dale_glass

This kind of rhetoric confuses me. Do you hate watch OpenAI (or whoever) ads or something?

I haven't a clue what the peddlers are saying because I run uBlock Origin and do my best to never see an ad. I don't care or even know what they're advertising, and don't take my guidance from big tech CEOs.

I don't really generate big chunks of code, but in general I found AI outputs to be quite sensible. It knows what the main modules are and uses them, I've not seen any as egregious as making a JSON parser from scratch without a need.

You can always exercise decision making.

Realistically, people do a lot of Googling "parse Json in Python" and going with whatever pops out, like "import json" probably. But there's other options out there.

With a LLM, you're almost certainly also going to get the standard json module if you just ask for a JSON parser.

But nobody says you have to stop there. For instance if I say I want "the fastest JSON module", it goes for orgjson instead.

And if I wasn't happy with that, I could ask it to just write a benchmark using every module available and actually measure what's optimal for my hardware and my data file.

Choices don't go anywhere, and a smart user can take advantage to evaluate possibilities for whatever factors might matter -- speed, size, number of dependencies, etc. I'd say smartly used LLMs make for better choices because you can do research much more easily. Benchmarks, prototypes, asking it to search for you whether a given library has a good reputation or not and is well maintained, etc.

Being realistic, a lot of decisions pre-AI work out to "Because that was what was on top of Google search".

Also there doesn't seem that much discussion of code in there.

> And what if you start digging into it? Won't other people say - "Hey Euler go back to ploughing the fields? We can have Djinn's answer all the math questions but this field isn't going to plough itself."

Seems like an overly pessimistic take to me.

If you have a magical Djinn happy to answer questions, why wouldn't you badger it a bit for the details? I can't imagine Euler of all people just wanted the answer.

Where's that from?

Now math isn't my wheelhouse, but every LLM I've seen provides far more than just the final answer and is quite willing to answer questions.

Almost all of it, yes.

The packet format is awful, full of complexity, backwards compatibility mess, and serious security problems.

The spec is pretty awful in that a lot of the complexity is due to keeping compatibility with things that are a terrible idea. And also in that the general concept of a long lived key turned out to be a bad one.

The library or lack thereof IMO cripped PGP/GPG back when it was relevant. It's nigh impossible to make tooling for it, or it was until Sequoia showed up, which was extremely late. Things like doing stats on keys, or doing anything but what the binary wants to do were terribly painful until Sequoia.

The Web of Trust was an interesting idea for a while, but failed to transition to the modern internet.

By the time you fix all the things what remains looks nothing like PGP, so it amounts to "all of it is bad".

Was there anything besides the Lytro?

Its problem seems to be that it was a solution to a problem that was fast disappearing, at the cost of unacceptable tradeoffs.

The Lytro sacrifices a ton of resolution and produces crappy images, in exchange for being able to refocus. And they did that in an age where excellent autofocus was already available.

These days we have cameras with AI models that recognize birds and animals and focus flawlessly almost every single time. And they don't sacrifice any resolution to do it. And they are proper, comfortable cameras rather than a weird stick.

My advice would be: don't do just math. Pick some practical hands-on skills, concentrate on things that are widely applicable in the real world and are valued by people outside of your own field, and be ready to shift sideways.

Do programming, electrical engineering, control systems, robotics, ML, statistics, etc.

Eg, we're probably very far off from having factories or new devices be entirely designed and built by robots. Somebody's going to need to get their hands dirty for quite a while still.

How about overbuild and interconnect? You don't have to overbuild for your personal worst case if you can buy power from somebody else if it comes to it.

It worked quite well for me!

The trick though is to gravitate towards places where there is a community and where you end up talking to a few of the same people. HN's probably not going to be it, an IRC channel, Discord, forum, etc, where you keep talking to the same dozen regulars is much better.

You can kill a battery much faster than 5 years if you don't treat it right.

My personal experience is that modern devices are far better about battery management, my phones used to wear out much sooner than 5 years back when smartphones were a new concept and Li-Ion was new and fancy tech. So all that "faffing" seems to be doing something useful, because even my quite old and forgotten tablet that I forget to charge seems to be doing quite well.

Things I don't like:

* It's not standard. You can't plug in some normal formatting library there and be done with it.

* You'll probably end up having to write conversions back and forth.

* It's fragile -- if anything gets misaligned it'll break

* It's tooling unfriendly. Think things like Nagios, Grafana, etc sending formatted notifications. Everything can do this or <b>this</b>, but practically nothing is set up to accommodate this format.

* It leaks into other subsystems. You can't eg, just search/replace/insert/delete words if you need to for any reason without breaking this.

Is it the worst thing ever? No, but it seems to go with the pattern that nothing in XMPP is normal or comfortable. Everything has this weird particular flavor to it and more complexity than necessary.

If it were up to me, my option would be to support 2 things: Markdown for simple cases, and embedded HTML for when you really have to get fancy. Each of those can be handed out to existing code and you don't ever have to keep adding extensions because what if people want colors or something now.

What part part, sorry?