The site that says no.
Most AI agencies tell every prospect to buy. These guides make the case for restraint: when a website needs strategy before speed, when AI needs a foundation before automation, and when plausible is not good enough.
- 01
Thirty years of web design, explained through one plumber's website.
Geocities to AI slop, six eras and an interactive wayback machine, and the single question that ever made the phone ring.
6 erasRead - 02
Everyone is shipping a SaaS in 2026. Most are already dead.
I spent about 1,800 hours building one, then it crashed. The demo-to-durability gap, the security afterthought, the market flood, the maintenance cliff, and the sunk-cost trap.
1,800 hrsRead - 03
Why vibe-coded websites will not last.
AI can generate a website-shaped object. It cannot replace positioning, taste, conversion strategy, content systems, or maintenance discipline.
Read - 04
When not to use AI.
The five scenarios where adding AI guarantees a worse outcome. The compatibility check, written longform.
Read - 05
The dangers of the wrong AI.
What goes wrong when AI is deployed on top of bad data, bad scope, or a culture problem. Three near-miss field reports.
Read - 06
Human-AI integration is the product.
Why the locked tagline is not a slogan. The deterministic verification step is doing the load-bearing work, not the model.
Read - 07
We walked away.
Three engagements we declined and why. The owner not bought in, the data infrastructure absent, the culture problem in disguise.
Read - 08
AI will not replace your bookkeeper.
It will, however, replace 70 percent of the email she writes during tax season. The Maria/Austin pattern, generalized.
Read - 09
AI in legal: the Mata lesson.
Mata v. Avianca, the Florida elective share fabrication, and what every solo practitioner should require of any AI on a law firm site.
Read