Saying No In The Age Of Coding Agents
Current times have these properties:
- The gap between idea and product is converging to zero because of coding agents
- We are on the precipice of transformational technological change that could go very well or very poorly
This is convincing me that the most important skill now is to develop a sense of what projects to say no to; one should have criteria for picking projects.
Here are mine:
- Is it fun? Does working on it give me energy? Do I feel joyful?
- Is it relevant to mitigating catastrophic risk from AI?
- Is it urgent? Is it important to know the answer now as opposed to 6-12 months from now (when the cost of working on it will likely have at least halved due to increase in LLM capabilities)
- Is it impactful? Does it provide an actionable answer?
- Is it urgent? Is it important to know the answer now as opposed to 6-12 months from now (when the cost of working on it will likely have at least halved due to increase in LLM capabilities)
- How many other people are working on answering the same question, or solving the same problem?
- E.g. this is a good reason to not worry about certain open source bugs or features.
- This also implies that sharing things earlier is better. It’s time to ship posts!
- E.g. this is a good reason to not worry about certain open source bugs or features.
2/3 are a very standard EA decision principles. I think one 1 is important too! Fun is a strong signal from your body about alignment!