citelao

> So doesn't that lack of understanding concern you?

Good question! It mostly doesn’t concern me. I believe these models have a huge “latent space,” so it’s not surprising to me that other people, with the right harnesses and prompts, can get them to produce more-valuable content.

I spent a lot of time calibrating how I use these models to produce valuable, good code, right?

But as I alluded with my question above, I haven’t seen a ton of evidence that these models have human-level argumentation in that latent space. If my assumption there is wrong, then yes, I’d have to reexamine my understanding of these models.

OP: my company pays for my work tokens, which is the source of the $3,000/mo. I don’t use substantial tokens with my personal subscriptions, so I didn’t factor that into my number. But good point, I’ll clarify.

OP: I'm not surprised you/other people are able to get better output than me.

And I agree that "less AI" edit passes reduce a lot of the hallmarks, but I think these models truly struggle with expressing complicated ideas. Have Opus been good at that for you?

OP: [last year][1], I shared how I was using AI day-to-day as an ordinary engineer. This year, I did it again!

It's crazy realizing that Claude Code only started getting big in the last year or so. It feels so different now, with all the new models & tools we have to use. I hope you enjoy!

[1]: https://ben.stolovitz.com/posts/how_use_ai_oct_2025

Probably you learned that from them! The first user is 31%, from their testing.

https://www.nngroup.com/articles/why-you-only-need-to-test-w...

https://www.nngroup.com/articles/how-many-test-users/

OP: thank you for making an account to comment :).

I'm certainly glad I have my good programming job right now. I'm not sure where the field is going, but it certainly ain't 2017 anymore. I will say: like the sibling comment mentioned, nursing (or PA?) can still be a very independent job, especially as an APRN or flight nurse. But I wish you luck!

OP: +1. If you want to do fire, you should be aware most calls are medical.

Even though where I live a lot of firefighters are just EMRs... the fire services are a lot more willing to take you if you already have your EMT (since you're not gonna abuse the program to get free EMT training).

EMT really is the fastest way to become any first responder.

OP: my service requires 2 twelve-hour shifts a month + a monthly meeting. I had to complete a training packet before I could ride as a "full" EMT. Many services have stricter requirements; some require weekly shifts.

Maintaining the certificate also requires ~40hrs of continuing medical education every 2 yrs, which I accomplish via short 1hr courses every few weeks.

OP:

> How do you balance keeping the skills sharp while working another job full time?

I don't! I've been volunteering with my current service for ~2yrs, and I still don't feel like a very "smart" EMT. It's coming, but slowly. And I'm trying to do more shifts (e.g. every week).

I simply try to "manage" my stupidity. I usually ride with a more experienced EMT, and any very-big calls will get a medic; I focus on learning & asking questions when it's appropriate; I try to drill _something_ every shift. If I really wanted to learn quickly, I'd take a few weeks off and run with AMR in a city... that definitely gets the calls.

> Also, I’m 38, so I tend to feel too old.

That's one of the biggest reasons I delayed. I felt too old at 29 & worried I would be out of place. FWIW, I think the class after us had an average age of ~35, so you might fit in more than you expect.

I resonate with wanting to be good, though. I think about it every time I get a call. My counterargument is (a) that 1 month of training is 1 month more than most folks have, and (b) in an emergency, you're always part of a team.

Even if you just fetch the stairchair.

OP: I've wanted to be an EMT for a long time, but I made excuses for years. I wanted to share some of them with y'all; maybe they resonate with you, or inspire you to do something you've been putting off.

Because worst-case, you might enjoy it.

I think the more appropriate analogy is there's more to a drill than the drill. When you buy a Milwaukee drill, because it shares a battery with other Milwaukee parts, it's easier to buy more Milwaukee tools.

So it helps to choose a drill brand with the ecosystem you like, not just one that's a serviceable drill.

(Or you could argue that a language's ecosystem is part of the dang language. Which I agree with).

That’s the prevailing statistic, but your original number isn’t that wrong either:

Humans can perceive much smaller latencies.

If you look at the Card & Miller reference, at least some humans can perceive differences in ~50ms vs 100ms latencies when typing (in my limited testing, it’s likely you can!). There’s some newer research I don’t have handy that I believe found error rates decreased and NSAT improved until around at least 30ms (if not 20ms).

On that note, humans can definitely distinguish 60hz vs 120hz reliably (about 8ms faster per frame).

Even faster: with a reference (eg when dragging on a touchscreen), humans can distinguish down to at least 1ms vs 10ms of latency: https://m.youtube.com/watch?v=vOvQCPLkPt4

And you can probably distinguish metronomes that are off by about 1-2ms. Much smaller for other things (like metronomes that slightly slower or faster than one another).

This is a special interest of mine XD

My statement is not super constructive, but yes, I also really liked this article. Local-first, privacy, even the example of managing personal restaurant reviews.

I don't have enough meaningful thoughts here to contribute, but you & OP: I hope we can build a federated protocol for sharing stuff again.

OP: ooooo, that's neat. Do you have a source? I struggled to find a source for even that $6,000 number.

OP: Perhaps it is, but this article honestly took me 12 months to research & write. I really struggled to find trustworthy sources. Do you know of any others I should look at?

Also, I rather like the idea that blogging can simply be stating the obvious: https://blog.jim-nielsen.com/2026/blogging-stating-the-obvio.... Hopefully my argument is helpful to someone!

Totally agree; I had to cut a few sentences about that :). (I also tried to steel-man the paper as much as possible).

Oddly enough, it seems like, although the value of "blind operation" is well-understood, it's not super well researched. As one of the papers I cite puts it:

> Little research deals with the optimal design of haptic features and how haptic feedback can support the user in searching for control elements.

https://pmc.ncbi.nlm.nih.gov/articles/PMC6676796/

OP here: I always disliked touchscreens in cars, so I didn't understand why automakers kept shoving them in. I always assumed I was weird in some way, and that most consumers preferred touchscreens or something (Reddit seems to argue this in circles all the time). I planned to keep buying Mazdas, with their lovely buttons and stuff.

But when Mazda unveiled their button-lite 2026 CX-5 about a year ago, I started investigating.

I'm pretty convinced that touchscreens today are primarily a cost-saving measure, and every other justification is secondary. I hope I can convince you, too!

OP here: I really enjoyed learning about the history of industrial design for this.

I wanted to understand the parallels between software engineering and industrial design (well, chairmaking) through the Industrial Revolution. Matthew Bird's [History of Industrial Design series][1] was fantastic, especially this quote [from John Ruskin][2]:

> Now it is a good and desirable thing, truly, to make many pins in a day; but if we could only see with what crystal sand their points were polished,—sand of human soul, much to be magnified before it can be discerned for what it is — we should think there might be some loss in it also.

I hope you enjoy!

[1]: https://www.youtube.com/playlist?list=PLEANvcdpJW0lUKVPJ_8ee... [2]: https://www.gutenberg.org/files/30755/30755-h/30755-h.htm

I've been using "Big noun pages" in my technical documentation for quite some time now, and I wanted to formally explain the concept to my teammates. Thus: this blog post!

I hope you find it helpful!

Perhaps this is a bit OT, since the article focuses more on self-development ("When training a muscle, you only get stronger with resistance"), but I wonder about the subtitle:

> Every week there seems to be a new tool that promises to let anyone build applications 10x faster. The promise is always the same and so is the outcome.

Is the second sentence true? Regardless of AI, I think that programming (game development, web development, maybe app development) is easier than ever? Compare modern languages like Go & Rust to C & C++, simply for their ease-of-compilation and execution. Compare modern C# to early C#, or modern Java to early Java, even.

I'd like to think that our tools have made things easier, even if our software has gotten commensurately more complicated. If they haven't, what's missing? How can we build better tools for ourselves?

Many languages have very complicated pluralization rules. For example, Russian has different plural forms for numbers ending in 2.

That doesn't prevent a good translation framework from working properly, but it proves why the sample code in your example & the OP would not work.

https://www.unicode.org/cldr/charts/43/supplemental/language...

I agree on all counts! Some generated music is basically indistinguishable at this point, which scares me.

Also, kudos to HN’s title algorithm; it gave my article a punchier title. AI, right? :D

OP here: I thought it might be interesting to document how I (a random engineer) have been using AI day-to-day. I tried to ground everything with real examples, so you can judge for yourself :)

I started writing this the day before TikTok shutdown, but I kept dwelling on it for months. How much should we archive? Why?

The end result is TikTok will always be temporary. I hope you enjoy!