cuddlyogre
· Level 6
random
Opus is a slobbering moron compared to Fable. I genuinely feel better about my skills that I'm equals with Fable in most of what I do.
1
Comments
I am now developing with DeepSeek V4 Flash to optimise my agent / test my agent. Still trying to get the most out of it. I'm out of Claude Max now, but I actually went for Sonnet the last few times and I was fine with it.
But have to admit that Fabel did magic on my interpreter 😄.
Hmm, yeah, but actually, by prompting more specific and detailed, you can use a lesser model. Maybe that's a route to investigate as well. My last product cost 10 hours to prompt and way more to generate using Deep Seek. Deep Seek did a fine job because of prompt style.
I mostly optimize prompts by asking AI what is unclear about it. It always comes with a massive list. Together we keep updating the prompt until no comments left. It takes a shit load of time but it pays back hardcore. Nothing worse than forgetting to specify the language 😂😅 Nah, it's all in my
claude.ms.The problem I'm having is I can directly point to the problem I am having and the fix I made and Opus will tell me that the fix doesn't actually work.
I'm not spending $200 a month for that. I would spend $200 a month for Fable, or something close. Maybe the new OpenAI model is worth looking at...
Or try DevPlace Code soon! It’s optimized agent wise instead of model! I will generate Factorio with it in the end 😉. (This is not Claude, this one renders perfect 😛)
@cuddlyogre see my claude.md setup: https://devplace.net/gists/3fc4fcfb7140-claude-md-structure
It’s massive 😜
My agent supports all Claude type files. 😊
I have a lot to learn...
We all do. But I answered more than 300 questions about software development or so, really sick. But from now on, consistent work forever. 😒