@hamspergler@MrBerneker@trikcode The problem ive had is that the more principles and skills added the more it impacts the reasoning in unpredictable ways
@J4GameDev This sort of post only encourages the exact behaviour you're condemning except by the people that agree with you rather than the ones that don't
@Noowainy@pizdabol_king So LLMs can solve all of the ambiguity and taste that comes with programming, but it cant do the same thing for game design. Something doesnt add up here...
@btn0s@EZE3D Its not nitpicking to ask how, when engineers would previously have said thinking about how to implement something correctly was the hard part, you seem to be operating as though that was never even a concern and is a mechanical thing for the llm to do, its just contradictory
@btn0s@EZE3D The basis is that LLMs abstract the problem solving and logical thinking part, if you're finding that answering leading questions from an LLM is more cognitively taxing than doing actual work then I dont know what to tell you.
@NoemiTitarenco when I prompt claude without all the harnesses and factories I get dogshit out the other side. So yeah, you're not wrong you _can_ do it, but the only reason you think theyre the same is because you dont know the difference.
@btn0s@EZE3D I use llms all day every day, what youre equating is cognitive overload of being shown things constantly for approval without you having understanding of it, it feels like youre doing more because you have to do more to understand less
@RaphV9@thorstenball@MichaelAlleged All of them all day every day? The decision to not make an awful public interface, the decision to name your variable properly, the decision on how to handle errors.
Llms repeatedly fuck these things up which is why they take longer and longer to work as a a project grows
@RudyNearYou@Leafar_P1@EZE3D Hes the sad loser not you projecting your desperate copium that you arent a talentless meat proxy by commenting on every single persons reply lol
@btn0s@EZE3D Its crazy how many people say this as though there wasn't endless respurces and communities for learning. The barrier of entry got dropped to 'you dont even need to use your brain' and thats what got us here.
@RaphV9@thorstenball@MichaelAlleged I had the opposite thought, those models could write code just as well as the current ones just at a smaller scale, they were shit and have always been shit at making decisions.
The models arent getting better theyre just churning out larger quantities of the same bad decisions
@plasma_node 2 years ago I said that llms could write very clean code if you knew how to get them to, but it was shit at decision making. This week I said llms can write really clean code but its shit st decision making. It got better at what didnt matter and didn't improve where it does.
@zaxiik@moulinauxbulles I dont only use it on xhigh, I see the same issue at all reasoning levels, I was just meaning even with the 'smartest thinking' its still bad a lot of the time