Technical

Exhaust every branch of an idea

Ask follow-ups, challenges and edge cases on separate branches, and work through each one until no open question is left.

Explore the whole book

A map of Harari's Sapiens, zoomed out to the whole book. Zoom in on any chapter, or clone the map to branch off with your own questions.

Open the full map

How it works

State an idea or a plan and the model responds. Every follow-up you ask opens its own branch, so you can go deep on one question and come back to the others where you left them.

  1. State the idea. Say what you are thinking and why. The first answer gives you something to push on.

  2. Give each question its own branch. “What could go wrong?” is one branch and “What am I missing?” is another. Each goes as deep as it needs to without burying the others.

  3. Try more than one approach. Branch twice from the same node to test two alternatives. Each branch carries the conversation from the root down to it, and neither sees the other.

  4. Let auto-expand dig. Switch a node to an expand node and set a depth. The model writes its own follow-up questions under it, answers them, and keeps branching to that depth.

When every branch has an answer, read the map for the arguments that held up and the risks that are left.

Why a map instead of a chat thread

In a chat thread you follow one line of questions at a time. Going deep on one point pushes the others out of view, and switching topics loses the thread. On a map, each question keeps its own branch, so you can go deep on all of them.

To test an idea from another angle, ask a different model the same question on a sibling branch and compare the answers.

Example

Say a founder wants to build an in-house analytics pipeline. One branch asks build or buy, another asks what happens if the pipeline fails at scale, and a third challenges the timeline. Each goes three or four levels deep. The build-or-buy branch shows the team has no data engineering experience, and the timeline branch shows the plan depends on a hire that has not been made.

Now try it yourself

I want to migrate our monolith to microservices. Convince me this is the right call.
It depends on what's actually hurting. Microservices solve scaling and team autonomy problems but add operational complexity. What's the pain point driving this?
Deployments are slow. One team's change blocks everyone else.
That's a real signal. But have you tried deploy pipelines per module within the monolith first? Microservices solve this but so does a well-structured modular monolith, without the distributed systems tax.
Grill me on this -- what will go wrong if we do migrate?
Service discovery, distributed tracing, data consistency across boundaries, network latency replacing function calls, and your team has never operated a service mesh. That's 6 months of infrastructure before you ship a single feature.
What if we only extract the one service that's the bottleneck?
That's the strangler fig pattern, and it's the right first move. Extract the deployment-blocking service, keep everything else monolithic. You get the autonomy win without the full migration risk.

Ready to try deep dives?