HN user

jeanvalmarc

134 karma
Posts0
Comments33
View on HN
No posts found.

These folks halfway across the world aren't bothering, nor do they actually threaten my freedom in any way shape or form.

As a counterargument, you can't know that. It seems like this argument could be restated as "I don't trust the US government to dispassionately determine who is a threat and even if they were a threat I am not in favor of killing people as a response." Which seems fair but the world is a complicated place and there are justifiable uses of drone strikes.

It's hilarious and probably terrible for society that you can sell your company but still control it. The most powerful companies become even less accountable to anyone. If you buy a share in an active business, sure the amount of control that share represents was factored into the share price. But it is still slightly absurd that these companies with zero accountability are allowed to be traded on public markets.

Here's a pretty detailed argument against dual-class shares from the UPenn law review: https://scholarship.law.upenn.edu/cgi/viewcontent.cgi?articl...

There have been a total of six deaths related to Autopilot misuse over 3 billion Autopilot driven miles by the fleet. Humans are terrible at risk management.

Or, humans rightly assessed that despite marketing to the contrary autopilot only really works on the easiest part of driving (highway miles) and still managed to kill 6 people.

I think treating it as a job where 75% of the salary is experience is probably the right attitude. FWIW I worked with a lot of grad students and anecdotally saw the best outcomes from people who within 2-3 months of showing up knew exactly what they wanted to do, were in the lab 10-8 to do it, graduated in 4-5 years and immediately got top-tier jobs.

I knew people who spent (7, 8, 9, 10, 11) years as a PhD student and they sure didn't seem happy or end up particularly good places...

If they'd done this right at the PY3K transition and only allowed the function-call-without-parenthesis syntax for `print` calls it could have been a good idea. But now that the community paid the decade-long price to get print as a normal function why backtrack?

Weird, as anecdotes go I've always thought the signal iOS app was very polished. I just checked three of those issues (camera access, shared links, search) on an iPhone 11 Pro Max and I wasn't able to reproduce. I suspect Signal may not have a QA group large enough to test on every phone/environment.

Pay 10% more than other non-FAANG companies (who are on a different planet) and offer a real mission and opportunity to make a difference in the product/world/industry. Make sure there's free coffee, smart people, and no jerks.

Amazon just lists it as a "subtotal," which I think is probably the best way to do it. I don't know about the "UX designer in a couple of hours" line: automatic shipping is a nightmarish bag of worms and isn't really a UX problem. What do you do if they order multiple SKU's that don't pack nicely into one box or are warehoused in different locations? Or if there is something weird about their shipping address? What do you use for box size and weight, and what approximation did you use for dunnage?

You can approximate UPS/FedEX costs by fitting a trend line, they are decently modeled by a linear (base charge + K*distance) function, but when you go to buy the label you might be way off. This puts you in the lose-lose-lose of either eating the difference of wrong estimates, overcharging for shipping and losing conversions, or just making people hate you by increasing the shipping over the estimate. Making a shipping API call is noticeably slow, so most people require an interaction after the address is entered.

Tim Sweeney's hot take was that "the two hardest problems in computer science are cache invalidation and shopping carts!"

do you think our government is better than theirs? For how long?

America has problems, but your order-of-magnitude calibration is way off if you think a concentration-camp operating (i.e. the real kind with over a million people who committed no crime and are held just for their beliefs and culture) expansionist (Hong Kong, Tibet, Taiwan) totalitarian (speech is monitored and punished in a "go to prison, right away" manner rather than "twitter deleted my racist shit and nothing else happened to me" way) government. So yes, occasionally ineffective democracies that value human rights and freedoms (i.e. the west) is better than the CCP government, and will continue to be better now and into the foreseeable future.

Is it desirable for a distributed Python package to write files to a hardcoded absolute path? Wheels seem to respect the `package_data` setup key, which is definitely inside the library path and seems to work pretty well for library data. I'd be a bit horrified if a package I installed from Pypi attempted to write stuff in `/etc/` or whatever. I guess I'm just asking, aren't these two separate problems (config management and Python packaging)?

Firefox 75.0 6 years ago

The number one rule of UX changes, people shit on everything no matter what you do. Counter-anecdote: it's a fine-to-good change.

K8 specifically might not be the future, but containers are the past (chroot) present (docker) and future. Sure, containers are a hacky shell script... but a hacky shell script that always runs the same way. I put in my time chasing down RPM dependencies, and debugging weird behavior because of a different version of some shared library I previously didn't know existed. I think the arguments presented here are focusing on implementation details, while the higher level picture is that containers-the-concept save a TON of developer (and devops) time and pain.

I think an important angle of "consider the cost" is that materials in R&D are pretty much ALWAYS cheaper than a team and time. For example here if the whole rocket was a 4mm thick 9m X 118m cylinder, the actual stainless steel would be like $700K [1]. While the rough headcount ceiling cost at SpaceX (7000 employees) is around $5M per day. And this way they get actual valuable engineering experience out of it, rather than just arguing in conference rooms like a more conservative company would have done. I think fail-fast prototypes are super valuable and are what good R&D looks like.

[1] https://www.wolframalpha.com/input/?i=%28pi+*+9m+*+4mm+*+118...

I've set up automated welding cells, and the "settings" excuse sounds totally reasonable to me haha. There are indeed a million settings (weld weave pitch and amplitude, wire feed speed, linear feed speed, gas flow rate, amperage, etc.) that are hard to tune 100% correctly and don't have "analytic solutions" (i.e. you tune some values like a monkey until it gives the result you want). Seems like this tank failure was their "destructive testing," although most people would probably have used smaller samples to save money. As someone else said: pressure vessels are hard!

GCP seems culturally a terrible fit for anything compliance-related. Azure and AWS both offer high-compliance regions (Azure Government / AWS GovCloud) that are absolutely necessary for many industries (medical, defense, etc). Slightly salty about this as I am currently migrating a legacy Firebase app because Google doesn't care...

So in summary Wikipedia is a responsibly run nonprofit that spends slightly less than it takes in while providing a valuable service to millions (or billions) of users?

Bonuses are something that’s easy to get outraged by, but look like an infinitesimal fraction of the business numbers. A judge approved a $235M bonus structure across 10,000 employees[1], while the wildfire liabilities are at least $18B [2], their 4-year infrastructure investment plan is $28B [3], and their top-line annual revenue is $17B. How many competent execs are going to want any part of that shitshow? $235M doesn’t sound like enough…

[1] https://www.sfchronicle.com/business/article/Judge-approves-...

[2] https://www.latimes.com/business/story/2019-09-09/pge-looks-...

[3] https://www.pge.com/en/about/newsroom/newsdetails/index.page...