Vibe coded plugins

[Urs]

Maybe, if established guys are doing it, they have issues in the current market, and need to survive. The market may be shittified for them, so they shittify it even more?

I guess we're in some kind of race. One can gamble that Anthropic et al magically become a viable business model, and software development/markets as we know it are done - "AI takes over". Or, IMHO more likely, LLMs hit a wall, Apple wins and everyone runs what limited intelligence is necessary on their M5+, "AI becomes a tool". Or, somewhat likely if Claude stops wiping hard drives, few people/companies can afford the winning model in the cloud, ten dollars per token, while most people are happy with the open source model on their M5+ ("AI as supercomputing solution").

[audiojunkie]

Maybe, if established guys are doing it, they have issues in the current market, and need to survive. The market may be shittified for them, so they shittify it even more?

I guess we're in some kind of race. One can gamble that Anthropic et al magically become a viable business model, and software development/markets as we know it are done - "AI takes over". Or, IMHO more likely, LLMs hit a wall, Apple wins and everyone runs what limited intelligence is necessary on their M5+, "AI becomes a tool". Or, somewhat likely if Claude stops wiping hard drives, few people/companies can afford the winning model in the cloud, ten dollars per token, while most people are happy with the open source model on their M5+ ("AI as supercomputing solution").

That's really all it is right now... AI is just just a tool--even if it never advanced beyond what it currently is today, and LLMs hit a wall and stopped advancing. It's not a perfect tool. And, it definitely needs the knowledgeable, experienced developer to be involved throughout the entire process if it is to be used correctly. But the potential to use it to quickly do the dirty work and then have it checked and vetted by an experienced developer is exactly what is going to speed up the development process and improve the code (think bug finding). This is for those who are actively involved in AI Assisted Development, of course, when compared with those knowledgeable, experienced developers that choose not to do so (we won't even mention Vibe Coders). Time is money. Getting a deliverable product as fast as possible to consumers is money. Having the extra time saved that can be then invested into the business in other ways, is money. Those companies who learn to leverage AI as the tool it is, flawed though it may be, have the definite potential to excel beyond those who choose to shun it (with all other things being equal).

[jamcat]

It’s always been the case that anyone with an idea could create it, if they’re willing and able to invest the time and money to make it happen. The only difference now is AI has truncated the amount of time and money required. Instead of spending $120,000 to hire a programmer for a year, you can now have AI do it in a month for $20. So really it’s no different other than the barriers have been lowered.

Urs seems to be wishcasting for those barriers to be restored. But that’s not going to happen. Genies don’t get put back in their bottles.

In the end, it’s the same things that have always mattered that still matter: the quality of the idea, and most importantly, the reach of the marketing. AI can’t help much with either.

[TechHaus]

for $20

Nope

[Tiles]

You could even work for free. You can run Google Gemma 4 and the new Qwen 3.6 35b locally in LM Studio or Ollama, given your hardware allows it. It's not Claude, but can already be pretty useful for smaller tasks. And is a big step forward.

[whassup]

Instead of spending $120,000 to hire a programmer for a year, ...

Thanx for that laugh!

[jamcat]

How much do you work for?

[ksandvik]

I sometimes wish these Vibe coders would set up a company, sell their product and spent time vibe-debugging the convoluted vibe generated code output based on strange bugs from end users' crash reports. Open Source does not count -- as you could just abandon the project.

[whassup]

How much do you work for?

I am not programming but heard things.
But if there are programmers in this industry who wanna say otherwise then cool.

[TechHaus]

You could even work for free. You can run Google Gemma 4 and the new Qwen 3.6 35b locally in LM Studio or Ollama, given your hardware allows it. It's not Claude, but can already be pretty useful for smaller tasks. And is a big step forward.

Basically, what Urs said!

I run the 9b and 4b level models on my Linux music / video editing desktop with an RTX3070 (8gb), and I got an M2 Mac mini with 24gb unified memory for $240 used, to run 35b / 32b / 26b / etc. models. All fine for chatbotting, but I don't even like their CSS output.

If i need a good outline for an SEO pillar post, they have stolen enough content from all the medium and substack posts to give me a good starting point. Happy I don't need to hire an SEO guy in 2026. But I still write the content like a luddite.

Washington Post and AP just published articles about the deals companies have made with the Pentagon to support wars. And i include Anthropic. They are no saints.

[Zeisner]

It's just another synthedit era. Most will flash in a pan. A handful will grow something meaningful. It's not that big of a deal in the current form.

I've never seen a SynthEdit plugin that turned out to be useful. Way too CPU/RAM hungry and unstable.

[jamcat]

for $20

Nope

Are you suggesting you would need a $200/mo Pro account, or that someone without a C++ background could realistically develop a full software project worth doing on a free account without Claude Code or Codex IDE integration?

[TechHaus]

for $20

Nope

Are you suggesting you would need a $200/mo Pro account, or that someone without a C++ background could realistically develop a full software project worth doing on a free account without Claude Code or Codex access?

Even the $200 accounts have been enshitified, recently (which is what Urs was saying, above!)

[jamcat]

You’re talking about Claude Code only, or OpenAI Codex, too?

[TechHaus]

You’re talking about Claude Code only, or OpenAI Codex, too?

All of it - I used to take Uber to the airport for $7 when they were subsidizing their services.

EDIT: Jensen Huang just said this week that a developer should spend half of their salary on tokens. That is where they need to be financially to make this work.