In line with cheaper, I'm also hoping we get better quantization or small models that can compete with big models. Running models locally is nice, but the kinda stuff I can run on my laptop or phone kinda sucks at the more complex stuff. I'm also looking forward to some light local agent stuff. I'm looking forward to telling the local model on my phone to research something and then have it come back with a deep research style dissertation for me.

Brave, LibreWolf, and Fennec

Yeah I feel like the food desert idea is a bit too strict. I've never lived in a non-food desert, but in a rural area everyone has cars so a 12 mile ride isn't exactly horrifying idea. I'm in a more urban environment atm, but literally less than an hour ago I did shopping at a grocery store two miles away via e bike, and I regularly peddle bike or walk there.

https://www.yahoo.com/news/world-economic-forum-rigged-data-105244292.html I've got no horse in the race on Brexit, but I can't miss the opportunity to clown on the eat bugs and rent everything guy. (Actually, I have eaten bugs once or twice. Mealworms tast pretty good when seasoned, although they're way more expensive than standard meats. If done at scale it might be a cheap source of protein that'd be healthier than chips or some other similar snack. But nevertheless that's my choice, and I'd rather not be forced to eat mealworms because all of my money is going to paying for my pod and uber to support my local megacorp; which costs more long term over buying a house and car).

Edit: reading replies, yeah, sounds like you found out it's a token limit. LM studio does let you set a custom token limit, to mixed results. Might be worth trying with a couple LLMs to see if they'd handle bigger files.

I've used Mixtral (the quantized Dolphin fork) to turn subtitle files into essay style documents. Works perfectly, until you get files that exceed its token limit and it summarizes or plain errors out. The biggest version of Mixtral you can run might do a good job of understanding vtt files (renamed to .txt if it refuses the original file). Speaking if which, it might not be Lama but a token limit. If you feed it a file that's more tokens that the AI will accept it can sometimes lead to really sub par results if you expect it to understand the whole file.

At the very least, it's probably still great for self hosting in that you can quantize or distille it into a model that could run on a more consumer PC or phone.

Iirc there's a few ways to get RSS from Nostr feeds, of which you could use an RSS to Email tool. Would be hacky but might be the easiest way to do so.

That hits a little close to home lol. I'm trying to move back to/around my hometown, and I expect to take a pay cut, but I'm seeing all these jobs where it's like "Okay, we require 5+ years of experience and a bachelor's degree, a handful of certifications are a plus. We're offering $2-$4 above minimum wage, depending on your experience and education levels. You must be okay working weekends, nights, and being on call."

Probably quite a bit of time, I've been doing a lot of experimentation with some different tools and projects. Often it's during quiet parts of work shifts when I can jump between my personal computer to write a few prompts and my work computer to do work, but I wouldn't be surprised if I found out I had some form of AI code editor open for 40-80 hours over the last couple months.

In most cases you could probably just tell them to screw off - unless you're registering a local business entity and accepting payments or selling ads, the worst they could do is try to add your domain to the great firewall of Britain since they have no jurisdiction over a server in another country (AFAIK). One of my sites just says "GDPR compliance: lol" (even though I'm technically in compliance with the GDPR since I don't collect any data). As long as you and your servers aren't in reach of the UK or whatever other region is trying to enforce laws I think it's safe from any needs for compliance.

Lol 90% of the time it's a fun thing with the feeling of infinite possibilities, the other 10% is me cussing out the AI and trying to manually fix broken stuff

Lol reminds me of this gem from grammerly

Oops, saw this in my notifications, probably meant cost not time. Used a Gemini free trial, then tried one month if ChatGPT for 20$, used a GitHub Copilot free trial last month before paying 10$ this month for a subscription, and put 25$ into open router so far. Trying to do everything on the cheap since I have no financial incentive to use the best, spent a little over $50 since I started vibe coding in April.

Liar! I've eaten Dollar Tree garlic bread and I'm just fine. *twitch* *twitch*