A product manager's entire OS, automated to the eyeballs

One PM. Six tools wired together. A morning routine that runs itself before the coffee is poured.

Project shot · final site, herodrop your screenshot here →

§ 01 · What we builtA morning routine that runs itself.

Every weekday at 7:02am, before I open my laptop, my Notion daily page is already populated with last night's Linear comments, the PRs assigned to me, and a digest of every Slack thread I missed in #engineering.

The full system is a chain of about a dozen Make scenarios triggered either by time or by Raycast hotkey. The PM-OS replaces the first ninety minutes of my old workday — the part where I used to read everything, write a triage list, and pretend I was "planning."

§ 02 · Problem being solvedThe job is mostly reading and ranking.

Product management at a mid-size company is, optimistically, 70% reading and 30% writing. The reading part — Linear, GitHub, Slack, Notion — is mechanical: skim, decide if it matters, file it. It doesn't require judgment until the very last step. So why was I doing it by hand?

I wanted to spend my mornings on the 30% that actually requires me — the writing, the synthesis, the conversations — and let software handle the reading.

§ 03 · What was the hardest partGetting the summaries to not be useless.

The first version of the digest was a wall of bullet points that took longer to read than the source material. I rewrote the ChatGPT prompt seventeen times. The thing that finally worked: telling it to write like a PM writing for themselves — five sentences max, no headers, name the people involved, end with the one decision required from me today.

§ 04 · What went wrongThe Linear API silently truncated long descriptions.

§ 05 · What you learnedThe bottleneck isn't the tools. It's the prompt.

One: the API call is the easy part. The prompt is the hard part. Treat your automation prompts like production code — version them, comment them, test them with edge cases.

Two: never trust a summary you can't verify in one click. Every line in my digest links back to the source. If it looks wrong, I can check in two seconds.

Three: the best automations replace the reading, not the deciding. The minute you let the LLM make a call for you, things go sideways.

§ 06 · The stack

The stacktoolrolecost / mo
Raycast
Command-palette trigger for every automation.
$8
Visit
ChatGPT API
Summarizes overnight Linear traffic into a daily digest.
$20
Visit
Notion AI
Daily page destination. Acts as my second brain index.
$10
Visit
Make
Glue between Linear, ChatGPT, and Notion. The actual automation runtime.
$9
Visit
Linear API
Source of truth for what shipped, what's blocked, what's mine.
free tier
Visit
Claude
PR diff review. Flags anything that smells of regression.
free tier
Visit
Total monthly spend$47 / month
~ end ~
M
About the author

M. Tanaka

M. Tanaka is a senior product manager at a fintech in Tokyo. She started automating her job to get her evenings back and accidentally turned it into a side hobby.

Built something stupidly clever? we want the receipts.

Submit your build