THINKING
IN PUBLIC
Not teaching — thinking out loud. Product strategy, AI implementation, growth experiments, and lessons from building.
A Skill in Claude Code Is Convenient. Right-Click Without Claude Is Even Better
The average office worker switches between apps 1,200 times a day, losing 23 minutes per interruption. Here's how moving two daily workflows — Markdown-to-PDF and audio transcription — from LLM skills into the right-click context menu made them feel like magic.
How to Build Your Own AI News Digest
I built an open source AI news digest that parses 30 RSS sources and writes posts in your style — here's how to spin up your own version for any niche using Claude Code.
Synthetic Research + ODI: 2 Weeks of User Research in 30 Minutes
A founder came to me with 15 feature ideas and zero clarity. I used LLM-generated personas and Opportunity Scoring to prioritize — in one coffee break.
I Couldn't Find the Right AI News Source, So I Built My Own
Couldn't find an AI news source that covered everything I care about — new releases, funding rounds, market shifts — so I built a bot that scrapes 27 sources and compiles a daily digest. Here's how it came together.
My AI tools were eating 23,000 tokens before doing any actual work.
My AI tools were eating 23,000 tokens before doing any actual work. MCP servers load their full schema on every prompt.
How to Boost Your AI Workflow Speed by 3.75x (Almost)
Speaking is 3.8x faster than typing, yet most people still type when prompting LLMs. Here's why voice input is the obvious upgrade — and which tools actually work.
First Episode of 'Big Potential' Podcast: Opus 4.6, GPT 5.3, and a Robot That Refused My Refund
Two product managers break down a week of AI news — Opus 4.6 and GPT 5.3 landing on the same day, AI agents stuck in infinite arguments, and a vending machine that decided refunds weren't in its best interest.
When building AI agents, I keep seeing the same core problems.
When building AI agents, I keep seeing the same core problems. Especially in wellbeing / support use cases: Tone of voice (TOV) — consistency, warmth, non-judgment across every interaction.
One Prompt → 3 APIs → a Ready Skill
I asked Claude Code to build a content-generation skill in my style — it pulled data from Notion, Telegram, and RAG on its own. One prompt replaced an hour of manual work.
Why Everyone Is Suddenly Obsessed With Clawdbot (And Why Claude Code/Cowork Doesn't Spark the Same 'Wow')
A look at why Clawdbot is gaining traction as a local AI assistant that actually executes — and where Claude Code falls short on proactivity and persistent memory.
Glossary: RAG, MCP, LLM, HRV
A plain-language glossary covering four terms that show up constantly in AI and health discussions: RAG, MCP, LLM, and HRV. No fluff, just the definitions.
Why I Love Vibe Coding
A quick look at how vibe coding led to building 7 services in a couple of months — including self-hosting n8n.
We Have a 1% Chance to Avoid a World Where One Player Decides If You Deserve AI Access
The Liberman brothers break down why AI is becoming geopolitical infrastructure — and why we have roughly a 1% chance of avoiding a future where a handful of American companies, backed by the US government, control who gets access to it.
I Spend $180–250/Month on AI Tools
A breakdown of the $180–250/month AI toolkit I currently use — ChatGPT, Gemini, Replit, Cursor, Claude, and APIs — with honest notes on what each is actually good for and where each one falls short.
More content on my channels