But with sites? I made a big picture spec sheet, and asked ChatGPT to expand on that. Then I fed that spec sheet into sites. 30 minutes later it spit out a perfectly working site of what I had tried to make a year ago. Not only that, it was much richer, and had more features. I asked for around 10 changes and new features, and all but one worked flawlessly. One time it managed to break functionality, but it was swiftly fixed. I didn't write or edit a single line of code. NOT A SINGLE ONE.

I've previously been on the buying side, and this was a website which could have easily ended up costing $10k-$15k with all the functionality. If not more. And I ended up getting that with my measly $20 subscription.

Web dev as a career is not dead, or dying, but I predict that A LOT of the people that would make those $0 - $10k sites are going to have a tough time ahead.

EDIT: This was also on par with my general experience with AI. For the longest time you can use frontier models, be unhappy with the results and tell yourself this sucks - and then out of nowhere some updates comes along, and suddenly everything just works perfectly well.

What is concerning, however, is when the competitors of these sports decide that their sport should become a filter for things adjacent to said sport.

The rationale is usually something like: "Competitive programming might not be a robust predictor for job performance, but most competitive programmers I know are also good developers, thus we must use competitive programming as a qualifying factor for hiring new developers." and weight it heavily compared to other signals.

Years ago I worked as a consultant on a projects which is quite similar to this, but larger scale and involving moving machinery. The task was basically to come up with sensors that would register when and which machines would be at what place in a structure. In the end we came up with a solution which only involved cameras and object detection...this was before AlexNet etc., so fun times with classical machine learning.

While it isn’t the toughest of trades, physically speaking, it will strain you neck, shoulders, and arms.

A lot of the work is pulling cables, and extremely mundane work. All that wiring in the ceiling you don’t see, was installed by someone.

Lots of terrible work environments, especially when it comes to industrial sites.

But in the end it feels rewarding. It felt like honest work, while also offering some stimulating work. 3 of 4 I started with pivoted into engineering within 10 years.

With that said, I’d rather become a automation tech. Basically the best parts of electrician, but more technical work.

Then I fed it to sites, and out came a working product.

A year ago I spent maybe 12 hours on making a specific website with tons of functionality. It was a process, to say it mildly.

Tried Sites now, where I fed the design specs in a MD file, and it one-shotted the entire website in 30 mins. Adding functionality has been a breeze.

Only once, so far, has any update broken functionality - but it was promptly fixed.

If for whatever reason you think you're being investigated or surveilled (who knows for what these days, could be enough to just associate with the various protest groups / orgs), you can just assume that the border will be where they'll try to clone / download your data (phones, laptops, whatever)

They all post vids of their vibe-coded dashboards where agents are doing market research and executing trades, while they kick back as collect to profits.

I've even seen these people promote some ASICs for home use, you know, to get those extra milliseconds shaved off.

After tech layoffs started, the get-rich-quick folks looked on what other fields pay well, and the decided that the absolutely most selective, the most credentialed and gatekept niche role in finance, that's the one to impersonate.

Basically, western labs are in it for the money / commercial monopoly. Chinese labs are in it for the tech? As long as they can keep distilling models, and get access to research other ways, they benefit. And if they can push western labs forward, they'll benefit from that themselves.

Comments URL: https://news.ycombinator.com/item?id=49909097

Points: 2

# Comments: 0

Back in the olden days, you'd emulate guitar cabinets using multiband EQs and filters, either analog or digital, sculpting the frequency response until it sounded close enough to the real deal. Then people started measuring the cabinet's actual response directly: send a known signal through it, record the result, and derive an impulse response. That IR could then be used with convolution to reproduce the cabinet's filtering accurately. Neural networks came later, mainly to model the nonlinear behaviour of amps and pedals.

That's basically the 15+ year progress we've had in the guitar world.

Say that you want to make a garage sale, but to get a better reach, you want to list item by item on marketplace. If you've never done it before, making 100 listings is such a pain in the ass you never want to do it again (I used to flip stuff for a living back in college, and it is a hassle)

At least with this, you can now just take pictures of everything, and tell the agent to research everything, list the stuff, and arrange for pickups.

Obviously not something I'd like to do for any serious deals or higher $ items. But for random stuff? I mean, it sort of beats hosting a garage sale or taking them to the flea market.

I'm not all pessimistic about this.

Even a very simple amp with 3 knobs that go from 0 to 10, would need like 11^3 = 1331 snapshots, if your parameter resolution is 1

For some amp like a JCM800, with the same resolution, you'd need 11^6 = almost 1.8 million snapshots, just for parameters like the front knobs. And then you have other things like tubes, bias, etc. Increase the resolution and number of parameters, and the number explodes.

Again, it is not a hype - but there are severe limitations if your goal is to make a digital twin of some amp. So just treat it as "snapshot of that exact tone", which you can then sculpt a bit with before the input and after the output.

There's so much money in it right now. There's such a momentum. There are zero incentives to slow down for those that are in charge.

I've accepted that in 5-10 years, the vast majority of human devs. and engineers will not touch a single line of code. It'll be small increments, with a couple of big ones here and there.

And there will not be any triumph for those that hold steadfast to the principle of human coding. They'll be tiny boutique shops that do custom stuff, in the same way cobblers are to the mega shoe factories.

Comments URL: https://news.ycombinator.com/item?id=49755152

Points: 2

# Comments: 0

But you know what, I’m glad to see that the fed is still operating independently. The doomer in me thought they’d fold to Trump, especially right before the midterms.

It also makes one wonder: What could AI solve if we managed to orchestrate billions worth of agents to take on a specific problem?

IMO the very best case scenario / potential for these are likely better than we think, but right now hidden due to logistical and financial reasons.

But if we assume that the model costs will continue to drop by a factor of 5-10 annually, there will always be a latency of a couple of years between what is completely out of reach, and what is financially viable.

Basically: If you knew AI could be affordable enough in 3-5 years so that even the most underfunded researchers could use it to solve cancer, how much would you value it now?

Software sales engineering could be another, which probably also pays more. But, alas, more competitive.

As a reference, for that kind of money one could put together a research group of 20-25 researchers, and keep them salaried for 5 years.

So while it is impressive, absolutely no doubt there, the SOTA access is so expensive that it is sort of unobtanium.

Luckily, the prices have historically reduced by a factor of 5-10 every year...but still, only those that swim in cash can afford this.