Luca MinudelWhat I’ve learned so far while coding with an LLM genie:
================================
This is what is working for me:
A - 1 goal => 1 chat => 1 one commit
B - Feedback beats "perfect" prompting
C - 3 strikes => switch model
D - My scratchpad > the LLM’s plan
E - Keystone question > roleplay
F - Follow my lead
What works for you?
Full story => https://www.linkedin.com/posts/lucaminudel_ai-artificialintelligence-generativeai-activity-7363117538395713537-AOKX
#AI #GenerativeAI
#FutureofTech #FutureofWork
#PromptEngineering #LLM
#AICoding #AIAssistedCoding #VibeCoding
============================================ | Luca Minudel
============================================ What I’ve learned so far while coding with an LLM genie: ============================================ A - 1 goal => 1 chat => 1 one commit -------------------------------------- Keeping the context small, chats short and focused worked wonders. Each small goal = a new chat + a new commit at the end. Task success rate went up, and conversations stayed sharp. B - Feedback beats "perfect" prompting ------------------------------------------ I got 10x better results by feeding the LLM real outputs, logs, and debug info. Even better: showing how to run the command and where to find feedback. This consistently helped the LLM break out of a loop of bad guesses, faulty "logic" deductions, and non-working solutions. C - 3 strikes => switch model ------------------------------- When even feedback loops failed, swapping models usually solved the problem faster than wrestling with the one stuck chasing its tail, often due to its limitations or temporary quirks. _________________________________________________ That's what works for me now. 👉 How about you? Have you tried similar tactics? Did it work or not? In which context? 🔮 I wonder if some strategies people write about don’t always generalise… With LLMs being non-deterministic and our natural tendency to anthropomorphise them, it's easy to mistake random variance for cause and effect. Pareidolia in action? _________________________________________________ 🎶 Bonus tracks 🎶 D - My scratchpad > the LLM’s plan ------------------------------------- A simple text file scratchpad for my own work plan gave me speed, focus, and flexibility. It let me edit and mix plans from multiple LLMs, bounce back when one crashed, start a new chat anytime without losing the plan or the progress, seamlessly switch LLMs in the middle of any task, E - Keystone question > roleplay --------------------------------- Instead of “pretend you’re a <role>…,” I start with a specific central question about the core of the task at hand. With the relevant details and using pertinent language, this set the stage better than roleplay. F - Follow my lead ------------------- When nothing else worked, I solved one instance manually and showed the LLM the pattern to follow. Asking it to replicate from my example worked well, especially for repetitive coding tasks. It also worked well when I could point to specific parts of the existing codebase where the same or similar problems were already solved. #AI #ArtificialIntelligence #GenerativeAI #FutureofTech #FutureofWork #PromptEngineering #LLM #AICoding #AIAssistedCoding #VibeCoding
HackerNoon
jo.vaneyck
Hacker News
रञ्जित (Ranjit Mathew)
Lukas Brausch