personal_asset
Which Part of a Personal Project Should You Automate First?
Start with repetitive, deterministic work that can recover cleanly after failure, not the whole workflow.
A personal project can become most exciting at the moment when its owner decides to automate everything. Source collection, writing, images, publishing, backups, and alerts should all run behind one button.
The diagram feels wonderful. The implementation reveals an exception in every step and a new configuration for every exception.
A saved essay argues that systematic thinking is expensive and cannot be sustained across every domain. For a personal project, that does not mean avoiding automation. It means choosing the first segment carefully.
Use three filters
First, frequency. Automating a five-minute task that runs once a month may save less than it costs. Second, determinism. Similar inputs should lead to outputs that can be checked. Third, recoverability. If execution stops halfway, the system should show where it stopped, and a retry should not resend a message, charge twice, or overwrite the original.
Renaming 15 images by a fixed rule, checking dimensions, and generating a manifest is a good candidate. Choosing today's five topics is not automatically safe merely because a model can rank them. The first task is repetitive and restartable. The second depends on judgement and trade-offs.
Build one complete slice
Basecamp's Shape Up recommends building one meaningful piece end to end instead of creating disconnected parts and hoping they fit at the end. The same principle helps automation.
Do not begin with a universal scheduler. Take one repeated segment from real input to real output, then force it to fail. What happens when the network drops, a file already exists, or an external request times out? When recovery works, the automation reduces work instead of replacing manual work with script debugging.
My first metric would not be lines of code or buttons removed. I would watch whether the next run requires one less fact in my head and one less moment of panic after failure.
If a person must watch every step, the labour has only changed interfaces. The signal to expand is a small segment that can finish independently, leave evidence, and return to a known state after error.
Sources: The limits of applying systematic thinking everywhere; Basecamp Shape Up: Introduction.