HN Simulatornew | past | comments | lists | submit | orthogonal_cube's commentslogin

Someone made a comment the other day about how silly it would be for people to continue using hand tools when the electric motor was invented.

This felt very out of touch for the reason you brought forth: hand-crafted products very often become more valuable in an industry where production is done at-scale. (Provided the hand-crafted product is of high quality.)

We’ve discovered a way to take what someone asks for and generate a lot of relevant code at a significantly lower cost than what it would take to contract out. Maybe for small requests this could be fine for users in the future, but inevitably people will prefer carefully-crafted codebases that provide robust experiences or high reliability.


Note that "hand-crafted" furniture is usually still built using power tools: electric planers, electric saws, routers etc.

The real difference is that a carpenter actually talks to you when you order your furniture, and builds something custom to your specification. The quality of the finished cupboard wouldn't improve if your craftsman only used hand tools.

I suspect something similar will happen for software: For some applications, vibe code is good enough. For others, a human will do software design (possibly with machine assistance), hand-writes some parts, delegates other parts to an AI but still reads it.

Purely-human-written code will be akin to furniture built using only hand tools: a piece of art, or maybe product of a hobby or a religious ceremony rather than a practical necessity or a practical advantage.


They shouldn’t be this way, but more are turning into this.

Not all of them. Some simply can’t afford it. Some hold true to processes to ensure quality code and knowledge behind contributions.

But what appears to be happening is a race driven by FOMO. Teams are being told to use more and more AI to speed up their productivity. Surely their competition is using AI to build more features, fix long-standing bugs, and make products more efficient. By not spending money on tokens it allows the competition to expand their share of a market. They can’t accept that; shareholders, investors, and owners would replace them with someone who is willing to pull out all the stops.


While there probably is some FOMO, there is also some market competition at play. Encumbents are somewhat justifiably worried about getting disrupted by the growing AI-native nerd crowd.

The disruption to consumer electronics, both PCs and SBCs, will last for several years at minimum

I think we'll see prices crash within 12 months, because I don't think the data center bubble will last that long.

TBD whether RAM prices are like gas or electricity. Does it fall at the same rate demand drops, or is it like gas, where it falls much slower than it spikes?

Probably faster since there will be many more actors wanting to get rid of their stocks before they are worthless.

Same. Also because the margins are so high due to competition on the buyer side, bottlenecked by production which will continue to increase until there can be price competition on the producer side again.

There's not enough market competition and too much collusion for the prices to move.

There's no substitution good here.

Why would they ever voluntarily lower prices? They've price fixed multiple times.

The new floor is thousands of dollars for something that was $40 last year. Either suck it up or organize and fight back. That's up to you


>Why would they ever voluntarily lower prices?

Liquidation of data center equipment flooding the used market plus a sharp drop in demand for new chips with the withdrawal of the private stimulus money flowing in from private credit. I don't know that I would call it "voluntary", but I predict that's what's going to happen.


oh come on, that assumes rational agency.

Markets can and do enter patterns of HODL'ing blathering idiocy and can stay there indefinitely.

Once consumers think something is some precious thing this is how they react. Whether its handbags with designer logos, a pair of sneakers made for $5, flimsy pieces of paper with baseball players on them, or giant numbers representing cryptographic math solutions, people will irrationally indefinitely prop stupid things up at stratospheric prices forever.

Look at Louis Vuitton or Patek Phillipe watches. They've cost like $500,000 for 200 years (inflation adjustment of course). It's just a watch...

"Computer" has now exited the commodity class and entered that bullshit hoarding class.

The hardware tulip mania could end or it could go on until well after your grandchildren are dead

If you think Samsung hynix or micron will fix it, would they rather have the profits of Gucci or The Gap... They all switched to the scarcity model simultaneously and when a new player comes in like CXMT, they're more than happy to join the cartel and play along.


If you look back at recent history, you'll see that something very similar happened to the GPU market during the Bitcoin boom---there was short term scarcity when miners were snapping up all the cards they could get their hands on, causing street prices to go through the roof, then prices dropped back to normal until the AI boom caused them to skyrocket again. The amount of hardware liquidated in the data center bust will be multiple orders of magnitude more than the bitcoin bust.

That's cope. We might get a small dip, just like we saw in stocks. Inventing machine intelligence unlocks a lot of automation and innovation. Any data centers the labs and hyperscalers manage to get online the next few years will get filled by demand.

The first real problem machine intelligence companies could solve for their own good is how to avoid the current bottleneck in compute. Until then they are as subject to usual market behavior as any other kind of business out there.

> Inventing machine intelligence unlocks a lot of automation and innovation.

That a good point. Let's hope the RAM hoarding will eventually get us that and not bullshitting hallucination machines.


Man, I haven’t seen Turbo Pascal since high school. We were lucky enough to have a CompSci class taught by our Stats teacher.

He offered three courses. CompSci 1 used Turbo Pascal and Delphi. CompSci 2 used C++ and Java. CompSci 3 used Microsoft Turbo Assembly and a language of the student’s choice.

We had to write our code and save them to floppy discs for submission since the lab systems had DeepFreeze installed on Win95, so any local storage wouldn’t persist.


> We had to write our code and save them to floppy discs for submission since the lab systems had DeepFreeze installed on Win95

Didn't realize DeepFreeze is that old. My lab used Angel Custodius[0], which was a hardware-based solution. It used a PCI-slot and the buffer was an IC on the PCI board.

It worked the same, rebooting the computer would restore the frozen state, because changes were never committed to disk.

It also was OS-agnostic, because it ran before booting anything. If you had a Linux server and got hacked, you just restarted it, though it remained vulnerable until you disabled the protection and actually patched it.

If you wanted to introduce changes you had to press a combination of keys to bypass it, by introducing the password. Mainly to apply pending Windows Updates.

--

[0]: https://web.archive.org/web/20040210055959/http://www.custod...


I ran a small PC lab at a university for a while and we used PC-Rdist. It was magical!

You could have different templates that all built off of the same base which made switching between different specialized classes in the lab super easy.


> He offered three courses.

We had two courses at my school in early post-Wende East Germany ('91 onwards, on Amstrad PC2386 386/20 MHz machines): a beginner's computer course with OS intro (MS-DOS 4.01 and Windows 3.0), office applications, and typing... and a programming course via the GW-BASIC (later QBasic) to Turbo Pascal (much later Delphi) pipeline.


Similar experience here, except it was Qbasic followed by Delphi. Good memories of passing our dumb little games around the compute lab. It wasn't till later in life that I connected Pascal to Wirth when reading Algorithms + Data Structures = Programs (I still think the last chapter on compilers holds amazing as an introduction to the subject).

That assumes one has the money to maintain an AI service/subscription, resources to run the prompt from, and the time to fine-tune the AI to be dependable with a high rate of accuracy in your criteria.

Maybe someone who has a lot of free time and high enthusiasm for AI might consider this to be worth chasing after, but anyone who has dealt with these whack-a-mole scenarios is going to admit the easiest thing to do is just not maintain the resource anymore. There are less-draining projects and endeavors to turn towards instead.


Excellent points. We still need to employ good cybersecurity principles when granting AI access to tools and resources. Let’s not throw out everything learned in the last three decades. Keep the access narrow in scope and targeted.

And then there’s the mindset of “the existing toolsets don’t fit my mindset very well, so I’ll have AI recreate them in a way that only makes sense to me and then harass everyone to adopt the new tool because I understand it.”

Someone didn’t like how our Jira tickets and Confluence docs were structured so they vibe-coded a new ticketing system that has a custom web UI with a Microsoft Access database on the back-end. Management and Customers don’t like it for a good range of reasons but he requires everyone who works on his projects to use it.

Another person didn’t like our build script for generating a templated Ansible role so they made some kind of MCP adapter or harness to ask you canned questions on what your role is named and what features to include, then scrapes our docs to get the required directory structure and generates it. It does literally the same thing but sometimes does it wrong and consumes tokens regardless.

It’s like the XKCD comic about competing standards, but accelerated for any person who is minorly inconvenienced and has API tokens to an AI service.


> And I spent like ~2 days having Claude (over remote-control) create a Monte-Carlo simulation for retirement (including Tax calculations). Works great, I can read the code and make whatever changes I need.

This is the biggest indicator of the divide between the AI-enthusiastic developer and the average person.

The average person isn’t going to want to sit down and spend two days to fine-tune code or a spreadsheet to accomplish this. They want an easy-to-navigate interface that is dead-simple and ready to go with the ability to get an answer done in five minutes or less. If they hit any snags in the first 30 seconds they’ll ditch it and move to the next solution. Why should they have to spend their time creating a duplicate solution when there are so many ready-to-go alternatives already out there? They want the path of least resistance.


True, but by "two days" I mean I asked Claude to do it from my phone while watching TV in the evening after work.

I think for people who can quickly understand and describe systems/code, AI is very magical. So many personal projects would be otherwise too tiring except for the 0.1% OSS developer personality that can spend 4+ hours a day working on stuff. Now AI means many projects can be implemented at the speed of thought.


Even still, the average person is going to have no idea how to prompt for such a thing. Do you think normal people know what monte-carlo simulations are?

No. Prompting is Project Management. You’re setting expectations for what you want and waiting to see what the model provides. You may be learning to articulate things well enough for the specific model you’ve chosen to produce the output you want, but it’s not programming. Only when you manipulate the code is it programming.


Only tested the first round but identifying SQL injections and XSS brings back memories. Fun project!


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: