Low-resource open source developers pretty much screwed now

cracauer@ said:
All this AI craze has a very unfortunate side effect.
Click to expand...
<snip>
cracauer@ said:
With LLMs that changes. You either need a monthly subscription to an AI service (and the $20/month option doesn't reach very far), or you need hardware to run an LLM locally. You'd spend about $3500, which gets you either:
Click to expand...
You almost make it sound as if an LLM is crucial to hobby based software development. I beg to differ.

In fact, I'd argue that this isn't a problem at all but rather a huge advantage as it will stimulate potential developers to actually learn and develop their own development skills vs. simply relying on tools which more than often get things wrong anyway.
 
> video snips, future development 'changes' / 'options' / FUD et al.

1. I watched Frank Klausz use a powered router to help cleanup up some dovetails. My point being: Frank has more than likely hand cut and chiseled tens of thousands of dovetails in his lifetime and whatever tool he choose to use is just another "time saver" in his quest to "lower production cost"; his knowledge and experience is what makes the tool (a powered router, in this instance) only a tool (-e.g., he made the decision that the powered router was just a saw that can go sideways as well).

2. Stickley furniture was an answer to the mass-produced market (and more interesting Harvey Ellis' versions were even more desirable) and created a whole market.

3. How may "regular" loops had Duff written before/after he used switch?

...knowing is different. Ai is not a panacea (stop perpetuating FUD regurgitation's). I use a lower case 'i' because if I choose to use an LLM I am the "intelligence", not it.

Again, just because the power saw was invented, wood will still cut the same without one.

So, what specifically are you trying to discuss/talk about? Are you honestly saying that you think the playing field is leveled now?


> a huge advantage as it will stimulate potential developers to actually learn and develop

I'm not sure I'd buy into the idea that the "age of LLMs" will stimulate new developers seeking out specifically only hard-won knowledge more so, but I do believe the current artists will shine that much brighter.
 
Why has AI coding become essential?
Has development become a race, that the AI coder will win, because he is faster?
I'm not a professional developer and only code the traditional way for myself, so maybe I am missing something...

Does anyone else consider AI's comparably high energy consumption compared to traditional problem solving/coding?
 
The convenience of automation. NASCAR drivers do not shift their vehicles into autopilot nor do Formula 1 drivers.

Skill is required somewhere in this space you would think.
 
k.jacker said:
Why has AI coding become essential?
Has development become a race, that the AI coder will win, because he is faster?
I'm not a professional developer and only code the traditional way for myself, so maybe I am missing something...

Does anyone else consider AI's comparably high energy consumption compared to traditional problem solving/coding?
Click to expand...

For me it's not about the coding. It's about review, testing, maybe some utility functions, test cases.
 
cracauer@ said:
And that doesn't begin to touch on the second problem here, which is that those $1000 mainboard/CPU/RAM combos are gone now, the RAM is no longer included. Thanks to AI companies driving up the prices of RAM, SSDs and HDDs. So right now you cannot afford a chromium master for non-LLM work either.
Click to expand...
I can purchase a P4 system today for a fraction of the cost than I could back then. A pentium 4 is easily good enough for open-source development.
 
cracauer@ said:
And that doesn't begin to touch on the second problem here
Click to expand...
... a third problem can be also this: if you were one of the few developers/companies able to make a living thanks to OSS projects with commercial support, now with AI, it will become (probably) easier forking/replicating your code and/or steal good idea.

If you become successful enough, many big/medium-size companies can use their resources for replicating your business.
 
kpedersen said:
I can purchase a P4 system today for a fraction of the cost than I could back then. A pentium 4 is easily good enough for open-source development.
Click to expand...

Unless you have EMT64 capable late gen models, P4 is a legacy platform.

With 64 bit CPUs you can say you can tackle open source development but in some specific cases where your software or the tooling does not need the accelerated instruction sets that CPUs gradually introduced. Even then to use standard tools you will have to recompile most of it.

About the 2nd hand market P4 is far from optimal choice, there are newer boards+CPUs which have same "worthless" price tag.

cracauer@ said:
For me it's not about the coding. It's about review, testing, maybe some utility functions, test cases.
Click to expand...
That's a far cry from their tales of 10X-20X performance boost.
Yes AI can help in any of development phases, including assistance with programming and code generation.
However it will largely be used in just a few of phases, where the staff is thinned. Like you say, a developer needs test generation and reviews and code analysis, a graphics artist may need few lines of javascript for the site, and so on. Each of those used some other tools before AI, we used deterministic code analyzers, test generation frameworks and such. AI is better than these, but it is not a quantum leap.

This comes down to some XY% where XY < 50, but with a financial cost.
 
Espionage724 said:
What are people coding? Doesn't practically everything already exist?

I'm curious about a real example; I've hosted websites and ran stuff on a lot of OSs over 10 years, but never had to code anything.
Click to expand...
There's still a bunch of stuff that doesn't yet exist, there's a ton of software that likely will be written to cover things like assistive technology, new games, tools to wrangle hardware that doesn't yet exist and the like. But, that being said, there is indeed software to cover a massive number of things and the remaining needs tend to be less widespread than in the past.
 
Espionage724 said:
What are people coding? Doesn't practically everything already exist?

I'm curious about a real example; I've hosted websites and ran stuff on a lot of OSs over 10 years, but never had to code anything.
Click to expand...
Sometime there are permutations of the same idea. For example a viewer of Git commits can use a graphical GUI or the TUI. It can be written in C, C++, Rust, Zig or as a plugin for Emacs or VS Code. The GUI can use GTK or QT or the web. Same idea, but many variants.

Text editors is a good example too. There is the family of "vi-like" editors: vi, vim, Helix, Zed, kakoune, neovim, etc... Everyone replicate some idea and innovate in others.

This is a rather incomplete list of programming languages. The main PL needs libraries doing the same things of libraries of other languages, but using the paradigm and approach of it. So for example a library for connecting to Postgres or SQLite for Java, C#, Zig, Rust, Go, etc...

Do you like the Lisp language? There are Lisps running on the JVM, on the Lua VM, on the BEAM or compiling directly to rather idiomatic C. Every run-time has different trade-offs, and it can make sense having them.

This is the list of the Apache projects. There are many big-data engines. Some work in RAM, other on the object-store. Some are good at batch processing, because they collect different queries and discover parts of the plan shared between them, and walk the data only one time. For every combination of features, storage, usage scenario, there is a different engine. Was Kafka message broker successful? Then they created Kafka compatible servers written in C++ that are faster than the original one.
 
drhowarddrfine said:
On the news today while driving around, they said there's some report out there stating that 95% of companies using AI find they are losing time and money by using AI tools.
Click to expand...
There have been a bunch of similar reports, for example here https://capitalnexus.substack.com/p/ais-reality-check-why-95-of-companies
AI generating big returns on the bottom line for their clients is an as-yet unproven hypothesis.

And there was a major wobble in the stocks bubble this week; south korea is particularly exposed because of their major dram and flash manufacturing companies.

Like the dot-com boom, winners will eventually emerge, I guess. At the moment everyone is still piling in, because of FOMO.

Low barriers to entry in terms of buying a cheap PC was always a big attraction of software; all you needed was the skill, a bit of business naus, some vision, and your time. The barriers to entry are now being raised again, if you need to spend serious amounts of money on AI to be able to compete. It's pretty sad to see. A lot of good companies got started on a shoestring, by a couple of guys in a garage with just a home micro and some electronics, and that won't be so easy in future, if they need to spend hundreds or thousands a month on AI.

So in the longer term, eventually, we might find that AI as a service becomes commoditised, in the same way that hardware is now; but the barrier right now is how the AI provider companies are ever going to recoup the huge capital spend and vast energy and water running budgets, let alone turn a profit, to justify their astronomical share valuations. They have to make things like self-driving real, for example, so that they can make redundant 20 million drivers and realise billions of dollars in profits as a result. The same goes for software developers; where you previously had 20 software guys, now you only need 2, and one of those is a desparate unpaid intern who will work his arse off for nothing, in the hope that you will eventually give him a job. That's the dream they are selling to the investors. Time will tell whether it's ever going to happen.
 
blackbird9 said:
There have been a bunch of similar reports, for example here https://capitalnexus.substack.com/p/ais-reality-check-why-95-of-companies
AI generating big returns on the bottom line for their clients is an as-yet unproven hypothesis.
Click to expand...
I mainly agree with what you wrote, but as side-note, a survey about AI returns, made more than 9 months ago can be completely unrelated to current situation.

For example in this interview a seasoned Linux kernel maintainer said that until 1 month ago AI reports were slop. But now, they are very useful: 2/3 of the signaled problems are real and the proposed patches need only some cleaning; 1/3 are rejected, but they are not complete false alarms.
 
Yeah, it's changing rapidly, with each new generation of AI. Each new model is massively more capable than the last, as is each new generation of the hardware it runs on. That's the bet that investors are making; that sooner or later, it's going to become real, and you will be able to replace large numbers of "lower value human capital" with machines, at which point, you stand to make a very large amount of money.

So you dare not be out of the market, for Fear Of Missing Out. Hence the crazy bubble. I think this is bigger now than the dot com boom ever was.

Of course there's also a geopolitical angle to this thing. The US and China are in a headlong battle for AI supremacy. Whoever wins, will likely dominate the world. It's a competition that neither side can afford to lose. Which means that money is going to continue being thrown at it, no matter what happens.
 
blackbird9 said:
And there was a major wobble in the bubble this week; south korea is particularly exposed because of their major dram and flash manufacturing companies.
https://www.cnn.com/2026/06/23/business/stock-market-kospi-dow-nasdaq-ai MSN
Click to expand...
That's old news—pun intended ;)
Even though CNN's message states that it has been updated: "Updated Jun 24, 2026", my guess is that it will be updated today again. Yesterdays aftermarked numbers of Micron are staggering: Micron. This drives tech markets, especially in the AI sector; the Nasdaq-premarket is already up by +2%. Micron, SK Hynix and Samsung are the three major memory producers in the world. Markets are fluid and individual stocks, especially anything AI related, fluctuate, a lot.

The "problem" with the KOSPI index of South Korea is that SK Hynix and Samsung together represent an overly big part of that index, I believe between 40 and 50%; for an index that is rather uncommon.

Bottom line: DRAM prices will likely remain high for the forseeable future.
 
Erichans said:
Even though CNN's message states that it has been updated: "Updated Jun 24, 2026", my guess is that it will be updated today again. Yesterdays aftermarked numbers of Micron are staggering: Micron.
Click to expand...
We should fund OSS with OSSGS: Open Source Stock GitHub Stars. I sold 50 Ruby-on-Rails stars 5 years ago and bought 500 Elixir-Phoenix stars, and I earned a lot of money in the process.
:)
 
Erichans said:
That's old news—pun intended ;)
Even though CNN's message states that it has been updated: "Updated Jun 24, 2026", my guess is that it will be updated today again. Yesterdays aftermarked numbers of Micron are staggering: Micron. This drives tech markets, especially in the AI sector; the Nasdaq-premarket is already up by +2%. Micron, SK Hynix and Samsung are the three major memory producers in the world. Markets are fluid and individual stocks, especially anything AI related, fluctuate, a lot.

The "problem" with the KOSPI index of South Korea is that SK Hynix and Samsung togethere represent an overly big of that index, I believe between 40 and 50%; for an index that is rather uncommon.

Bottom line: DRAM prices will likely remain high for the forseeable future.
Click to expand...
You're right, it really is 'old news', in 'markets time'. Yes, there's been a "memory bounce", micron results have been phenomenal, and now it's all back on, so pile-in again, and "buy all the (semi) things"! Don't worry, be happy. 😂

There was another little glitch a few days ago, when it was in the news that China might have acquired a latest generation ASML lithography machine. I think ASML dropped a bit that day! I'm sure that names were taken, too.:-)
Yeah, "RAMageddon" isn't going to end, any time soon.
 
Zare said:
Unless you have EMT64 capable late gen models, P4 is a legacy platform.
Click to expand...
I don't believe a legacy platform has any bearing on open-source development. Much of the code underlying open-source is much older than a P4.

A P4 would still provide a stronger development environment than a modern Raspberry Pi. Especially since C long ago solved the gap between porting to different processors of varying ages.
 
kpedersen said:
I can purchase a P4 system today for a fraction of the cost than I could back then. A pentium 4 is easily good enough for open-source development.
Click to expand...
Every piece of hardware is "easily good enough" for some open-source project. The problem is that it is "easily good enough" finding many open-source projects for which the P4 based system is not good enough. The 99% of them. In the 1% there is awk, vim, some plan9 based project maybe.
 
You must log in or register to reply here.
Back
Top