sampullman

I picked up the 128gb version when it was $2,199 and it runs Qwen 3.6 reasonably well with a 128kb context. Not very useful for complex tasks but it can handle some web stuff.

They're talking about modern games running on screens with high refresh rates.

I've built a few little games for myself both with and without AI, and completely agree. AI can help prototype an idea faster, or clone something very specific, but it can't make your control scheme feel good, invent a unique mechanic, etc (at least not yet).

I used the IRS free fillable forms this year and it was much easier than I expected, including contractor, investment income, and foreign income declarations.

bia hơi is pretty light

You adjust pretty quickly. Taking away compiler error messages would be fun though.

There's a bunch of stuff I include, depending on the project. Some general ones are commenting style and coding standards. In theory it should be able to do it without that by looking at the repo style, but I haven't found that to be the case (especially with overly verbose/repetitive comments).

A specific example in another project is the testing/verification procedure. It's a wasm/WebGPU and the test harness is fairly complex. There are scripts to handle it, but by default Claude will churn for a while to figure it out and sometimes just give up. It definitely saves a lot of tokens/speeds things up.

How ofter are you switching tips? It's been a while since I did any real soldering, but I don't remember often needing to switch in the middle of a session.

To make this claim, wouldn't you need to understand the meaning of the script? It's probably not about monads, but you don't know for sure.

Why? I don't have a porch at home, and it's too hot to sit outside and read.

Did you reply to the wrong submission?

So you think the Vaporfly prototypes Kipchoge wore in 2018 placebo'd him into crushing the world record by 78 seconds?

You don't need to be competing on the world stage to enjoy some of the benefits of Alpha flys or those pumas. 500 for the new Adidas does seem a little silly though.

I usually don't mind, but tend to split reviews into two types. Either I understand the context and can quickly do an in depth review, or I have to take some time to actually learn about the code by reviewing the surrounding systems, experimenting with it, etc. But in both cases I would at least run the code and verify correctness.

I think it becomes a chore when there are too many trivial mistakes, and you feel like your time would have been better spent writing it yourself. As models and agent frameworks improve I see this happening less and less.

I drag a tiny fetch wrapper around with error/json handling, timeouts and basic interceptor support. It doesn't cover everything axios does but it's nice enough and I haven't had to touch it in a couple years.

For reference: https://github.com/sampullman/fetch-api/blob/main/lib/fetchA...

I usually get it a few seconds ahead of time at least, in Taipei. I figured it's more related to the proximity than anything else.

You might be surprised, there are plenty of low effort attacks out there that just install a crypto miner and phone home periodically without doing much to cover it up.

For that I'm not so sure. I tried both early 2025 and was disappointed in their ability to deal with a TCA based app (iOS) and Jetpack compose stuff on Android, but I assume Opus 4.6 and GPT 5.4 are much better.

That's interesting, I actively use both and usually find it to be a toss up which one performs better at a given task. I generally find Claude to be better with complex tool calls and Codex to be better at reviewing code, but otherwise don't see a significant difference.

It's not possible in Taiwan's current political/social climate. I'm not so confident to say 50 years, but 20+ feels conservative.

Agreed, I've blocked all notifications for years. Maybe it got worse recently, but I thought they were annoying since at least Big Sur.

That's a good point, I'm not 100% sure it's worth throwing away the potentially beneficial uses. There might not be a solution that's both feasible to implement and avoids banning useful things. In the end I usually come back to it being the parent's responsibility to monitor usage, limit screen time, etc., but it hasn't been working so well in practice.

I think there's a little more nuance than that, but it seems roughly correct.

Wouldn't it be better if apps/websites targeting kids didn't use A/B testing to be more addictive?

It's 1.5 miles in 14:25, I think most people can handle that. There are plenty of ways to exercise that aren't plain running. Biking, skating, swimming, Tai chi...

There is lossy PNG compression that works very well for images using a limited color palette (pngquant, lossypng, etc).

It can be harder, but it's specific to the country/system. Here it Taiwan you can walk into any clinic with stock and get a (NHI covered) vaccine any time.

There are other things to complain about of course, but the rules for what's covered ate generally logical. Non-covered medication is affordable to, which helps.

That's true for code editing, but it's nice to not have to reach for a different solution when editing huge files. Sometimes I like to open up big log files, JSON test data, etc.

If you buy a game and can't tell it's made with AI, isn't that just as good?

That's true, and I more or less already use it that way for things like one off scripts, mock APIs, etc.

The person I'm replying to believes that there will be a point when you no longer need to test (or review) the output of LLMs, similar to how you don't think about the generated asm/bytecode/etc of a compiler.

That's what I disagree with - everything you said is obviously true, but I don't see how it's related to the discussion.