I like to say AI is like the super soldier serum from the MCU. Teams take it and either become Captain America or Red Skull, it depends on what you were to begin with.

I wish. The "captain Americas" are more like Tony Stark's jury rigged suit used in the desert (for now). It's not the suit truly doing innovation, it's the man inside who knows the suit (and its many limitations/flaws) inside out. Using the strengths long enough to escape before the hunk of metal inevitably fell apart.

The fact that it’s called “the quick pitch” screams Claude-written pithyness pulled from some context that doesn’t match the article’s style (like investment pitch decks).

The whole article screams Claude-written.

tired: this meeting could've been an email

wired: this company could have been a prompt

As a tech educator I 100% agree. LLMs are not going to become a "new compiler" where we don't have to worry about the code any more. There's a reason we trust deterministic systems.

I've been worried about this a lot, I even created an agent skill called do-i-understand that's designed for novice devs (and experienced too, because atrophy) where the LLM asks you questions about the PR you're about to submit. I've found it helps a lot: https://github.com/AnthonyPAlicea/skills/blob/main/skills/do...

One way or another, there will be a skill reckoning.

We don't run our own unit tests on compiled code because it generally can be trusted. LLMs are an extraordinarily leaky abstraction.

Saying "as long as it works and tests pass" suggests that we can test for every possible scenario. We can't. And tests can be flawed on top of it. Which is why an LLM is no more a compiler than a human coder (as you say) is.

Why do you trust the non-deterministic LLM with your skill?

It absolutely might make mistakes. But the skill requires all questions to be based on lines of code with specific line numbers which the human needs to review. That's intentional to serve as a safeguard against hallucinations.

A deterministic LLM is possible. Would you trust it?

they probably mean deterministic in the sense of traditional computer programs consisting of if-else decision points and ordinary cpu logic, as opposed to numerical models (which can of course be deterministic too, as you pointed out)

Model inference is just that same logic but faster.

I think the point is determined. No-one can determine what the LLM might do.

> Model inference is just that same logic but faster.

This statement is nonsensical.

There are 2 levels of determinism that matter here IMO. Level 1, given the same input you get the same output. Level 2, you know what to change on the input to get a specific change on the output.

LLMs are currently capable of level 1, but not capable of level 2. Trust follows what is actually deterministic.

Trust should, but instead is being gulled by those incapable LLMs. And I don't see that changing, given the typical human's seemingly unlimited vulnerability to this con.

I mean in the traditional software sense. We trust autopilots to fly planes. Would you get on a plane flown by an LLM?

Then I think I'd rephrase you: There's a reason we trust trusted systems.

:)

Even their fans agree, LLMs are inherently unreliable. The fact some people trust them regardless is due to a deep flaw in human psychology that I believe has not been significantly exploited by any previous tech. There will be tears.

I think two things contribute to the trust in LLMs:

1. Decades of generally reliable systems (I trust my calculator because it always says 1+1=2) has trained people to believe what computers say.

2. LLMs "speak" with great confidence, which has, as you say, a psychological effect.

I agree.

I’ve done a lot of work with the Office file formats over the years, flipping to .zip and messing with the format manually. Lately, because I’ve made courses with 500+ PowerPoint slides with lots of morph transitions, I iterated to find an approach that generally worked with Claude but is never perfect (using the Claude for PowerPoint plugin).

I switched to https://bento.page/ when it showed up here on HN and that’s been much more accurate I think because HTML and CSS is even more represented in training data.

First code I ever edited.

First time I played it was 8th grade, in Randy Seaman's Algebra class. He also had wolf3d on that machine.

Google switching to hallucinating AI summaries has been to me an absolutely shocking abdication of care for both their users and their own reputation.

It extends beyond search as well. I have had multiple incorrect Gmail summaries that, if I had only read them instead of the actual email, would have resulted in financial harm.

> Buttons regularly perform navigations. Clicking a logout button navigates the current page to a logged-out one; clicking a “search” button navigates the current page to the query results.

This seems to conflate appearance with semantics. If an element causes a navigation, I make it a link. Whether it looks like a button is irrelevant, that’s CSS.

I always choose one or the other by intended behavior first, and that always works out great.

That said, I like the idea.

> This seems to conflate appearance with semantics. If an element causes a navigation, I make it a link.

Neither of my examples work with links. Logout buttons must be buttons because they trigger unsafe requests, and "search" buttons are buttons because they submit the form as a query parameter. Both of these examples, are, however, navigations.[0]

> Whether it looks like a button is irrelevant, that’s CSS.

It's not irrelevant, because even if you style a link like a button (or vice versa), it will still behave like the thing it actually is. That's why it's important to have native HTML behavior that lets people use the right semantic element.

[0] https://html.spec.whatwg.org/multipage/browsing-the-web.html...

“Logout buttons must be buttons because they trigger unsafe requests”

But this is just more of the pedantic developer nonsense that everyone is complaining about here.

I’ve built tons of sites in my time, and the logout “button” is always a link to “/?logout=1”. In 30 years, this has never caused a problem.

Well, the idea here is “things that look like buttons do page navigation.”

And the proposal is that we should more closely align the presentation and function.

You are correct that if you don’t care about alignment, there is nothing wrong with the status quo.

Presentation in what medium? A screen reader? You’re talking the user-agent stylesheet, which is a particular chosen presentation that I can change with my own CSS. I can make a button look like a link.

You’re suggesting aligning to an inherently and deliberately changeable presentation.

The idea is fine, but worry about what an HTML element “looks like” is not a good reason.

> You’re suggesting aligning to an inherently and deliberately changeable presentation.

Yes, you have understood the proposal.

Imagining you have a navigation to “/logout” that server side invalidates your token and redirects to home, would you implement that as a link? If so, what happens if someone tries to “open in a new tab” it?

Yea I think a form that posts is the right tool for the job here. It is clearly unsafe semantics.

All those .BAS files were an education.

I did the same in QBasic (and Tandy Color Basic)!

I feel like the old Choose Your Own Adventure books were a sort of precursor to programming for a lot of us. Basic if/then/goto.

For me the spark is reinvigorated. I can try ideas before I didn’t have the time for, prototype and test and only pursue what proves to be worth the time.