” offering and doesn’t have a monopoly in any area. If there was no other options I would agree

Every ORM I've ever used has some raw SQL escape hatch you can use when you hit that edgecase. For the 90% of DB access that really is simple, ORMs are a pleasure. For that other 10%, if your alternative is raw SQL, just use the escape hatch and you're not worse off than if you had skipped the ORM.

I work at a company that does a lot of code generation, and it gets uglier the longer you do it. It's much harder to write the code that generates the code you want than to just write the damn code in the first place. The abstractions & assumptions made for your code generator will eventually begin to break down, and when that finally happens everything goes from a simple refactoring to way overly complicated update to the generator.

10 per day is a ridiculously low limit. Also, NAT and shared networks (think college campus) mean that hundreds of people often share the same IP, making that limit even more ridiculous.

Reality would be that limit would have to be much higher.

Pretty insane you can fly these with no license or certificates when basically all drones require having your TRUST certificate or full on drone license to fly.

A proxy war between NATO and Russia isn’t that rare of an edge case.

That’s demonstrably false. Google the red hill fuel tanks

This is such a poor interpretation of the data that it almost seems intentional.

They're cherry-picking an arbitrary year for comparison over an arbitrary time range. There's many years you can select to draw similar bad comparisons. For example, select 2021 & 2015 and look at Feb 28 - March 5. Are you going to believe a sensationalist claim that "90% of SST rise between 2021 & 2015 arrived in 2 weeks"?

Hopefully not, because that's a bad comparison and ignores that temperature fluctuates up and down

The situation sucks, but their continuing to live there and be insured has to be subsidized by somebody somewhere. The area is clearly insanely risky to the point where no premium justifies offering coverage.

If coverage was offered, who should be the one subsidizing it? Private companies? No way, they never would. Government? That's just a regressive policy funneling money to a relative well-off (homeowners) group of people.