Graham McNicoll

I cannot look at a hundred experiments and tell you what they have in common. Nobody can. That is not a knock on anyone. A hundred experiments is just past the size a human brain holds at once. For a long time the answer to this was documentation. Run the test, write it up, and later, in an ever growing document. The thing is, nobody reads the document. If you're lucky, it will be used to make sure you don't duplicate an experiment. Rarely is it used when making a new product. But the way we build products has changed. You do not sit down and read a folder of old experiment docs before you start. You ask your agent to build the thing. So the synthesis has to live where the building happens. When someone points their agent at a new feature, it should already be reading across every related test you have run, what worked, what failed, and applying that before it writes a line. If you're only using experiment results for making sure you're not duplicating work, you're not taking advantage of the rich information you've learned about your users. Past experiments to compounding into an advantage. Make your experiments available to your agents, have it use the findings for future products. We built this into 5.0. Right now, is your experiment history an asset or an archive? https://lnkd.in/gmZq8nym