swordsith

good will

First thought looking at the first options was I would absolutely never prompt this way. The LLM may as well be a misaligned genie in a bottle, it has to be treated as such.

Personally I've always seen AI 'memory' as a pain point for people in their experience using LLMs than a benefit from the agent remembering the last unrelated thing you were working on. It wastes context similarly to 'skills'. The most efficient workflow imo is having a few well written (not by ai) md files across a clean codebase.

It almost seems like they might've been projecting.. lol

Yeah not telling people you're doing the equivalent of css on their website is diabolical dude. You should be ashamed.

my solution to nasty websites with js ridden everything is to scrape and remake the entirely new dom populate it with only content you verify, sure you might lose some detail but its easy to copy website ui if you desire. can be done with something like tamperminkey or a browser ext

The devil is in the useragent and other request metadata

awful website jfc

ragebait used to be hard to identify

The source of the bloat is less vibecoding and more architecture missteps and reliance on using web interfaces for everything.

Sometimes after a larger source draft with AI before I even touch the program I start to pick up on some of the holes in my prompts caused bugs/unintentional mechanics to be potentially woven in. Knowing what you want and how it should be made is half of it, but if you don't supply the bot with extra guardrails eg. don't modify the contents of x, because of y don't allow z. etc. They will do what you ask of them, usually less than you'd hope.

Saying we're at the Peak of LLMS is like saying we never went further than everest, these are probabilistic state machines sure, but the real value is the effort put into tooling and training these models in RL situations.

Could send random characters along with 'keep going' and it would change nothing, the seed is whats causing deviation between these responses.

Auto is not a model, auto is the mystery box option that cursor gives people who are too slow to read a documentation and flip a few switches. With auto you're going to get routed to any model that happens to be the most available and cheap at the moment. Rather you should specify composer or grok as they both charge under cursors auto category - which by the way can be used just fine through cursors cli and ide. Also auto pool usage isn't infinite, just hard to exhaust.