Built a lot of FiveM scripts with AI help lately, noticed a problem. Anyone else?
Been using ChatGPT and Claude for FiveM scripting, and they're genuinely useful for boilerplate and logic, but they keep getting things wrong in ways that are annoying to debug, wrong native context (client-side native on the server), deprecated ESX functions, and hallucinated QBCore exports that don't exist.
I'm looking into building something specifically for FiveM scripting that's actually grounded in the real native reference and current framework source, rather than just a general-purpose LLM guessing at FiveM stuff.
Before I go too deep on this, I want to sanity check the idea with people who actually script:
- Do you use AI in your scripting workflow at all? What works, what doesn't?
- What would a FiveM-specific tool need to do to be genuinely better than just using ChatGPT?
- Would framework support matter to you, QBCore, ESX, ox_lib, etc., or do you mostly figure that part out yourself?
- Free with limits vs. a small paid subscription, what would make sense for something like this?
Not trying to pitch anything, genuinely trying to figure out if this is worth building or if the problem isn't big enough to matter. What's your honest take?