// Blog
Practical lessons from building AI workflow systems.
AI finally pays off for small businesses: $500–$2,000 and 20+ hours saved a month, with returns most owners see in the first half-year. Here's where the money actually comes from, what it really costs, and why a stable foundation matters more than the model.
Some developers say Claude Opus 4.8 feels worse than 4.7 — more refusals, shorter answers, no clear gains. Here are the real causes and the fixes that actually work.
For two years the AI conversation was about capability. In 2026 it quietly became about cost. Here's what the shift means — and why teams that built lean are suddenly the ones with the advantage.
A non-programmer's account of steering a VPS-hosted AI automation stack from a phone: servers, automations, nightly backups — and why it works at all.
Context engineering is emerging as the real skill gap between hobby AI projects and production systems. Here's what we learned after 50+ days of daily AI-assisted development.
How we run a full AI-assisted development and automation stack on a single VPS for under €15/month — and the trade-offs that come with that choice.
AI-assisted projects silently accumulate information loss between sessions. Here's the real pattern, why it happens, and the systematic approach we use to stop it.
We deployed a Caddy config update that appeared to succeed but changed nothing. Here's the exact failure, why it happened, and the one-line fix that works every time.
I was skeptical of the AI easy money claims — so I ran the experiments. Amazon affiliate through Pinterest, AI image tools, drift that kills real projects: here's what the numbers showed, and what we built from it.
Single-agent AI setups hit ceilings fast. Here are the coordination patterns we use in production — including the mistakes we made getting there.