#codeassistant — Public Fediverse posts

I understand that #AI companies are shifting from monthly fee to token based pricing. In fact there is no way someone/something will write you dozens thousands lines of somehow usable code for lower dozens of dollars monthly.

So here are my toughs:
1) they made people addicted/dependent on AI for “free” and now they start milking, do I see here paralel with drug dealers?
2) their management is really utterly dumb, in general I do not believe that.
3) they have been for sure subsiding costs till now and hoped they will sort the operational costs and improve efficiency meanwhile, well either point 2 applies or something went really south.
4) companies like Microsoft can afford to some degree burn money like this in huge pit, but it will be popcorn sitcom to see how OpenAI and friends will deal with this, as they did not produced any income cash flow and investors will start now asking viability of such companies.
5) AI will stay with us, but in different shape, I believe it will be pushed to the edge and enduser computing with smaller specialized models.
6) if 5 applies and I believe so, OpenAI knew tat and they intentionally screwed RAM supply chain to make this impossible/horribly expensive and unaffordable.
7) lets say it will stay in cloud, people play for tokens, it is Q of time they will start challenging why they shall pay for tokens burned by AI debugging its own crappy code, producing even more crappy test and ending in broken record cycle. They will draw parallel with human, if human screw programming and spend more time on delivery, it is hard for him to charge you more for his mistakes. Why this shall be accepted from AI companies. And yes this mostly applies for vibe coding.

#codegeneration #LLM #codeassistant #sustainability

Heads up: the new version of @Posit RStudio software (2026.04) has a genAI assistant integrated. It's subscription based, but can be turned off in the settings.

It'll be the first time I won't mail the IT department to ask for an update. We'll have to test how easily students can activate their own subscriptions on exam PCs before we even think about updating.

If anyone can shed light on enterprise control of settings for this, much appreciated.

#RStats #genAI #CodeAssistant #RStudio

AI Can Write Code, But It Can’t Debug Without Context

Large Language Models (LLMs) are often marketed as the ultimate productivity boost for developers: “Write code faster! Debug with AI! No more manual work!” After a recent experience, I can confirm that LLMs are incredibly useful for writing and even structuring code (I’ll write about this probably in a later blog post).

But when it comes to debugging, one should make really sure that the tool has access to all the relevant context (and don’t disable your brain). But .. let’s see, what happened:

Since a couple of days (uhm .. nights mostly, after work), I was writing a web application. The copilot-experience was very good and it really helped tremendously. I never really ran into a situation where I had to debug. And I was curious when (if?) I’d run into that – and how things turn out then.

[…]

Outdated code snippets from Stack Overflow jeopardise software security
https://cispa.de/en/jallow-stackoverflow and
https://publications.cispa.saarland/4053/

This could be relevant for #Copilot and all other #codeassistant tools training data.

#security #software #codesnippet #stackoverflow #github #cispa #cispahelmholtzcenterforinformationsecurity