Leave room for why not

A full roadmap turns every new idea into a request. Cal.com plans two weeks ahead and leaves the rest open, and that is where its notes app and dictation came from.

Two launches nobody planned

Cal.com launched a calendar, a notes app and dictation this week. I joined a month ago, so I only caught the last stretch. What stayed with me is that the notes app and dictation were never on a roadmap. Someone on the team asked why we didn't have our own, and spent a weekend building a rough one to find out. The founders tried it, liked it, and said yes.

I've worked on teams where that weekend would have gone nowhere. The plan had no room for a good idea.

The plan that looks responsible

A quarterly roadmap is the tempting version, and it looks like good management. Twelve weeks, every one assigned, everyone knows what they are doing in March. I've made those plans and felt organised.

What it costs shows up on the Thursday someone has an idea. Every week already belongs to something, so the idea goes on a list for next quarter. By then the person has stopped caring, or someone else shipped it. A full plan turns every new idea into a request.

Two weeks, and the rest left open

Cal.com plans about two weeks ahead. We know the next project before the current one ships, and that is as far as anyone commits. The tools we build with change every few weeks, so a longer plan would be a guess anyway.

The open weeks are where the notes app came from. An idea gets a short spec and a couple of days as a working build, and then it ships, waits, or goes in the bin. Binning is cheap, so nobody has to defend an idea before trying it. The founders write code next to everyone else, so a yes takes one conversation.

I think of it as a diamond mine, and every idea is a shaft. You dig a lot of small ones, most come up empty, and you only find out which by digging. The notes app is a diamond. Only rich companies used to mine this way, because getting from an idea to something a person could try took a large staff and years. That part now takes a weekend. ChatGPT was a shaft like that once, a research preview, and NotebookLM was a small experiment inside Google. A small team can dig the same way now.

The hours nobody is awake

The night makes room too. The team is spread from the Americas to Asia and works async, so someone ends their day by lining work up for the people logging on eight hours east. Agents take the hours in between. Several engineers have built their own pipeline: queue small tickets with a tight spec in the evening, and an agent plans each one, builds it, opens the pull request and answers the review. One colleague queued seven before bed and woke up to seven pull requests, one of them already merged without a person touching it.

Nights like that are why nobody rations the tools. The best rule I've heard since joining: if you ran out of tokens this month, you used enough.

Where the long plan is right

Some work should be planned far ahead. A database migration, a pricing change, anything a customer has been promised by a date. The overnight pipeline has the same limit. It takes small changes, and anything large or risky still waits for a person. I wouldn't run a hospital's scheduling on why not. The open weeks are for shafts you can abandon.

What a competitor can't copy

A competitor can copy a notes app in a month. Copying a team where trying something costs a weekend and no meeting takes much longer. When a shaft costs a weekend, you dig ten where a planned team digs one. That is the ten times I see against every team I worked on before.

You can always add a plan. The craft is in the weeks you leave empty.

andreaskruszakin@gmail.com