lostdog

Yeah, you can see this with Google's search results too. They're trying to improve on some internal metric, but the metric was clearly generated from ratings by people ignorant of the topics. And so the search results get worse, but appear better internally.

Great to see that they have not learned from this experience, and are repeating the mistake with Gemini.

There is no such thing as a good flow for showing sponsored items in an LLM workflow.

The point of using an LLM is to find the thing that matches your preferences the best. As soon as the amount of money the LLM company makes plays into what's shown, the LLM is no longer aligned with the user, and no longer a good tool.

How about: tech companies must implement mandatory screenings of users' messages and posts to look for violent intent.

Sadly, the default values are an even bigger source of bugs. We just caught another one at $work where a field was never being filled in, but the default values made it look fine. It caused hidden failures later on.

It's an incredibly frustrating "feature" to deal with, and causes lots of problems in proto3.

I'm not sure if there's any hope of US healthcare improving for a long while. The ACA was the only recent improvement. It was a small change, but the electorate decided that it was bad, pushed its supporters out of office, and elected a series of people who promised to "repeal and replace" it, but weren't really serious about any improvement. And then when the "repeal and replace" crowd failed, they were not held accountable, but reelected anyways.

There also isn't much interest in improving healthcare from either side right now. The right has nothing. Their current platform is ignorant views about vaccines. The left has a stronger interest in Palestine and housing abundance right now, though all of that is dwarfed by trying to keep the rule of law going, and preventing us from falling out of a democracy. Healthcare is way way down the list for everyone sadly--even Bernie doesn't talk about it much anymore. The electorate has spoken, and they are not interested.

It gets super frustrating to have to empty/null check fields everywhere you use them, especially for fields that are effectively required for the message to make sense.

A very common example I see is Vec3 (just x, y, z). In proto2 you should be checking for the presence of x,y,z every time you use them, and when you do that in math equations, the incessant existence checks completely obscure the math. Really, you want to validate the presence of these fields during the parse. But in practice, what I see is either just assuming the fields exist in code and crashing on null, or admitting that protos are too clunky to use, and immediately converting every proto into a mirror internal type. It really feels like there's a major design gap here.

Don't get me started on the moronic design of proto3, where every time you see Vec3(0,0,0) you get to wonder whether it's the right value or mistakenly unset.

Mosh is so good, but their unwillingness to implement OSC52 for copying to clipboard sent me back to tmux.

There is not that much available data for these, so it will take new techniques to get AIs that are truly good.

Does solar scale linearly when you have to get onto roofs to install it? And when each roof is available for installation at different times, so only small crews can do it piecemeal?

Every time I copy paste something in python I have to check the whitespace. Often there's a problem that needs fixing, like the first line indented differently, or everything is off. Sometimes I'm not paying perfect attention and get hard to catch bugs.

I'm now certain that significant whitespace is simply wrong.

It's a defect. We should fix it by making them do a recall.

And that's why the real name policies helped kill Facebook. The best way to section off your friends from family from work is a separate somewhat anonymous account for each.

Absolutely. I learned so much from my degree. Even at a decent school, the top level classes walked me through algorithms, compilers, operating systems, and graphics in a way I might not have done myself. Plus, the best people in my class were great, and discussing how to problem solve with them for coursework was very very helpful. The whole thing was very helpful as practice in a bunch of different areas of CS.

On the practical end, your degree is still a big filter for how companies screen applicants. A lack of a CS degree makes it way more difficult to get in the door. I know there are counterexamples, but 99% of my coworkers had CS degrees. Unless the money is a huge burden, or the school you would go to really is crap, then yeah, the degree makes the rest of the journey much easier.

As advice, try to get evaluated out of the intro courses and skip to higher level stuff. Often if you can show a prof that you can program (before the school year begins) you can place right into a course that is at your level.

The AI arguments are also facile. A CS degree is still the best training to be ready to use AI effectively. You understand more about how it is built than everyone else, plus you know way more about every tool surrounding the AI, and would have an easier time making your own.

I know some people have success with alternative paths. But the main road is the main road for a reason here.

Edit: Please also look at the best schools for CS, and don't just discard them as options. My coworkers out of these schools got even more out of their classes and costudents than I did. Your son sounds talented, and the best schools can help talent go even further.

Mosh does not support OSC52, so it's a barrier to getting copy/paste to work.

I have tried exactly this, and it does not work. Nobody who did direct work ever had recommendations for peer workers. They just don't care. Most subs work directly with generals and not each other that much.

The general contractors know which subs are (currently) good and bad, so their recs are good, but you have to be friendly with them for them to share (instead of taking the project on themselves).

Ok, but let's make the non-compete bidirectional.

If I leave or am fired from your company, then you are not allowed to hire anybody who works in my field for at least one year.

Inkscape. I love it, but a bit of touchup work would be really nice. I would especially love boxes and connectors to be even more seamless, so I can replace all the digital whiteboard apps with it.

"Comprehensive Error Handling: Helix's error handling system is novel to helix, it allows for either complete error handling or a more pythonic approach to error handling, this allows for more control over how errors are handled in Helix."

Wow, these really do seem to be LLM written, and I would guess ChatGPT. This long README has a bunch of generic, redundant statements over and over again.

Put me behind the wheel of a bus and I'm pretty sure I could get it somewhere safely.

A lot of the "science" we do is experimenting on bunches of humans, giving them surveys, and treating the result as objective. How many places can we do much better by surveying a specific AI?

It may not be objective, but at least it's consistent, and it reflects something about the default human position.

For example, there are no good ways of measuring the amount of technical debt in a codebase. It's such a fuzzy question that only subjective measures work. But what if we show the AI one file at a time, ask "Rate, 1-10, the comprehensibility, complexity, and malleability of this code," and then average across the codebase. Then we get measure of tech debt, which we can compare over time to measure if it's rising or falling. The AI makes subjective measurements consistent.

This essay gives such a cool new idea, while only scratching the surface.

Once a startup hits series A, early exercise is often $X0,000 and a lot of people cannot afford to take a speculative bet with that kind of money.

Yes, it is possible to care about more than one thing.

YC was founded more than 20 years ago, and I don't recall any lobbying on the issue of employee tax treatment. If they do care about this they are being super quiet about it.

Interesting that YC is jumping on this tax code thing, but the shitty tax situation for employees trying to hold onto the equity they've earned is just ignored.

The number one problem for startup employees is that the AMT tax makes it impossible to exercise options in a company that is doing well. YC is showing us that they will fight for changes to the tax code when it benefits their bottom line directly, but are silent when it comes to helping startup employees hold onto equity that they have earned.

Think carefully before joining a YC startup as an employee.

MAGA is DEI for morons.

To be fair, they need jobs too! But giving them all the White House jobs does not seem fair or effective to me.

They could have ended the diversity statements, but kept all the research.

They decided to end all the research too.

He said he would end it on day one. That's a promise he didn't keep.

It's not my fault he made such a dumb, impossible promise. It is true he broke it.

You are posting blatant misinformation. The MS-13 tattoos were painted into an existing picture.

February 29th, 2027, the so called "Hacker News" is declared a criminal and illegal site. The wartime powers delegated to the Supreme President allow him to imprison domestic enemies and remove them. Gradus_ad is right in the middle of explaining the difference between "begging" and "panhandling" to the hobo he is harrassing, when disguised agents grab him off the street for commenting on a criminal site. He is whisked off to a correction facility in Hungary (it is now illegal for this publication to call them "gulags") and never heard from again. Luckily the agents who disappeared him say everything was handled legally!

ICE is supposed to keep records, and the courts are supposed to create a transparent record in the case of a dispute.

But ICE hid the evidence and prevented the courts from looking into it.

The feds have been lying in their court filings for the past few months, so don't take their complaint at face value.