AI experiments, in the open
AI is a multiplier, not a magic trick. Here's what I'm testing, what worked, and what didn't — with the receipts.
What's running in the lab
Ask Shubham
A chatbot trained on every article, case study, and framework — answering in my voice, with citations.
Result: RAG pipeline over published content. Private beta in Q3 2026.
Semantic site search
Intent-based search that understands 'growing SaaS on a small budget' and surfaces the right five pieces.
Result: Vector index across all content. Early results: 3x better relevance than keyword search.
AI content QA rubric
An automated editorial rubric that pre-checks drafts against a quality bar before a human ever sees them.
Result: Cut editing time 40% while raising the average quality score.
30-day autonomous social test
Let an AI agent run a social account for 30 days with guardrails, then measured what actually happened.
Result: Engagement held; voice drifted. Conclusion: AI accelerates, humans must steer.
Follow the experiments
I document the wins and the failures in the newsletter and in the free tools.
Try the live tools