Buy, build, or vibe-code: the decision framework for every tool in your software stack

Issue 02 of “The Leverage Point” was a build story. I cancelled my business management software and replaced it with a system I built in an afternoon.

This issue is about the decision(s) that came before the build.

The build was the easy part. The decision is where operators can get it wrong, in both directions. So here are the four questions I now run before committing to any new paid software solution.

The three lanes (your options)

Every piece of software in your business sits in one of three lanes.

Buy. SaaS. This is the fastest path to value, and someone else carries the maintenance, security patches, and uptime. In exchange you rent forever, you conform your workflow to their opinions, and per-seat pricing scales against you as you grow.

Build. Here, you hire developers. You get exactly your desired workflow, but it costs five figures, takes weeks to months, and ends with a maintenance relationship you might not have planned for.

Vibe-code. You plus AI-powered coding tools. You get working software in days, sometimes hours, usually on AI subscriptions you already pay for. I've started calling this lane agile app deployment because "vibe coding" undersells what's actually happening, but vibe coding is the name that stuck, so there it is.

A year ago the third lane barely existed for non-developers. Now it's real, and it changes the math on the other two. That's the part most buy-vs-build advice hasn't caught up with.

The four questions, in order

Run them in sequence. The first question that gives you a clear answer ends the exercise.

1. Is this workflow how you differentiate, or is it commodity plumbing?

Payroll, accounting, email delivery: plumbing. Nobody hires you because your payroll runs well. Buy it, pick the boring market leader, move on. Only workflows that make you different earn the right to custom software.

2. Does an off-the-shelf tool fit 90 percent of how you actually work?

This is not about how you might work someday. It’s about how you work now. If a SaaS fits 90 percent, buy it and stop deliberating. The last 10 percent is not worth owning a codebase over. The trouble starts when the fit is more like 60 percent and you're paying full price to fight a tool. That was me in Issue 02: the software wasn't bad, it was wrong for a business whose workflow changes on the fly.

3. Does it touch other people's money, logins, or regulated data?

Payments, multi-user authentication, anything with compliance letters attached: buy it or hire professionals. This is the question that keeps vibe-coding honest. The lane is for tools where you are the user and your data is the data. It's not for a thing your clients log into.

4. What's left after the first three?

“Does it have a small blast radius if it breaks on a Tuesday afternoon?” That's the vibe-code lane. More of your stack sits there than you'd think.

The receipts

My own stack, run through the framework.

Still bought, happily: Google Workspace and my accounting software. Workspace is plumbing (question 1) and infrastructure I'd never rebuild. Accounting fails questions 1 and 3 at the same time: commodity plumbing with compliance attached. No way I'm vibe-coding that.

Also bought, as the infrastructure the third lane runs on: Supabase for databases, Cloudflare for domains and security, and inference (AI) credits from Anthropic (Claude), OpenAI (ChatGPT), Google (Gemini), Perplexity, and OpenRouter. That inference spend is moving out of experimentation mode and into developed systems. Lovable.dev runs the website, for now. It's the right tool today and it gets replaced when the SEO-optimized build is ready.

Vibe-coded: the business operating system from Issue 02, client tracking, time and billing, radar summaries, built in an afternoon on subscriptions I already had. I also have my “second brain” system, managed on Supabase plus Obsidian, because no notes SaaS fit the way I actually send and retrieve knowledge. Each one failed questions 1 and 2, passed question 3's safety check, and landed in lane three.

The pattern across all of them: I buy plumbing and infrastructure, I vibe-code the workflows that make MODRN (my brand) different.

Where the third lane bites

The third lane has teeth.

You are the vendor now. When it breaks at 11 am before an important client call, there’s probably no support ticket to file. There's you and your vibe-coded solution. Budget for that before you build.

Security is a skill you don't get from vibes. Single-user tools running on your own data are one thing. The moment a tool faces other people, question 3 applies with no exceptions.

The failure point points back at you. A tool nobody else can run or fix is fine for a solo operator, but a liability the day you hire someone. Write down what your system does and how it works while you still remember. Documentation in the age of AI is critical.

None of the points above kills lane three. It just prices it correctly. An afternoon to build is real, but so is the ownership that comes after. Budget accordingly.

Do this in the next 24 hours

Pull up your software spend. Pick the one software line item that annoys you most every month. Run it through the four questions in order and actually price the third lane before you pay again, in hours instead of dollars. If the answer is "one afternoon and I already pay for the tools," you have a decision to make.

What I shipped

A content hub for this newsletter: a single-file kanban board that moves every piece from idea to ship, with a sign-off gate at each stage. One .MD file, no vendor, no login. Built the same afternoon-sized way this issue describes.

Tools I'm testing

Claude Code's /loop command on the new Fable 5 model. I pointed a self-pacing loop at this issue's pipeline and it researched, outlined, and woke itself up when it needed my input. Honest take: the bottleneck is now me answering its questions.

Tools I'm building

An complete A-Z direct mail lead-generation system managed entirely by AI. I’m taking the AI leverage offline. More on this in issue 04.

Pattern I'm watching

Operators quietly replacing per-seat SaaS with small owned tools, one workflow at a time. Not a mass exodus, a slow leak. The renewal conversation is changing from "which plan" to "do we need this at all."

Find the leverage,

-Terence