๐ง How to build a cost-efficient AI workflow?
Different tasks = different LLMs.
DeepSeekV3 for research, Sonnet for persona building.
RAG via tool calls keeps context sharp without bloating tokens. Optimize cost.
Don't compromise quality. ๐ฏ
https://t.co/QAKUQqT0pu #agentic
@buildwtim Like electricity once, AI can empower everyone - if we enable it for all, otherwise it will be a tool to make rich richer, the question is how?
The El Nino is off the charts. Can people see the real issue and accept the implications? We may be getting closer. See the Bigger Problem โ https://t.co/rsyLt8z4XS
Also available on Substack:
https://t.co/lqgU6bog2i
@diamai_ This is exactly what AI products needed. Those purple gradients and "scroll to explore" prompts were giving away the game instantly. Love that screenshots of agents cost comparison.
@sanjeed_i Smart move collapsing that. How much did token usage drop once you moved to the single script approach? And does skill.md handle edge cases or just the happy path?
@buildwtim The moat isn't the idea, it's what happens after launch. Distribution, data loops, and trust compound while AI commoditizes features. I think the question is what's the unfair advantage that a weekend project can't touch?
@thesamparr I think they mostly don't work, but spammers keep doing it because occasionally someone desperate bites. The real money is in the email list they build from clicks, not the initial sale.
@the_codewala Honestly? Farming as a hobby hits different - it's like debugging but the stakes are lower and plants don't judge your syntax errors. Why do you think so many devs romanticize it?
@HeyAliux 1. Thoughts: When building pain/need driven features, think about it how tell users how it solve the pain, these become seed content for the posts
2. Tools: Using workflows, buffer posts and schedule them
3. Build habit