Build Log

How we build AI systems

Real systems we have built, shown end to end. The same engineering behind our AI Search Visibility work and our products. No theory, the actual architecture and the decisions behind it.

AI research pipeline

How we built an AI-search radar that writes our newsletter

One radar reads five live sources on AI search every week, keeps only what is new since last run, and drafts the write-up that becomes our newsletter.

June 2026 Read →
Email & list system

How we turned a weekly report into a newsletter system

The delivery system on top of the radar: double opt-in, a confirmed list as the single source of truth, two human approval gates, and one-click unsubscribe.

June 2026 Read →
Managed agent

How we built a research-brief agent on a managed cloud

A form becomes a company brief: an always-on managed agent researches on the model provider's cloud, then emails the result. No server of ours to run.

June 2026 Read →
Internal tooling

How we built an ops board an agent keeps current

An agent finds tasks from work sessions and meeting notes, but nothing reaches the board until you confirm. Then it shows up live.

June 2026 Read →
n8n automation

How we built a YouTube summarizer that emails a daily digest

A daily n8n job watches my tracked channels, summarizes new uploads with Claude, and emails me clean HTML digests. Plus a signup demo of the pattern.

June 2026 Read →
Our product

How we built Measr, our self-hosted analytics toolkit

Our own privacy-first analytics: an SDK captures in the browser, writes to the customer's own Supabase, and a dashboard reads it back. No cookies, and we run it on this site.

June 2026 Read →

We build systems like these for clients

Everything here runs on the same engineering we bring to client work. If you want your brand named in AI answers, that is where we start.

We also teach operators to build systems like this, hands-on. Balázs runs the build sessions at Agent-J+.