Reply To: [unofficial weekender] First of 2026

[limburger 22353xp 2]

I’ve added my recent experiments / scribbling / prototyping ideas to my warroom project.
That should keep them somewhat safe, because I’ve noticed that LLM’s like Gemini and Claude.AI tend to make it tricky to retrieve any documents they have created.

It’s a bit like talking on discord, because the data these thing generate tends to evaporate the longer a conversation lasts.
Some may say that this happens because the ‘free’ versions have limited storage capacity, but I think it is part of the design as well.  These things work best when there is a limited scope as any long conversations expose their alzheimer like tendencies and increases the chances of hallucinations f*cking up your results (which is what happened almost every time I’ve used them).