My AI-Assisted Software Development Workflow

A Unique Approach to Building with AI?

Here’s my process when I build. You can call it “vibe coding,” but I prefer “assisted software development.” And, I don’t know if my approach is unique, but the more I’m online and on LinkedIn, and the more I see people complaining about AI—how hard it is to use, how “non-intelligent” it is, or how it overcomplicates things—the more I think that the way I work might be a little different than most as I just don’t struggle with the same issue others do.

One of the main things I do is use multiple versions of the same LLM for different parts of the workflow.

Here’s what I mean:

Phase 1: System Design and Architecture

When I’m building an application, I typically start by designing it in a regular Claude chat (the web browser version). For this stage, I use a model like Sonnet 4.6 (but now Sonnet 5; things change quickly in the realm of AI), which is excellent conversationally and intelligent enough to help me build and design my systems without burning through my token usage. Model distinction is important because you don’t need a token-heavy model like Opus 4.8 (and now the newly reinstated Fable 5) just to build out an implementation plan. Sonnet 4.6 is good enough to flag issues and create a high-level execution plan, and then I

Read more

I used to pride myself on being a ChatGPT power user, but then I switched to Anthropic's Claude ecosystem—by which I mean Claude Desktop (Cowork and Code), the standard browser platform, and Claude Developer Workbench—and it transformed the way I build with LLMs. I'm still a power user, and I've essentially become a developer by conventional means rather than the traditional path.

In that process, I've developed some systems that work, and I wanted to share them with others building in this space. The trifecta I mentioned above has become this assisted software developer’s game-changing tool. And it all starts with Claude Projects.

Context Management & The Multi-Chat Workflow

Project Initialization and the "Limiter" Prompt

When I start building something new, I set up a Claude Project and immediately apply a foundational set of system instructions. The most critical instruction I drop into every single project is:

Always verify before you create or edit anything.

Large Language Models (LLMs) are naturally eager. If left unchecked, they will "run away" with a task, recreating files unnecessarily or regenerating the same context or code over and over. When an LLM over-generates like this, it quickly loses context, forgets past constraints, changes things you wanted to keep, and burns up session tokens.

The problem is that you might love a specific output but want to make a minor tweak. If the AI is too eager, it might rewrite the entire file from scratch and ruin what you already built.

The solution is to implement a strict limiter prompt. I instruct the AI to always review the current state and explicitly confirm its planned changes before generating or editing any files.

While I dynamically update project instructions as the project develops, this specific verification rule is a permanent fixture. It completely prevents runaway task creation, which saves an immense amount of time and token usage.

Here's what I drop into every project instructions window:

Read more

The Original Mission vs. Client Acquisition

When I first started this Substack, my core goal was simple: to provide real value to other builders in the AI space. Somewhere along the line, I lost sight of that. I started focusing too much on pitching to traditional business owners, trying to convert them into clients.

Overcoming Market Resistance

The reality is that there is a massive amount of friction and resistance when it comes to business owners adopting AI. You can’t force it; they need to come to their own realizations about what actually works for their operations. Instead of pushing a hard sell, the most effective approach is simply to show them what I have successfully built and let them convince themselves of its value. Because of that friction, I’ve decided to pivot my focus back to where I started: supporting other builders.

For the Non-Technical

Given my non-technical background (I’m effectively a self-realized Assisted Software Developer), I have a genuinely unique perspective to share, based on the practical, real-world systems I have built for myself and my clients. In the process of learning and developing in this space, I’ve created some systems that work well for me, and I wanted to share them with others also building in this space.

What is Assisted Software Development (ASD)?

ASD is the process of building real tools and working systems with AI handling the technical execution, while you bring the problem definition, the judgment calls, and the system architecture. It’s not “vibe coding”; it’s more than that. ASD sits in a grey area between no-code templates that fall apart the second you need something custom and traditional software development. The development and coding processes are similar, just without the formal background. It’s AI-first coding, and it’s a game-changer for anyone who loves to create but doesnt know how to read or write code.

This new section of my Substack brings an ASD practitioner’s approach to building things that actually work, even if you’ve never written a line of code in your life.

In upcoming articles, I’ll explore how I work within the Claude ecosystem to create real-world applications like Top AI News Today, MoveSmart, lead-generation automations, a voice memo-to-auto-SEO blog-post AI Agent for a client in Orange County, CA, and more.

Most of the articles in this section will be available only to paid subscribers, as they are based on thousands of hours of research, experimentation, execution, painful debugging, and hard-earned lessons.

Subscribe now

I’ve scheduled my first real post in this section—How To Set Up a Claude Project: A Non-Technical Builder’s Guide To Creating Token-Efficient and Safety-First Claude Projects—to publish this coming Friday (June 26).

I’m looking forward to leaning in and sharing what I’ve learned. I hope you’ll find it valuable.

Happy ASDing!

—BottBott