I Got Laid Off Last year. So I Built My Own Job Hunting AI App.

What happens when a creative stops complaining about a broken system and just builds the fix?

Six months ago, I got an email.

You know the one. The kind where you read the first line and your stomach drops before your brain even catches up. After months of grinding, of late nights, of giving everything to something that wasn’t mine, I was out.

I won’t pretend I handled it gracefully. I sat with it for a while. But underneath the shock, something else crept in. Relief. Actual, genuine relief. The kind that tells you something was wrong long before the email arrived.

So I did what any slightly unhinged creative would do. I took the time off.

I traveled. I shot videos. I had conversations I’d been putting off for months. I slept properly for the first time in years. And for a while, it felt like the best decision I’d ever made.

Then my savings account had a different opinion.

The Problem I’d Forgotten About

When the numbers started looking uncomfortable, I dusted off my résumé and jumped back into the job market. And I was immediately reminded of something I had buried from the last time I’d done this.

Job hunting is exhausting.

Not because good opportunities don’t exist. They do. But finding them requires a kind of sustained, repetitive attention that drains everything creative in you.

Every morning I’d open six browser tabs, scroll through a dozen Telegram channels, check LinkedIn, monitor Twitter #hiring posts. Then do it again in the afternoon. Then again at night, just in case.

And still. I kept applying too late.

The role would go up, hundreds of people would apply within hours, and by the time I found it the window had already closed. I was spending more time searching than most people spend at work, and still coming up empty.

There is a specific kind of demoralisation that comes from effort that doesn’t convert. I knew the problem intimately. I just hadn’t paused long enough to question whether I was approaching it the right way.

Then one night, sitting in front of another empty search results page, I asked myself something obvious.

Why am I doing this manually?

The Decision to Build

I’m a creative. I make things. I solve problems by building something that didn’t exist before. And here I was, refreshing job boards every few hours like a habit I couldn’t break.

Something about that felt deeply wrong.

I’d been watching what Replit could do. Not just the demos. Real projects that real people shipped through conversation rather than months of planning and grinding. I decided to find out what it could do with my specific problem.

I opened Replit. I pulled up Agent 4. And I just talked to it.

Not in technical language. Not in specs. I talked to it the way I’d explain a problem to a friend over coffee. I told it I was tired of manually searching. I told it I wanted something that would do the scanning for me, understand what I was looking for, rank what it found, and deliver only the results worth my time every morning.

Then I watched it build.

That experience, watching an idea become a real working product through conversation, is something I don’t have adequate words for yet. It didn’t feel like using a tool. It felt like collaborating.

What Came Out of It

I called it Watson. After Holmes’s partner. The one who does the legwork so the detective can focus on what actually matters.

Here is what Watson does.

You sign up and tell it what you are looking for. Your target roles, your skills, your experience level, where you want to work and how. That is the only setup you do. Everything after that is Watson’s job.

It scans job boards, Telegram channels, and Twitter every single day. Every listing it finds gets scored against your profile on a scale of zero to a hundred, based on how well the role, the required skills, the seniority level, and the location actually match what you told it. Not keyword matching. Weighted relevance.

Every morning, a clean digest lands in your email. And if you connect Telegram, it lands there too. Only the jobs that scored well. No noise, no irrelevant listings, no missed opportunities because you were asleep when something good went up.

There is also a tracker where you move jobs through a pipeline as your search progresses. Saved, Applied, Interviewing, Offer. So nothing falls through the cracks during what is already a stressful period.

The Part Nobody Talks About

I want to be honest here because there is a version of this story that sounds like a magic trick, and it wasn’t.

Replit’s Agent 4 built a working foundation faster than I could have alone. The structure, the database, the interface, the email system. It handled the volume of work that usually takes weeks. That part was genuinely remarkable.

But I was still there. Catching things that weren’t quite right. Rethinking flows that felt off. Pushing the product toward something I’d actually want to use, not just something that technically worked.

What shifted for me wasn’t that I stopped being involved. It was that I stopped starting from nothing. Instead of staring at a blank canvas, I was reacting, refining, and shaping something that already had a headstart. For someone with a creative background, that difference is enormous. The hardest part of making anything is the beginning. When the beginning is already behind you, everything opens up.

What I’m Most Proud Of

A few things landed better than I expected.

The email verification flow. It’s a small detail but it matters. New users can’t access the app until they confirm their email. It sounds obvious but most early products skip it. I didn’t, and it meant that from day one the user base was real people who actually wanted to be there.

The match scoring. Running Watson on my own profile made the value obvious immediately. A 90 percent match looks completely different from a 40 percent match, and the difference is actually meaningful. It reflects what matters, not just which keywords appear most often.

The onboarding tour. When a new user lands on the dashboard for the first time, a short walkthrough appears once, showing them where everything is. Most products drop you into an empty screen and hope you figure it out. That’s a creative failure as much as a product one. First impressions shape everything.

What I Would Change

A few things I’d approach differently.

Sort the email deliverability out from the start. Watson currently sends from a shared testing domain. It works, but some emails land in spam. Getting a custom verified domain set up is not complicated. It just kept getting deprioritised and it shouldn’t have.

Build the real-time scraping earlier. The architecture supports it but most sources are running on structured data for now. Genuine live scraping from each source is where the real value compounds. It’s the most time-intensive piece to do well so I kept pushing it. That was the wrong call.

Why I’m Writing This

Not to tell you to quit your job and build your next idea with AI. That’s your decision.

I’m writing this because six months ago I was on the outside looking in, genuinely lost about where I was going. And what got me moving again wasn’t a plan. It was making something. Something specific, for a specific problem I was living inside.

Watson exists because I needed it. I use it now. And it works. I’m finding roles earlier, applying while the window is still open, and spending almost none of my energy on the parts of job hunting that are just friction.

If you are sitting on a problem you keep meaning to solve, there has never been a better moment to just start making something.

Build out your Idea with Replit right away!

Watson is an AI-powered job hunting assistant built on Replit. It scans 50+ job boards and channels, scores every listing against your profile, and delivers a daily digest to your email and Telegram.