

I am typing this on a 5 year old Android phone. It has 128GB of memory and 8GB of RAM, very decent cameras, a beautiful OLED screen and a processor that is more than fast enough for everything I do with it. And even now the battery still lasts two days with normal use. It cost me about €300 at the time.
Unfortunately the Android version is getting so far behind that some apps are starting to get a few issues, so I have been checking out some black Friday deals for new phones, but they look very disappointing.
In the current market it seems like I’d have to pay about €500 to effectively just get a side-grade. All €300 offerings look like just a straight up downgrade in any way apart from the more recent android version.
So I think I’ll hold on to this one a while longer. Hardware-wise it’s still in perfect condition, and if software support really becomes an issue then perhaps I’ll try out a custom ROM.




I’ve also experimented with this. In my experience, getting the NPCs to behave the way you want with just a prompt is hard and inconsistent, and quickly falls apart when the conversation gets longer.
I’ve gotten much better results by starting from a small model and fine-tuning it on lore-accurate conversations (you can use your conversations with larger models as training materials for that). In theory you can improve it further with RLHF, but I haven’t tried that myself yet.
The downside of this is of course that you’re limited to open-weight models for which you have enough compute resources available to fine-tune them. If you don’t have a good GPU then the free Google Collab sessions can give you access to a GPU with 15GB of VRAM. The free version has a daily limit on GPU time though so set up your training code to regularly save checkpoints so that you can continue the training on another day if you run out. Using LoRa instead of doing a full fine-tune can also reduce the memory and computational resources required for the fine-tune (or in other words, allows you to use a larger and better model with your available resources).