I used to write, design, and post everything by hand. Four channels, plus a newsletter. Solo.
That math stopped working about a month in. There are only so many evenings where "okay now I post it somewhere" is the fun part. So I built a system that does the boring half, and I keep the interesting half -- writing and deciding what ships.
This is how it actually works, the honest version. Not a screenshot tour of a clean dashboard.
The problem was never the writing
The traffic to my own head -- that was the subtle part. When one person does every channel, the bottleneck isn't the work. It's the context switching. You write a good idea once, and then you re-format it for Instagram, re-tone it for X, squeeze it into a pin, and condense it into a newsletter. Four times the effort, for one idea.
Most of that re-formatting is mechanical. So I stopped doing it by hand.
The core: write once, repurpose everywhere
The whole system sits on one rule: one core piece, fanned out. I draft a single long-form piece -- the anchor. That one piece becomes the Instagram carousel, the X thread, the Pinterest pins, and the condensed newsletter version. Same idea, four surfaces, no rewrites from scratch.
The important part is in how. Nothing is a copy-paste. Each channel gets its own real adaptation, because each channel has a different job. Instagram is proof-first and visual. X is casual, one or two takes, sometimes a joke with no CTA at all. Pinterest is keyword-first. The newsletter is the thinking. Same voice, different register.
Where the automation actually lives
The stack is boring, on purpose:
- The orchestrator (an agent I run locally) does the trend scan, the drafting, and the fan-out. It writes in my voice -- it has a file of writing rules I keep, and it checks every draft against them before I ever see it.
- A scheduler queues the posts. Drafts, review, and scheduling all flow through it, so I approve in one place.
- Images render from templates. The brand fonts live on my machine, and I control the one accent color per post. The agent fills in words and color, not design.
- Memory keeps it consistent. Every shipped post's claims, numbers, and links get logged, so a new post can't quietly contradict something I already said in public.
None of that is exotic. That was the point. I did not want a system that needs me to babysit it at 11pm.
The approval gate is the whole product
Here's the part I care about most. Nothing publishes without me saying so.
The agent drafts and proposes. It never hits "post" on its own. I look at the review queue, cut what's weak, bump what's strong, and only then does anything get scheduled. The automation removes the chore, not the judgment.
That gate sounds like a slowing-down. It's not. It's the reason the system works at all. The un-stoppable content machine is how you get a feed that reads like a bot. I'd rather post less, well, than more, identical.
What I got wrong first
I tried to over-engineer the beginning. Six profiles, every channel, all running daily. That was me mistaking production for a content strategy.
It collapsed under its own weight, because production capacity is not the bottleneck. Review is. One human reading and approving everything is the ceiling, and I wanted capacity I did not have time to approve.
What I actually needed was the guardrails first, and the growth second. Consistent posts that are honest and don't contradict each other are the baseline that makes later metrics worth looking at. Metrics on a feed that reads like a bot tells you nothing.
What I would change if I started today
Start with ONE channel, tiny, and prove the loop through the approval gate. Two weeks of real, approved, shipped work. Then clone it.
I started by trying to run everything at once, and the list of things I optimized first was the wrong list. I should have optimized for "can I keep saying yes to this without hating my evenings."
Why a solo dev should care
If you're a dev, you have a weird advantage here that most writers don't. You can build the tooling that makes publishing sustainable, instead of buying fragile hype. The thing that finally worked wasn't a magic AI subscription. It was treating content like a small software system -- input, pipeline, gate, output -- and being honest about where the actual constraint is.
The constraint is always review. Everything else is a script.
I'm not going to pretend the dashboard is a screenshot-tour for you, because mine is half notes and a backlog. But the shape of the system is the shape that works: write once, repurpose honestly, never ship without a human call, and automate the rinsed-and-repeated parts only.
That, and there's genuinely something nice about ending the evening having made the call instead of having done the data entry.