My Journey to an AI-Driven Workflow
I initially approached AI with caution. Integrating artificial intelligence into my daily work felt like a novelty rather than a necessity, and I lacked a clear vision for how these tools could genuinely enhance my productivity. Yet, through a series of deliberate experiments, mindset shifts, and strategic pivots, I’ve transformed my approach—and reshaped how I tackle a wide range of tasks.
My first forays into AI-assisted development were modest. I used Neovim paired with CodeCompanion and GitHub Copilot, appreciating their inline suggestions and chat interfaces but encountering friction at every turn. Copilot behaved like an overly expensive language server, and CodeCompanion’s custom keymaps and unpolished UI demanded more effort than payoff. Although I respected these early assistants’ promise, I remained unconvinced of their long-term value—until a single explainer video illuminated a far richer landscape.
That ten-minute overview introduced me to the major AI players—Google’s Gemini, Anthropic’s Claude, Meta’s Grok, and OpenAI’s ChatGPT—and explained techniques like chain-of-thought prompting and system prompts that provide long-term context. It also showcased emerging AI editors such as Cursor and Zed, which offer advanced autocomplete, refactoring suggestions, and inline feedback. In that moment of clarity, AI shifted from curiosity to cornerstone in my toolkit. I suddenly understood exactly how these capabilities could integrate into my workflow.
Armed with this new perspective, I subscribed to OpenAI and began crafting Custom GPTs tailored to my domains. Each personality carried its own knowledge base and leveraged OpenAI’s memory system to learn my preferences over time. On my phone, the ChatGPT app supplanted casual web searches—turning quick queries into deeper, conversational explorations. I found myself comfortable asking “dumb” questions, knowing that AI could serve as both research partner and personal tutor whenever inspiration struck.
Energized by success, I built Brainiac: a simple CLI tool powered by a local Ollama model to generate blog metadata—titles, keywords, and summaries—with a single command. Its successor, Brainiac-2, taps GPT-4 via OpenAI’s API, delivering richer insights and more consistent output. What began as a proof of concept has become a reliable workhorse rather than the next unicorn—though I’m already envisioning grander iterations to push its capabilities further.
Despite these triumphs, I still ran into friction. VS Code’s default keybindings and configuration conventions felt jarring compared to my preferred Rust workflows. Switching between modes, hunting for shortcuts, and wrestling with settings interrupted my focus. When I discovered Zed—with its native Rust support and familiar keybindings—the frustration melted away. Suddenly, I was back in my groove, able to work at peak efficiency.
Today, my AI ecosystem spans a suite of specialized avatars living on the ChatGPT browser site. I have a Senior Engineer persona for architectural guidance, an AWS Associate coach for exam prep, and a semantic commit-message generator that transforms git diff --staged | pbcopy into polished, convention-compliant commits. These assistants amplify my solo efforts, making me feel like part of a larger team.
Looking ahead, I’m brimming with ideas to weave AI more deeply into my projects: an MCP server to power Cobblepot’s SQLite databases, a retrieval-augmented system for NYC events discovery, and an AI-driven email digest aggregating news, product updates, and insights from my favorite sites. The era of powerful AI models is only beginning, and their integration into creative and technical workflows will only deepen. I encourage you to explore, experiment, and form your own perspective—because learning to leverage these tools now means you won’t be playing catch-up later.
These tools aren’t a temporary fad—they’re here to stay, and their capabilities will only grow more deeply woven into our creative and technical workflows. I encourage you to explore, experiment, and form your own perspective. Whether you embrace AI as a collaborator or test its limits before adopting, one thing is clear: it’s better to learn how to leverage these models now than to play catch-up later.