bloomca

> A simple system of, "Here's what we think you owe, based on the information we have on hand - sign and submit if you agree" would work for most people.

They already do that -- if you calculate your taxes wrong, they will send the adjustment (they will do it both ways, pay you back or ask for the remainder). I guess they might not be aware of all the deductions, but standard deduction beats itemized one for the majority, so they can 100% automate this whole process if they decide to. For complex cases and businesses, sure, you are on your own, but at least most W2 should be covered.

I believe any free space we have will be taken unless you actively fight for it. And while it is hard to maintain free space, clearing up taken space is even harder -- the discussion is usually what to put instead of something.

Pretty much this. Even in topics I'm decently knowledgeable, I feel like the vibe from SO answers will be "just read the spec" or source code; so I usually was able to string other answers together.

With LLMs you can start with first principles, confirm basic knowledge (of course, it hallucinates but I find it's not that hard to verify things most of the time) or just get pointers where to dive deeper.

Uber definitely screwed the workers and probably existing taxi companies, but for the users it was a huge W, at least in many parts of the world. Taxi companies are notoriously scammy and it seems to be a very universal experience.

I think the article itself mixes these concepts. It starts with the world-class level example, and then jumps into learning new skills over a week.

In order to be great, indeed, you don't need to be obsessed, and very rarely you cannot achieve it -- you just need to be relatively consistent. And I 100% agree that if you start early, maintaining a skill on the "better than vast majority" level is not too hard (the hard work was done in the past).

But to achieve truly outstanding results, you definitely need to dedicate tons of time and ideally be obsessed with it (at least very-very interested).

It unironically is. Maybe if you can afford not working, it's fine, but with a regular job you simply don't have that much free time. In vacuum, there is a lot of it, in practice once you get a few commitments, it gets very challenging.

There are some tricks -- if you are already good at something, maintaining that good level is usually not that hard (you already know how and can recalibrate your level to the available time), but picking up new concepts is very time consuming.

I'd also say that the article talks about living through something, that's probably more involved than just passionate.

The article brings a lot of examples of people who lived and breathed something, so it was essentially their life.

Telling people to dive into something even if they don't feel it is dangerous advice (especially for kids) -- people will just get burnt out.

Of course, there is nuance here -- you don't necessarily want to achieve world-class in something, and sometimes pushing is indeed worth it.

How would that work? How many bytes per character? How different fonts would work?

It is literally a highway at this point for any B2C tech product. All the info they are collecting surely can be used for very targeted ads. Not hard to predict at all.

I have been working on my terminal editor, but I parked that for now -- https://github.com/bloomca/love. It is possible to load a file and edit it, copy/paste works, you can select text, etc. The next step is to integrate with the tree-sitter for syntax highlighting and then with LSP, but it took a bit more time than I wanted.

Another project of mine is to play music from my audio CDs by myself. I built a simple Rust library to read TOC and raw PCM data from a CD drive -- https://github.com/Bloomca/rust-cd-da-reader (works on Windows, macOS and Linux), and a ripper -- https://github.com/Bloomca/audio-cd-ripper, which rips all tracks and encodes it as FLAC and fetches metadata from MusicBrainz.

The next step is to play it. I looked into using cpal (https://github.com/RustAudio/cpal), but I feel like using low-level audio API for each platform is a better approach for learning.

Managers 100% lose their abilities, their focus shifts to completely different concerns -- codebase health, enabling people, tracking velocity metrics, etc. They still understand high-level concerns, of course (if we are talking about strong technical background), but they'd struggle a lot if just dropped into the codebase.

Tech leads can exist in many variants, but usually they spend the majority of time in code, so they don't lose it. If they become too good at managing and change their priorities, they _will_ gradually drift away too.

Code is the final artifact after everything is shipped. But while the development is active, it is more than that (at least for now), as you need to know implementation details even if you are really proficient at the domain knowledge.

Although I do agree that there is a possibility that we'll build a relatively reliable abstraction using LLMs at some point, so this issue will go away. There probably be some restrictions, but I think it is possible.

Most cities in Europe absolutely do not have suburban zones where most people live. Non-center areas also have their own leisure, going out, shopping places (usually mixed with the residential areas).

The suburbs purely for residential space where you have to go somewhere else for activities do exist, but usually over time they grow their own infrastructure for shopping and hanging out without needing to go far.

Some models will still trickle down; hell, better/cheaper hardware should enable to run hefty models available today, and they seem to be already okay-ish with such queries.

I don't understand why would you opt in to share your data. Is it because you believe that it would help to improve the model and you would benefit from it? Or something altruistic?

What do you think is better? The web is indeed questionable, but it is literally the best we have, it is still reasonably simple to deploy a web app.

Desktop app development gets increasingly hostile and OSes introduce more and more TCC modals, you pretty much need a certificate to codesign an app if you sideload (and app stores have a lot of hassle involved), mobile clients had it bad for a while (and just announced that Android will require a dev certificate for sideloading as well).

edit: also another comment is correct, the reason it is like that is because it has the most eyes on it. In the past it was on desktop apps, which made them worse

Microsoft has way too much of legacy software people use, banning it all overnight will not go well at all. They understand that as well.

They tried to pull a similar move with WinRT/UWP, but nobody wanted it, so now you can continue with Win32.

They would love to do so, but legacy compatibility is a major business advantage.

This is how macOS works, without a signature they will tell you they can't guarantee it doesn't have malware and you need to go to settings and choose to run anyway (and most people don't even know about it).

Microsoft would love to do that too, but it just has too much of legacy software to introduce such a major hurdle.

I _believe_ on macOS, if they revoke your certificate (the one you used to sign the app), the app moves into the "We can't verify that this app is free of malware" category, so in theory they can still do that.

But it is a nuclear option, and it would be a big deal if they did so for something "minor".

It's not just about that. I am sure if the court would force them to allow sideloading, they'll make sure to never promote your app if you decide to offer both options to the users.

"Allow access" is pretty orthogonal, I don't know how it all works in mobile OSes, but I assume everything is virtualized there, so you can't just access whatever you want without user granting a permission (e.g. through a file picker system component).

You can also ship sandboxed apps on Desktop without the store (although I am not sure on how hard it is to auto-update them, usually stores handle that part), at least on Windows and macOS.

Stores handle storing the apps themselves and distributing updates, that part of the cost is real, plus they do manually review submissions (to some degree), but 30% is insane for that.

Is it allowed to charge more in storefronts which take these cuts? Why nobody does that?

What about Steam? Can a publisher sell a game for ~$45 in their store and $60 in Steam, or is it against some TOC?

I think about the same. Right now we are at the normalizing the ID verification stage and banning specific content in certain countries/states, once we are desensitized, VPNs will come next, and then some government solution to track everything you do online.

They can go after hostings as well and everybody can take down a lot of things out of fear.

> Nearly 20 years later, is that the career badge that you'll always feel good about?

Well, if the result work has negative connotations, you wouldn't even mention it (especially after 20 years). However, as you said:

> enough that she (and her firm) got paid a lot and maybe even did landmark work

At the end of the day, that's what mostly matters. Sure, some people believe in what they are doing and put insane hours, but most just do it for money. And if they manage to get a lot, then yeah, it was all justified.

---

> Second, the wipe-out rate is pronounced but not catastrophic

I agree with this -- people who are deeply invested in their projects are often already do the second shift. So if you are motivated enough, that's kind of the same, plus people can be in a position where they have no external obligations (often when they are young).

It is bad long-term, but for a relatively short term for many it is a decent gamble.

> 2. When NOT to use the truck... when talking or the bike is actually the better way to go.

Some people write racing car code, where a truck just doesn't bring much value. Some people go into more uncharted territories, where there are no roads (so the truck will not only slow you down, it will bring a bunch of dead weight).

If the road is straight, AI is wildly good. In fact, it is probably _too_ good; but it can easily miss a turn and it will take a minute to get it on track.

I am curious if we'll able to fine tune LLMs to assist with less known paths.

My understanding is that the base layer (similar to what FlatPak provides) is shared and downloaded by the snap manager so it is portable as long as you want to download it.

The end result should be similar to FlatPak where you have practically no dependencies as it should package almost everything.

It is relatively easy to replicate if you open 2 apps and start typing in the first one which opened, both on macOS and Windows (I don't use Linux enough to notice this issue).

A proper solution is probably faster startup times, but overall it pretty much never happens? Idk maybe I'm lucky or just conditioned to ignore it.

I've been looking at native development for quite some time now (WPF/WinUI/SwiftUI, starting with Win32 and AppKit), and honestly Web technologies are much better than that. The fact that it is cross-compatible is just a cherry on top.

If finally WASM gets a cheap and easy way to manipulate DOM, I think even more stuff will move towards web tech like Electron and hopefully Tauri in the future.

What kind of forms do you need? I honestly feel people think they have to use all sort of dependencies. There are some you definitely need, like the router or any state management (you can build both, but there is a decent amount of boilerplate and efficient hooks can be tricky).

But for forms I honestly would recommend to start with plain React and only abstract things you need in your project.

Node.js is a runtime, not a language. It is quite capable, but as per usual, it depends on what you need/have/know, ASP.NET Core is a very good choice too.