Pi
All case studies
02 How · case study
Live

pickingsolutions.tech

151 Posts. Zero Developers.

151 published posts. Zero developers. No code written by hand.

One person, one AI co-pilot, and a content engine that runs itself before you wake up.

This is how pickingsolutions.tech got built — and what it cost to get it to run without you.

151
Published posts
0
Lines of code written by hand
200+
Sessions to build
~10mo
Phase 3 start to self-running
31
Automated jobs supporting the site
5
Content flywheel stages
9
AI agent profiles in personal stack
0
Paying users at Lumen launch
Timeline
Jun 2025

Blank Repo. Phase 3 Begins.

First session. Next.js 14, Prisma, PostgreSQL, Tailwind CSS. No prior full-stack experience. AI writes the code; the operator decides what gets built. The build-as-direction model from day one.

Jun–Aug 2025

Core Site: CMS, Blog, Categories

151 posts eventually published here. The editorial infrastructure — categories, author pages, the build log — built piece by piece across the Builder phase. 200+ sessions of incremental work.

Sep–Oct 2025

Lumen: The Pricing Page That Didn't Ship

Lumen dashboards and a premium layer built on top of the same repo. Pricing page was priority #1. Real user feedback arrived: "I don't know what this site is for." The pricing page was deprioritised entirely. The lesson cost sessions. It was the right call.

Nov 2025

Phase 4: Orchestrator. Automation Layer Begins.

Shift from building the product to building the machine that runs the product. launchd jobs start multiplying. Hermes agent profiles come online. The site stops being something you maintain and starts being something you direct.

Early 2026

Content Flywheel Designed and Deployed

5-stage pipeline formalised: Seeds → Propagation → Harvest → Review → Publish. content_reviewer.py deployed — autonomous approval for items scoring ≥22, max 5 surfaced daily for manual review. 298 stale backlog items processed on first run.

2026

151 Posts. Site Runs Itself.

Morning briefing at 07:00 seeds ideas. Flywheel propagates by 10:00. Reviewer processes the queue at 11:00. X publisher fires at 08:00. The editorial operation runs daily without manual intervention. The operator reviews what matters; the machine handles the rest.

I didn't build a content site. I built a machine that runs a content site — and then learned to get out of the way.
Key learnings

1. Build the machine, not just the product.

The instinct is to build the product and then think about how it runs. That's the wrong order. The content flywheel — 5 stages, automated daily — is what makes 151 posts sustainable for one person. Without the automation layer, the site would have stalled at 40 posts. Building the operational infrastructure first, not as an afterthought, is what separates a site that runs from a site that requires you.

2. User feedback killed the pricing page — and that was right.

Lumen was priority #1. A premium layer with dashboards, a pricing page, a subscription model. Real feedback from real people: "I don't know what this site is for." That feedback cost sessions of work to process. But building a product nobody understood would have cost more. The pivot — deprioritise Lumen, lean into content and the building-in-public angle — was the correct decision. The honest record of what didn't work is on this page because it's the most useful thing you can leave for someone building something similar.

3. Direction is the skill — code is not.

Every line of code on pickingsolutions.tech was written by AI. The operator's contribution was knowing what to build, in what order, and why. The architectural decisions — Next.js over static, PostgreSQL over a headless CMS, Abacus.ai over self-hosted — were made on domain reasoning, not technical preference. When you're not writing code, the quality of your decisions becomes the quality of the product.

Supporting evidence

How I Built a Content Site With No Developers

The full story of 151 posts, one AI co-pilot, and zero lines of handwritten code.

coming

The Lumen Pivot: What User Feedback Actually Looked Like

Real feedback, real cost, and why deprioritising the pricing page was the right call.

coming

Building in Public: What 151 Posts Taught Me About Content Strategy

The editorial strategy behind the site and what compounding content actually means.

coming

How the Content Flywheel Works — and Why It Took 10 Months to Get Right

Five stages, autonomous review, and the operational infrastructure that makes it sustainable.

coming