Hand-Crafted Content Systems

I build content engines that compound.

Strategy, systems, execution, and results — end to end.

See the work Start a conversation

or just scroll

I · approach

How I work

I.A

Real inputs over LLM filler

Every piece worth writing starts with a customer or an SME. AI is leverage on real signal, not a substitute for it.

I.B

If I can't find it, I build it

If no tool does what I need, I build my own — shaped to how I actually work, not how a vendor imagined it.

I.C

Narrative that converts

Story and funnel aren't at odds. The piece that truly lands is the same one that earns the demo.

II · work

Selected work

A publishing pipeline that replaced a $1k/month tool

Fig. 01 — Publishing pipeline architecture, 2025. Replaced a 20–30 minute manual process and a $1k/month tool.
Field notes & walkthrough
Situation
At Goodshuffle Pro, publishing a blog to WordPress took 20–30 minutes of manual work every post. Because of how the site's backend is set up, you hand-place each section and image and set the formatting one at a time. High friction, every single post.
What I built
First I codified the flow as an n8n workflow: Google Doc in, [image placeholder] tags matched to a Drive folder, pushed to the WordPress REST API with alt text and featured_media set. That cut a 15–20 step process to a click. Then I rebuilt it in Claude Code: paste the Asana task and ~3 minutes later it returns a ready-to-publish draft with formatted body, comparison chart, takeaways, images, CTA, and FAQs, plus a company announcement.
What happened
A 20–30 minute manual process collapsed into one or two clicks. Tools like AirOps charge thousands a month for similar pipelines; this one runs on an API key.

A content library that turns a question into a ready-to-send email

Fig. 02 — Content library, built 2026. Indexed website and YouTube assets, semantic search, and one-click drafts by persona and funnel stage.
Field notes & walkthrough
Situation
Our content lived all over the website and YouTube, from webinars to case studies to blogs, with no good way to find the right asset for a given prospect. Sales and CS dug through folders by hand, and content usage rates stayed low because nobody could see what existed.
What I built
An internal system that indexes every asset across our website and YouTube, filterable by type, funnel stage, vertical, and persona, with previews and “best use when” notes. On top, a semantic layer: ask a question, it pulls and compiles the right sources, then drafts the answer as a prospect email, text, or internal memo, ready to send.
What happened
Finding and packaging the right content went from a manual hunt to a single question. Now sales, CS, and marketing can all self-serve the right asset for a persona and stage in seconds, built to lift content usage rates across the org.

A skill library that runs a one-person content team

Fig. 03 — Content-ops skill library, two dozen-plus codified workflows around one operator and growing. Built 2025.
Field notes & walkthrough
Situation
As effectively a one-person content engine at Goodshuffle, I needed to scale myself without re-explaining my standards every time. The recurring, judgment-heavy work was eating time that should go to the writing and strategy itself.
What I built
A library of two dozen-plus codified skills that encode my editorial standards and turn recurring work into repeatable systems, from drafting the newsletter to formatting blogs to prepping standup. Each shares the same backbone: when to use it, the rules, and the guardrails that keep the output sounding like me, not generic AI. Drop an interview into the newsletter skill and it comes back a near-final draft, ready to edit.
What happened
Content operations became systematized rather than manual. A newsletter that used to start from a blank page now comes back ~90% ready, on-brand and formatted, so my time goes to interviews and strategy instead of production. A one-person function ships at the cadence of a larger team's.

III · bio

About

Celita Summa
Pl. 01 — Celita Summa, 2026.

Stack

  • Claude Code
  • n8n
  • WordPress
  • HubSpot
  • Python
  • Mixpanel

I trained as a writer and ended up coding.

I run content marketing at Goodshuffle — software for the event rental industry. Over the past year I've built around twenty AI-augmented systems that let a one-person content function ship at the cadence of a full team.

Before that: content roles at Uberall, plus seven years of freelance B2B SaaS work. I once gave a TED-style talk on Verdi's late style at a team retreat, which probably says more about how I think than any skills list would.

IV · write to me

Get in touch

Currently writing & building from New York. The best way to reach me is snail mail, but the fastest way is email.