This week's sessions + two new resources you can use right now


We had two AI packed sessions last week in Beware the Default and I wanted to send a "quick" recap plus give ya'll some goodies...

🍻 AI Happy Hour

On Wednesday, three different people asked three versions of the same question.

  • Charles asked which tool to start with
  • Deon asked why his dashboards kept breaking
  • Max V. asked when to use Cursor vs. Claude Code

No one seems to have a decision framework for where to build what.

So I built one.

"Where Should I Build?" Decision Framework

  • Six tiers w/ a two-axis quiz that measures your technical comfort AND your project complexity separately, then maps the intersection to a recommended path (it's not as confusing as that sounded).
  • Copy-paste starter prompts at every tier.
  • A Lovable vs. Bolt comparison that explains the actual pricing mechanics behind each answer. And a Terms 101 flashcard deck covering front-end, back-end, PRD, tokens, GitHub, terminal, API, markdown, and more.

Start with "Find My Path" if you're not sure where you belong. Then open the tier it recommends and copy the starter prompt directly into the tool. And then get on buildin'!

The whole thing came from one session.

Beyond the tool question, the principles that came up on Tuesday are worth repeating.

Normally the resources from these calls are reserved for community members but we must be feelin it from the Happy Hour...

Check it out here.

A few other things we covered...

Start with the emotion, not the feature
When we were walking through how to build a coaching app, the first question that I tell people always to ask is,

"What is the primary emotion a user should feel when they open your app?"

That answer shapes everything. Colors, button sizes, animation speed, copy. I showed an example where we designed a credit monitoring app around the emotion of "calm" and the AI suggested making buttons larger because users freaking out over someone stealing their credit cards might have shaking hands. You would never think of that starting from a feature list.

Plan before you prompt
AI will try to get you from where you are to where you want to go as fast as possible. That sounds good until you realize it's skipping steps to...get your there as fast as possible. Force it to think through a plan first.

  • "Give me a step-by-step process for how you would build this."
  • "Tell me why this order is better than the alternative."

When it has to justify the sequence, the quality of every step goes up because each one needs to connect to the last.

Do as much as you can in Claude chat before you move to a building tool
Every message in Lovable, Cursor, or Claude Code costs tokens. If you brainstorm, refine your prompts, test your conversations, and build a full PRD inside Claude first, you save credits and ship a better first version. Tell Claude where you're headed so it preps your files in the right format for the next tool.


The Build Lab

Friday was all execution. Tam walked the group through setting up Obsidian + Claude Code from scratch.

The short version of why we landed on Obsidian:

Everything is a markdown file.

That's it. Markdown is what AI reads best (don't listen to that nonsense about html). It's what Claude Code writes in. It's what every new tool will play nice with and your vault is just a folder on your computer.

  • No proprietary format
  • No API that breaks at 2am
  • No $10 per 1,000 credits (we learned that one the hard way testing Notion agents, $320 in a month of testing alone)

Tam ran the full setup live. Obsidian installed, Dataview and Templater plugins configured, a custom GPT interview to design each person's vault architecture based on their actual workflows, and then Claude Code built the entire folder structure and templates in real time.

Still feelin good from the Happy Hour, we are giving out the interactive walkthrough to everyone. Should take about 30 minutes.

Obsidian + Claude Code Setup Guide

It covers every step with a clickable tracker, common troubleshooting fixes, and a copy-paste prompt for the /log-session bonus skill (which automatically summarizes your Claude Code sessions and writes them to your vault so you never lose what you learned).

A few principles from the session worth keeping.

  • Simple beats sophisticated. We have all built dashboards that we stopped using. Several versions of them. The more connections and APIs you add, the more things break silently. You end up operating on stale data and don't even know it. Our new approach is simpler. Keep everything in clean markdown files. When you want a view, just ask Claude Code to spin one up on the spot. Your dashboard changes based on the question you ask instead of being a static thing you have to maintain.
  • Bidirectional linking makes your vault smarter over time. This used to be a manual pain. Now you just tell Claude Code "go find the files that are related and link them for me" and it handles the hashtags and connections. When you look at a client note, every project they're involved in shows up automatically. When you look at a project, every related person and resource is right there.
  • Slash commands turn your vault into a control center. I type /today in terminal and my agenda, task list, and highlights populate automatically in Obsidian. You can build a /braindump command that processes a voice transcript, sorts it into the right folders, and generates a dashboard. Repeatable. Every time.

One more thing.

I'm putting the finishing touches on something called the Rubber Ducky Robot app. It guides you step by step from a raw idea to a finished PRD (that's a Product Requirements Document, basically the blueprint your building tool needs to do good work) and then holds your hand through actually building it in Lovable, including giving you the exact prompts at every stage.

Named after the old coding practice where developers would debug problems by explaining them to a rubber duck on their desk. Sometimes just having something that forces you to think out loud is the whole fix.

This one will be for community members only.

We do these sessions twice a week. Wednesdays at 4pm ET (Happy Hour) and Fridays (Build Lab). They're open, they're live, and we build real things on the spot. If you have something you're stuck on, bring it. We'll work through it together.

Come hang out with us...we promise you'll love it or Big Zain will buy you dinner (seeing if he read this far).

Beware The Default

Stay Curious,

Click below to discover your Cognitive Fingerprint.

Subscribe to Max Bernstein