Tools My Tools Can Use
A small project for a non-profit client revealed why I'm choosing different tools now. I no longer want tools that expect me to use them. I want tools that my tools can use. The future isn't no-code, it's AI-orchestrated code.
I've been choosing very different tools lately than I would have even a few months ago.
A small project for a non-profit revealed a change in my thinking and perhaps how we will all work.
Recently, as I've gotten comfortable calling myself a software engineer again (or, more accurately, not worrying about the labels), I took on building a small app for a non-profit I'm working with. They needed a couple interfaces, a database, and some data pipelines. My first instinct was to connect their existing survey tool to Airtable, a lovely platform where you can cobble together what are essentially websites and databases without writing much code.
But as I was thinking it through, I realized something: building it fully bespoke would actually be less work for me.
The New Workflow
I have Claude Code and Cursor set up to interact with tools and APIs directly. And suddenly, I don't want to work with tools that require me to click around in some web interface to configure things.
Stripe, with its comprehensive APIs and command line tools? Perfect. I can configure everything programmatically.
Supabase with its CLI? Once the API keys are set up, I can manage databases and permissions without ever opening a browser.
Google Cloud's CLI? Partially there. Better than nothing.
I no longer want tools that expect me to use them. I want tools that my tools can use.
This preference for CLI-first tools isn't just about my comfort. It fundamentally changes what's possible.
Why This Matters Beyond Just Convenience
In the past, I would have chosen a no-code solution for this non-profit project for one big reason: ownership. If a client doesn't have an engineering team, giving them a bespoke app is great—until something breaks or they have questions. Then it becomes a real challenge.
I can now show them how to point an LLM at their codebase and ask, "Hey, what were the meth-addicted trash pandas who built this thinking?" And it will actually explain the code in terms they can understand.
The documentation and support requirements have fundamentally changed. Non-technical stakeholders can now have meaningful conversations with codebases through AI.
When Dashboards Became Obstacles
This shift reveals something important about how we've been building tools.
I've long realized that AWS was no longer a tool for me. When I open that dashboard—that essential everything store for cloud services—it's not even designed for a casual engineer anymore. Setting up the permissions, billing, and monitoring is a project unto itself.
But I can see a world where adequate markdown documentation and tooling lets Claude Code help me understand which of AWS's seventeen database systems I actually need. Where I can have my own bespoke version of Vercel—that brilliant hosting platform with amazing DevOps support—running on DigitalOcean, AWS, or anywhere else, orchestrated entirely through AI agents and CLI tools.
The Unix Philosophy Meets AI
What's particularly powerful is that Claude Code, running on my Unix-based Mac, has access to the entire ecosystem of Unix tools. It's not just about writing code anymore—it's about orchestrating entire workflows through tool usage. The whole MCP and tool usage discussion dovetails amazingly with the Unix philosophy of small special purposed tools that can be interwoven to great impact.
This isn't just a productivity hack. It's a fundamental shift in how we think about building software. The boundary between "tools for developers" and "tools for AI" is dissolving. And the tools/products/apps that will win are the ones that understood this shift early: the ones with great APIs, comprehensive CLIs, and programmatic access to everything.
The future isn't no-code or low-code. It's AI-orchestrated code, where the tools we choose are determined not by how easy they are for us to click through, but by how effectively our AI assistants can wield them on our behalf.
That future is already here for some. The divide isn't between coders and non-coders anymore. It's between those working with AI-orchestrated tools and those still clicking through dashboards.