Skip to main content

Content pipeline · Live

The pipeline that writes about what I actually shipped.

Pluma reads Orbit, my own project tracker, for whatever’s genuinely done in the last seven days and drafts it into X, LinkedIn, blog and Build Log posts — in my voice. I read every draft before anything goes anywhere. Nothing posts on its own.

Scout → Strategist → Writer → Editor → Publisher

THE SOURCE

Real finished work, not a blank page. Every draft starts from a task Orbit already logged as done.

THE VOICE

Trained on what I approve. Every approved draft feeds the tone of the next one.

THE RULE

Nothing posts without me. Approval is a real gate, not a formality.

The signal

It’s reading something real, not a blank feed.

That glowing sphere is straight from the real app — Pluma calls it Orbit Intelligence, a live view of what’s coming in from Orbit each week: tasks completed, projects active, drafts worth sharing. The sphere itself is decorative, not a literal chart. Everything it represents is real.

Pluma's Orbit Intelligence globe visual from the live Command view

How it works

Four steps, then it waits for me.

01

Scout reads the week

Queries Orbit for tasks and payments finished in the last seven days, and drops anything touching credentials or security before it goes any further.

02

Strategist finds the story

Groups whatever survives into real narrative threads and decides where each one runs — X most often, blog for the week’s single strongest thread, LinkedIn least.

03

Writer and Editor draft it

Writer drafts each platform’s version in my voice. Editor trims X posts to fit 280 characters, checks the app name isn’t buried, and writes a real SEO pass on anything longer.

04

I review, then it goes out

I read every draft in the queue and press approve or reject. Approved X posts go out in about 15 minutes; blog and Build Log entries publish on their own daily check.

The team

Five agents, one pipeline.

Scout

Scans Orbit for what’s genuinely worth sharing. Skips anything touching credentials or security, outright.

Weekly

Strategist

Groups the week’s work into real narrative threads and decides where each one runs.

Weekly

Writer

Drafts each platform’s version in my own voice, working from my approved history.

Weekly

Editor

Trims X posts to 280 characters, writes a real SEO pass for blog, and checks the app name isn’t buried.

Weekly

Publisher

Posts approved X drafts in about 15 minutes, and publishes approved blog and Build Log entries once a day.

Continuous · Daily

See it in action

The real app, not a mockup.

Every screen below is the actual live app, not a recreation — the queue itself is mine to edit, not public.

Pluma's live Command view: agent pipeline, Orbit Intelligence, and priority queue
Command · the whole pipeline, one glance
Pluma's live Review view: a real draft with agent trace and approve/reject controls
Review · an editing desk, not a black box
Pluma's live Voice view: measured voice signature and style guide
Voice · calibrated from what I actually approve

The cadence

How often it actually runs.

Weekly

The full Scout-to-Editor pass runs once a week, drafting X, LinkedIn, blog and Build Log candidates from the last 7 days.

Daily

A lighter, Build-Log-only pass runs every day, so a shipped feature shows up there in about a day, not a week.

~15 minutes

Approved X posts publish on their own within about 15 minutes of me clicking approve. Everything else waits for its own daily or weekly check.

No invented scores. Readiness and style compliance are plain, checkable rules, not an AI’s opinion of itself, and nothing here reports an engagement number I can’t actually pull from a real API. Scheduled tasks only run while my Claude Code app is open, so “daily” means a day of it actually running, not a day of the clock.

Voice

It learns from what I approve.

Pluma doesn’t write from a fixed persona. The Writer and Editor steps open by reading a short style guide I keep — who I am, how I actually sound, how I structure a piece, what’s worth writing about — plus my last eight approved drafts, and match against that instead of a generic AI voice.

Every time I press approve, that draft becomes tone reference for the next one. The more I approve, the more calibrated it gets.

Direct · first-person sentences
Evidence-led · concrete numbers over vague claims
Declarative · sentence length

Measured from real approved posts using plain text statistics, not an AI’s opinion of its own voice.

Why I built this

The work was already real. Nobody was finding out about it.

I kept shipping real things every week, across five or six projects, and telling nobody. The proof lived in a git log and in Orbit, my own tracker — useful to me, invisible to everyone else. Writing about it properly took time I didn’t reliably have, so it mostly didn’t happen.

Pluma is the fix. It reads what Orbit already knows is done, drafts the first version in a voice that already sounds like mine, and hands it to me to approve or bin. I still make every call on what goes out. Pluma just makes sure I actually get asked.

Everything on this site’s Build Log, and most of what shows up on my X account, started here. I run it every week already — if you’d want the same machine pointed at your own work, that’s what the waitlist below is for.

— Yeison · yeison.co.uk

Early access

Want this pointed at your own work?

Pluma already runs every week on my own projects. If you’d want the same machine reading your own tracker and drafting in your voice, leave your email and you’ll be among the first I reach out to.

No spam, ever · I’ll only email when there’s something real to say.