alt Hacker NewsOh, fuck off. For one, I work in an LLM startup and work on the harness for our own models, and we're 8-digit profitable, so I'd say I know a thing or two about prompting. For two, there is no prompt that makes Luna work. It ignores your fucking prompt. The OP related how Cursor's model has a quotation mark tic and blatantly ignored a single constraint not to use quotation marks. Luna also has the same deficiency. You can give it a single paragraph prompt with a single sentence constraint not to do X and it will do X every time if it's in the average path for the prompt.
sorry my comment was a bit rude. what i meant was that if you're not vibe coding and instead doing a lot of the architectural work yourself, the less powerful models can still be quite effective for doing the grunt work.
when i built my own harness i had to use sol 6.1 medium, luna, deepseek, etc made too many mistakes, but i was full-on vibecoding it. in other projects when i put more effort into it, models like luna, deepseek, etc do just fine.