Lore

Ship Early, Let Usage Define Scope

Linear's internal agent began as a scrappy prototype meant to automate "boring" work; its scope grew organically once real usage surfaced demand nobody had specified upfront — e.g., users dumping raw sales-call notes on it, or giving it vague Slack instructions and expecting it to figure out the right action ("Computers can do a lot of work for us. So let's get rid of all the work we don't want to do and give it to computers."). The argument: ship a minimal version early and let what people actually try to do with it define the product's real scope, rather than trying to specify that scope in advance. This pairs with Expert Dogfooding as an Implicit Eval Substitute — the same real usage that expands scope is also the source of evaluation signal.