tudorizer

This parallel is something that I've been mulling over for the better part of this year.

Are we simply getting old and bitter?

Personally, I would add a previous cycle to this: social media. Although people were quick to point at the companies which were sparked and empowered by having unprecedented distribution.

Are we really better or worse off than a few decades ago?

The one-time price is interesting.

If the platform goes away in 1 year, it essenetially becomes 39$/year.

Any plans on how you'd make this a longer lasting product?

The gist of the article is a fun thought experiment.

Why count lines of code? Error handling is nothing to sniff at, especially in prod. Imagebin had a small handful of known users. Open it up to the world and most the error handling in Go comes handy.

For PHP, quite a bit was left on the shoulders of the HTTP server (eg. routes). The final result of Go is a binary which includes the server. The comparison is not fully fair, unless I'm missing something.

same. Would have prefered a lo-fi stick figure drawn on a napkin. The cartoon Max detracts from the rest of article, which is a good read.

> Vibe coders are clients seeking an alternative, not developers.

Agreed. That's genuinely a good framing for clients.

I accept any equivalents of reading comprehension tests to prove thay I watched the video, as I have many of Andrej's in the past. He's generally a good communicator, defo easy to follow.

Sure, I wasn't sure how to call the robot layer. Is is "Iron Main Suit"?

Entertain me in an exercise:

First, instruct a friend/colleague of how to multiply two 2 digit numbers in plain English.

Secondly (ideally with a different friend, to not contaminate tests), explain the same but using only maths formulas.

Where does the prompting process start and where does it end? Is it a one-off? Is the prompt clear enough? Do all the parties involved communicate within same domain objects?

Hopefully my example is not too contrived.

TIL. Thanks for sharing

Agreed. This e-commerce example is quite a good highlight for LLMs.

Same can't be applied when your supplier needs 300 68 x 34 mm gaskets by the BS10 standard, to give a random, more precise example.

> That's a lot of people to talk to in a day more or less, since the talk happened. Were they all there and you too, or you all had a watch party or something?

hehe, I wish.

The topics in the talk are not new. They have been explored and pondered up for quite a while now.

As for the outcome of the donut experiment, I don't know. You tell me. Apply it repeatedly at a big scale and see if you should alter the initial offer for best outcomes (as relative as "best" might be).

Fair. No process is 100% efficient and the depths of many topics become ambiguous to the point where margins of error need to be introduced.

Chip fabs are defo far into said depths.

Must we apply this at more shallow levels too?

Can we have it now and skip 3.0?

Might sound like a generic tangent, but it's the conclusion people will leave from the talk.

> Did you talk with them? Otherwise this is just creating an imaginary argument against some people you just assume they didn't listen.

I have, unfortunately. Start-up founders, managers, investors who taunt the need for engineers because "AI can fix it".

Don't get me wrong, there are plenty of "stochastic parrot" engineers even without AI, but still, not enough to make blanket statements.

Working code or not is irelevant. Heck, even human-in-loop (Tony-in-the-Iron-Man) is not actively the point. If we're going into "it's all about" territory then it's all about:

- training data - approximation of the desired outcome

Neither support a good direction for the complexity of some of the system around us, most of which require dedicated language. Imagine doing calculus or quantum physics in English. Novels of words would barely suffice.

So a context window as big as the training data itself?

What if the training data is faulty?

I'm confident you understand that working code or not doesn't matter in this analogy. Neither does LLMs reaching out for the right tool.

LLMs has its merits. Replacing concrete systems that require a formal language and grammar is not.

`1 + 1 = 2` because that's how maths works, not because of deja vú.

I watched the entire talk, quite carefully. He explicitly states how excited he was about his tweet mentioning English.

The disclaimer you mention was indeed mentioned, although it's "in one ear, out the other" with most of his audience.

If I give you a glazed donut with a brief asterisk about how sugar can cause diabetes will it stop you from eating the donut?

You also expect deterministic outcomes when making analogies with power plants and fabs.

95% terrible expression of the landscape, 5% neatly dumbed down analogies.

English is a terrible language for deterministic outcomes in complex/complicated systems. Vibe coders won't understand this until they are 2 years into building the thing.

LLMs have their merits and he sometimes aludes to them, although it almost feels accidental.

Also, you don't spend years studying computer science to learn the language/syntax, but rather the concepts and systems, which don't magically disappear with vibe coding.

This whole direction is a cheeky Trojan horse. A dramatic problem, hidden in a flashy solution, to which a fix will be upsold 3 years from now.

I'm excited to come back to this comment in 3 years.

amazing!

I just installed it and enjoying the integration.

'tis a broad umbrella, but my co-founder and I are deep into transforming excess green energy (wasted because of curtailment) into ML / AI compute.

Focusing on training and fine-tuning, because of lower bandwith requirements.

We're collaborating with a few AI researchers through offering a sponsorhips program for the right people. If you know anyone, send them our way.

PS. If anyone has experience with Marimo, give me shout!

[1] https://enverge.ai/

so where is this actually hosted? couldn't find on the home page

On a quick glance, it seems it's possible to run this as a service similar to JupyterLab, right?

I'd be keen to offer it as an alternative to Jupyter on my little GPU platform experiment.

I wrote this article as a preamble to a collection of tools and resources titled [ML Development Cycle](https://tudormunteanu.github.io/ml-development-cycle/).

It's meant to be open and public.

So far I received quite a bit of feedback and hopefully, it can only get better. Thanks!

amazing!

How do you gauge "GPT-4o-level assistance"? Have you tried side-by-side comparisons?

why would it be satire?

Indirectly: choose boring tech because it's well documented, widely supported, and training data for LLMs abound.

Do you run LLMs locally? I'd be keen to have an overview of your stack.

Hello, fellow CTO Craft member!

People do like to mention those faxes a lot, don't they? Or pocket translators.

I think there are other uses of tech far beyond it that are far better integrated in the Japanese society that in the western world.

I doubt that's the most important part. Maybe for yourself.

The overall tone and high-level business decisions are much more interesting.

Love the fact that it can essentially API-ify the uploaded CSVs.

What's the expiration time on pasted CSVs?