← Back to posts
Mar 19, 2026
I Used to Spend Weeks Building What Now Takes Me Hours. Here's What Changed.
AI coding tools aren't just about writing code faster. For freelancers and small teams, they're quietly changing what's even possible to build alone.
A week ago, I sat down to build an application for a client. Not a simple one either. We’re talking custom internal tools specific to their business. The kind of project that used to mean five to nine weeks of heads-down work, minimal sleep, and at least one moment where I questioned my entire career pivot into tech.
I finished a working version in a weekend.
Now, I want to be careful here because “finished in a weekend” can sound like the kind of thing people say in X threads right before they’re trying to sell you a course. So let me be specific about what that actually looked like and what it means for freelancers and small dev teams right now in 2026.
-----------------------------------------------
What’s Actually Happening with AI Coding Tools
-----------------------------------------------
The conversation around AI and coding used to be about autocomplete. Tools that finished your sentences, suggested the next line, saved you from typing the same boilerplate for the hundredth time. Useful? Sure. Transformative? Not really.
That version of the story is over.
What’s happening now is different in kind, not just degree. Tools like Cursor, Claude Code, and GitHub Copilot aren’t just completing your code anymore. They’re reading your entire codebase, reasoning across multiple files, running commands, flagging errors before you see them, and in some cases working autonomously through entire features while you step away. Choosing an AI coding assistant in 2026 is no longer about which tool gives the best inline suggestion. It’s an architectural decision about how you want to build software.
That framing matters because it changes what questions you should even be asking.
“The question used to be: can AI help me write this faster? Now the question is: what can I build now that I couldn’t realistically build before?”
-----------------------------------------------
What This Actually Means If You’re a Freelancer
-----------------------------------------------
Here’s the honest version of this story that you won’t find in a press release.
When I came from a sales background into software development, the thing that terrified me wasn’t the learning curve. It was the gap between what clients wanted and what one person could realistically deliver. A client would describe something they needed, and in my head I’d be quietly doing math: how many hours, how many context switches, how many late nights to make this happen at a price they’d actually pay.
That math has changed. Significantly.
AI coding tools don’t just help you write code faster. They compress the distance between idea and working software in a way that changes your capacity as a solo operator. Three things I’ve felt most concretely:
You can take on more complex projects. Not because you suddenly know more than you did, but because the cost of exploring unfamiliar territory drops. When I was building the internal tool, I wasn’t starting from zero every time I hit a new problem. I had a thinking partner in the tool that could reason through the problem with me.
Documentation and handoffs stop being the thing you always skip. One of the quiet disasters of solo freelance work is that documentation almost never happens. There’s no time, and there’s no one to hold you accountable. AI tools that can auto-document complex codebases and summarize pull requests change this dynamic in practical, not theoretical ways.
You stop underselling yourself on scope. This one is harder to quantify, but it’s real. When you know your effective output has increased, you quote differently. You take on work that would have previously been out of reach for a team of one.
-----------------------------------------------
The Part Nobody Talks About Enough
-----------------------------------------------
I want to push back on something, though, because this piece would be dishonest if I didn’t.
AI coding tools are genuinely powerful. They are also genuinely imperfect in ways that matter. The tools that work autonomously on complex tasks are still making meaningful errors in high-stakes situations. Research from places like Anthropic and Carnegie Mellon has found that agents aren’t reliable enough yet for processes where getting it wrong is expensive.
For freelancers, this shows up in a specific way. The tool can help you build something quickly. The tool cannot replace your judgment about whether what was built is correct, secure, or actually solves the client’s problem. That judgment is still yours.
Which, honestly? That’s also where your value lives. The AI is fast. You are the one who understands the client’s actual situation, knows what questions to ask, and can catch when something technically works but still misses the point entirely. That combination is what you’re selling.
-----------------------------------------------
Where to Start If You Haven’t Already
-----------------------------------------------
If you’re a freelance dev or running a small team and you haven’t seriously integrated AI coding tools into your workflow yet, here’s the honest entry point I’d suggest:
Start with one tool and use it inside a real project, not a sandbox. Cursor is a solid choice if you want AI baked into your editor without switching your whole environment. GitHub Copilot makes sense if you’re already deep in that ecosystem. Claude Code is worth exploring if you do a lot of multi-file reasoning and complex refactoring. Most have free tiers or low-cost starting points, so the barrier is more about time than money.
Then pay attention to where you feel the friction disappear and where it doesn’t. The tools are not all the same, and the best setup for a solo freelancer building client sites is not the same as the best setup for a small product team maintaining a SaaS codebase.
The thing I keep coming back to is this: I did not get into software development because I wanted to spend my time fighting boilerplate and rebuilding infrastructure from scratch on every project. I got into it because I wanted to build things that solve real problems. AI coding tools, at their best, close the gap between where your energy goes and where it should go.
That’s not hype. That’s just what it feels like when a tool actually works.
Let’s build it beautifully,
Fab
I finished a working version in a weekend.
Now, I want to be careful here because “finished in a weekend” can sound like the kind of thing people say in X threads right before they’re trying to sell you a course. So let me be specific about what that actually looked like and what it means for freelancers and small dev teams right now in 2026.
-----------------------------------------------
What’s Actually Happening with AI Coding Tools
-----------------------------------------------
The conversation around AI and coding used to be about autocomplete. Tools that finished your sentences, suggested the next line, saved you from typing the same boilerplate for the hundredth time. Useful? Sure. Transformative? Not really.
That version of the story is over.
What’s happening now is different in kind, not just degree. Tools like Cursor, Claude Code, and GitHub Copilot aren’t just completing your code anymore. They’re reading your entire codebase, reasoning across multiple files, running commands, flagging errors before you see them, and in some cases working autonomously through entire features while you step away. Choosing an AI coding assistant in 2026 is no longer about which tool gives the best inline suggestion. It’s an architectural decision about how you want to build software.
That framing matters because it changes what questions you should even be asking.
“The question used to be: can AI help me write this faster? Now the question is: what can I build now that I couldn’t realistically build before?”
-----------------------------------------------
What This Actually Means If You’re a Freelancer
-----------------------------------------------
Here’s the honest version of this story that you won’t find in a press release.
When I came from a sales background into software development, the thing that terrified me wasn’t the learning curve. It was the gap between what clients wanted and what one person could realistically deliver. A client would describe something they needed, and in my head I’d be quietly doing math: how many hours, how many context switches, how many late nights to make this happen at a price they’d actually pay.
That math has changed. Significantly.
AI coding tools don’t just help you write code faster. They compress the distance between idea and working software in a way that changes your capacity as a solo operator. Three things I’ve felt most concretely:
You can take on more complex projects. Not because you suddenly know more than you did, but because the cost of exploring unfamiliar territory drops. When I was building the internal tool, I wasn’t starting from zero every time I hit a new problem. I had a thinking partner in the tool that could reason through the problem with me.
Documentation and handoffs stop being the thing you always skip. One of the quiet disasters of solo freelance work is that documentation almost never happens. There’s no time, and there’s no one to hold you accountable. AI tools that can auto-document complex codebases and summarize pull requests change this dynamic in practical, not theoretical ways.
You stop underselling yourself on scope. This one is harder to quantify, but it’s real. When you know your effective output has increased, you quote differently. You take on work that would have previously been out of reach for a team of one.
-----------------------------------------------
The Part Nobody Talks About Enough
-----------------------------------------------
I want to push back on something, though, because this piece would be dishonest if I didn’t.
AI coding tools are genuinely powerful. They are also genuinely imperfect in ways that matter. The tools that work autonomously on complex tasks are still making meaningful errors in high-stakes situations. Research from places like Anthropic and Carnegie Mellon has found that agents aren’t reliable enough yet for processes where getting it wrong is expensive.
For freelancers, this shows up in a specific way. The tool can help you build something quickly. The tool cannot replace your judgment about whether what was built is correct, secure, or actually solves the client’s problem. That judgment is still yours.
Which, honestly? That’s also where your value lives. The AI is fast. You are the one who understands the client’s actual situation, knows what questions to ask, and can catch when something technically works but still misses the point entirely. That combination is what you’re selling.
-----------------------------------------------
Where to Start If You Haven’t Already
-----------------------------------------------
If you’re a freelance dev or running a small team and you haven’t seriously integrated AI coding tools into your workflow yet, here’s the honest entry point I’d suggest:
Start with one tool and use it inside a real project, not a sandbox. Cursor is a solid choice if you want AI baked into your editor without switching your whole environment. GitHub Copilot makes sense if you’re already deep in that ecosystem. Claude Code is worth exploring if you do a lot of multi-file reasoning and complex refactoring. Most have free tiers or low-cost starting points, so the barrier is more about time than money.
Then pay attention to where you feel the friction disappear and where it doesn’t. The tools are not all the same, and the best setup for a solo freelancer building client sites is not the same as the best setup for a small product team maintaining a SaaS codebase.
The thing I keep coming back to is this: I did not get into software development because I wanted to spend my time fighting boilerplate and rebuilding infrastructure from scratch on every project. I got into it because I wanted to build things that solve real problems. AI coding tools, at their best, close the gap between where your energy goes and where it should go.
That’s not hype. That’s just what it feels like when a tool actually works.
Let’s build it beautifully,
Fab