The four-chore loop

Step one: find it

Where did you save it? Honest answers include: a note titled 'prompts' (one of several), a pinned chat, a message you sent yourself, a text file on the desktop. The search tooling around all of these is borrowed. Notes search is built for notes. Chat search is built for conversations. Nothing in that stack understands a query like 'the earnings prompt, the one with the risk focus, the version that worked.'

So you scroll. The scrolling is the expensive part, and not because of the seconds. You were mid-analysis, or mid-draft, or mid-diff. Now you're reading meeting notes from October.

Step two: fix it

The prompt you find is stale. The client baked into it is two projects ago, and there's a formatting instruction you added for one specific deck. So you edit it in place, which quietly turns your 'reusable' prompt into a log of whatever you needed most recently. Do this for a few months and the saved version is nobody's version.

The standard workaround is a master copy you duplicate before each use. Congratulations: the loop now includes copy management.

Step three: fill it

Now the placeholders. You scan the text for the parts that change, swap them by hand, and hope you caught everything. The placeholder you miss becomes a confused response. The value you forget to update sends the wrong client's name into the output.

Nothing in a plain-text prompt marks which words are structure and which are slots. That distinction lives in your memory of writing it, which is sharp a day later and gone after a month.

Step four: paste it

Copy. Switch to the model. Click the input. Paste. Remember the attachments, go find the files, attach them too. By the time you hit enter you've crossed four surfaces, and the original thought, the reason you needed the prompt at all, has been waiting at the back of your mind through every one of them.

The expensive part

Add up the clock time and prompt reuse looks merely inefficient: a few minutes of hunting and editing against the three it would take to retype. If that were the whole story, none of this would matter much.

The real bill is the context switches. Every app transition drops the state you were holding: the argument you were building, the structure of the doc, the path through the code. There's a whole literature on how slowly people resume deep work after an interruption, and my experience says to believe it. The loop doesn't just take time. It takes the thread.

Collapsing the loop

What I wanted wasn't a faster notes app. It was no trip at all. One keystroke over whatever app is frontmost, a few characters to filter the library, Variable fields already showing last time's values, and one action to inject the rendered prompt into the app I never left.

Summon, search, fill, inject. The same four beats, zero departures. That loop is the reason Linea Prompt exists; the Blueprint Library, Variables, Context, and Injection History are all in service of not leaving the app you're working in.

Try counting your own apps tomorrow. If the answer is one, ignore everything I just said.