after many, many iterations i’ve finally landed on a my own custom built harness and setup that fully scratches every itch
i feel like everyone has a slightly different set of requirements and the only way to truly satisfy it is to build your own
fortunately it’s never been easier
i mourn for the times i didn’t use all my tokens whenever i’m staring down the barrel of maxed usage
does anyone use open weight models as their daily driver?
now that i’m fully utilising my mac mini i kinda want to get a mac studio too
i’ve rebuilt my ai harness yet again and it finally feels like a massive upgrade on anything else available
it’s feels like ai harnesses are a bit like todo lists in that everyone has a unique set of wants and needs
so you either end up compromising, adopting someone else’s workflow, or building your own
also basically impossible to make money from them
good people of @cursor_ai, please provide a sane way to get usage data
you’re killing my tokenmaxxing dashboard
there’s two distinct types of ai harness
one where you treat ai as a tool and one where you treat ai as a partner
i’ve been flip flopping between the two but i think it depends on the context of what you’re doing
so perhaps there won’t be one harness to rule them all
what model and thinking do you have codex on?
i consistently blitz a weeks usage in a couple of hours
15+ years of primarily working with people via IM has put me in a good position to work with AI all day
i love weekend warrior projects where you give some random idea to an llm and then build out the whole thing via your phone
you should make guests attempt to split the g before starting the conversation @collision
if you’re a high profile employee at one of the ai labs, you basically get free juice to make @x rev share bank