The Input Problem: How Zo.E Finds Something to Say
3 min read · 613 words ·
You can build a beautiful machine and starve it in week three.
That's the failure mode I see most often — and the one I nearly fell into myself. Hours spent wiring up workflows, testing prompts, getting the formatting just right. Then the system goes live, runs for a fortnight, and grinds to a halt. Not because of a bug. Because nobody engineered the front end: the part that tells the machine what to say.
Content generation isn't the hard part. Content discovery is.
There's a distinction worth making here between creative output and engineered input.
Creative output is what most people focus on: the writing, the formatting, the post, the article. That's visible. That feels like the work. Engineered input is the system that decides what gets created, when, and from what source material. It's invisible, but it's load-bearing. Without it, the creative output machine is just a very sophisticated blank page.
Zo.E has two workflows that solve this: WF1 (Article Discovery) and WF3 (Article Sourcing).
WF1 runs on a schedule. It looks for source articles — pieces of content from RSS feeds or curated web sources that are relevant to the niche I'm publishing in. WF3 handles the actual pulling of that content: extracting it, structuring it, and moving it into the queue. Together, they make sure WF2 (Content Generation) always has something to work with. By the time a post is being drafted, Zo.E has already surfaced a relevant source article, run it through WF7 (Content Research) for additional context, and passed it to WF2 to generate original commentary.
The machine is never staring at a blank page. That's the point.
When you set up the template pack, WF1 and WF3 ship with placeholder sources. One of your first configuration tasks is swapping those out for feeds that actually match your niche.
If you're in enterprise sales, you want RSS feeds from the publications your audience reads — industry blogs, trade press, analyst briefings. If you're in systems and automation, you want feeds from that world. The principle is the same either way: you're not telling Zo.E what to think. You're giving it a steady stream of real-world source material to react to. That's thought leadership, not content generation — you're building commentary on what's happening, not generating articles from thin air.
Take the time to build a short list of reliable sources. Five to ten good feeds is plenty to start. Quality over volume: one solid feed from a publication your audience trusts is worth more than twenty low-signal RSS imports. Once WF1 and WF3 are pointed at the right sources, the queue feeds itself.
The safety valve is Telegram.
When I have a specific angle I want to run with — a take on something that just happened, an idea that came up in a client conversation, something I want to say before the news cycle moves on — I send it directly. A topic, a brief, sometimes just a sentence. That message bypasses the discovery workflow entirely and goes straight into WF2 as a prompt.
It's the override. Most weeks I don't need it. But it's there, and it's fast. The system handles the steady state; I handle the moments when I want to get out in front of something.
If you're setting up Zo.E and you skip the WF1/WF3 configuration, the system will run for a bit on the placeholder sources, then slow down, then stop producing. That's not a bug. That's what happens when a machine runs out of input.
Engineer the input. The output takes care of itself.
Ta,
James
Founder | Nunlimited
