ask: what stops you from shipping this feature in 2 days instead of 4?
formula:
what stops you from {{desired or better outcome}} instead of {{current timeline/approach}}?
Why SHOULD we {do|buy|invest|build} this? >>> Why SHOULDN'T we {do|buy|invest|build} this?
I find negative questions help grill ideas better than positive ones. You come from a "what can go wrong" view and have to convince yourself to proceed (vs starting with reasons to do it)
Deductive Reasoning vs. Chain of Thought (CoT) 🤔
🔹 Deductive Reasoning:
-- "If A and B are true, then C must follow"
-- Applies established rules to reach conclusions
-- Works from general principles to specific conclusions
Example: "All humans are mortal. Socrates is human. Therefore, Socrates is mortal."
🔹 Chain of Thought (CoT) Prompting:
-- "Let me think: First A happens, which leads to B, causing C, so D is the answer"
-- Explicitly shows each step in the reasoning process
-- Helps LLMs break complex problems into manageable steps
Example: "To calculate 13×27, I'll first multiply 13×20=260, then 13×7=91, and add them: 260+91=351"
When to use each:
👉 Use Deductive Reasoning when applying clear logical rules to reach conclusions
👉 Use CoT Prompting when you need the AI to show its work or tackle multi-step problems
Short and sweet summary
Weekend project is coming along well - building an API to track and monitor companies. I’ll be shipping a (free) YC dashboard powered by this API soon! 🤝
And pricing --> In both standard and extended thinking modes, Claude 3.7 Sonnet has the same price as its predecessors: $3 per million input tokens and $15 per million output tokens—which includes thinking tokens.