I’ve been watching my conduct evolve as I take advantage of AI extra.
Coding new tasks, I began first by describing a giant thought : generate a brand new web site that ingests podcasts and transcribes them. However I rapidly found this strategy was like asking somebody to construct a skyscraper with out blueprints—the AI choked on requests too broad & missing construction.
So I broke the challenge down into smaller blocks. Write a perform to obtain the record of podcasts within the final week from these websites. However I observed how low-leverage that felt : I knew I used to be yoking a robust animal to tug alongside a baby’s wagon.
Right now, I take advantage of AI to write down a primary PRD, ChatPRD is nice for this. Then I ask the AI to learn it and create an implementation plan. I approve the plan after which insist that the AI write exams first for every perform after which check every part.
At this level, I can let the AI run wild for 10-Quarter-hour. With the plans and check instances, it may iterate to success itself.
As I watch the code cascade via the window, I’ve the identical feeling of being pushed by a Waymo. Having fun with the journey rather more now that the toil is managed by a pc.
Prime of thoughts for me now : how do I parallelize this so I can handle 10 browser tabs or terminal home windows concurrently? My limiting issue is structuring concepts & questions in a technique to succeed with the AI working.
Exterior of coding, I’m seeing myself default to AI search & reverting to Google.com once I suspect one thing is fishy within the AI’s output.
That is now additionally true past on-line searches : somewhat than asking a human skilled, I’m experimenting with asking an AI first. Why haven’t equities moved onto the blockchain in the identical method debt has?
The rub on this effort is understanding when to belief the AI. Like every human, the longer an AI bloviates, the much less I imagine it.
My deeper concern is what nuances has the AI excised? What am I lacking from its abstract?
How have you ever modified how you’re employed with AI?
