7 Lessons from Launching My First Product

I quit my job, built an AI product, and got a reality check

In April, I left Decentralised.co to build software products with AI tools. I’d already started doing this at my previous role, where I built two tools that were each successful in their own ways—one reached 100k users, while the other generated significant inbound interest for the firm.

Next, I wanted to venture out solo and build without a safety net.

My first experiment was Dweli, a personal website builder that was different in two ways:

1. It automatically aggregated user content from across platforms like LinkedIn, Twitter, and Substack, keeping everything constantly in sync.

2. It deployed the content in AI-friendly formats like LLMs.txt, making individuals discoverable by AI agents.

Two months, thousands of lines of code, and countless conversations with ChatGPT and Cursor later, I finally launched Dweli two weeks back. And... it didn’t feel great. Most things didn’t go as planned. I wasn't proud of what I'd shipped. Sadly, midway through the journey, I already knew it would end this way.

I'm writing this post to share what I learned building Dweli. It's a note to myself—a collection of reminders as I continue on my journey. In an age where building software has never been easier, there's a chance you're building something too.

Maybe this post will help you avoid similar pitfalls.

Start with Why

My initial motivation seemed genuine enough—I'd fallen in love with personal websites after discovering Onur Şuyalçınkaya's onur.dev. It captured his work and personality in ways no algorithmic feed could. Once I built my own site and realized LLMs could access it freely, unlike my locked-down social profiles. This felt like an opportunity: in a world where AI agents become our primary search interface, everyone would need their own web presence.

I also had a two-month San Francisco trip planned for June. I wanted to experience the "builder energy" I'd been promised on X. Having a product would give me an in—credentials, a talking point, an answer to "what are you working on?" at meetups.

This trip deadline warped everything. What started as a tool I believed in morphed into a checkbox to tick. By the time I landed in SF, I wasn't building Dweli because I believed in the vision. I was building it just to have something to talk about. A prop for networking conversations. Nothing more.

When I finally launched, I felt nothing. No excitement, pride, or anticipation. A half-hearted launch effort got me 20 signups, half of them friends, none paying. Even worse, when I talked about Dweli at those SF meetups, there was no conviction in my voice. People could tell. What should have been a conversation starter became a conversation killer.

Lesson 1: You need absolute clarity on why you're building and who it's for. Is it to create something useful? To generate revenue? To raise venture capital? Building just to have an impressive answer to "what are you doing?" is NOT good enough. Your motivations will show up in the product and in how you talk about it.

The Sunk Cost Trap

Around the week four mark is when my conviction around Dweli evaporated.

But I'd already spent a month on it. Written thousands of lines of code. Learned new technologies. How could I just... stop?

So I kept going. Each day made it harder to quit. I'd invested too much to walk away. The thought of having nothing to show for all that work felt unbearable.

This was textbook sunk cost bias. I knew about it intellectually. But when it's your time and your code, rationality goes out the window.

By refusing to "waste" four weeks of work, I wasted nine. By insisting on having something to show, I ended up with something I wasn't proud to show.

Lesson 2: Cut your losses fast. When you realize a project isn't working, the time you've already spent is gone. You can't get it back by spending more.

Be Wary of AI

Before starting Dweli, I did what felt natural—I shared my vision with ChatGPT and asked for help. We settled on a tech stack with tools I'd never used (Cloudflare, Redis, Docker, Fly.io, Loki). Then I asked for an implementation plan. I wanted to ship fast. My SF trip was five weeks out, so I requested a day-by-day roadmap that would deliver an MVP in three weeks. Shoot for the stars and land on the moon, right?

Here's the thing about AI: it gives you what you want but won't push back when you're being delusional. Building what I envisioned in three weeks was frankly insane—especially for someone learning most of these technologies from scratch. Yet ChatGPT handed me a plan that made me assume it was doable.

Lesson 3: don't mindlessly follow what an AI tells you. Recognize and account for its blind spots. Verify and question everything that matters. AI is an incredible tool, but it won't save you from your own delusions. It'll happily help you execute them.

It takes Time

Reality hit fast once I started executing on ChatGPT’s plan. What should've been done by Day 3 slipped to Day 4. I took one Sunday off, and suddenly I was three days behind. At one point, I abandoned the plan altogether and just did whatever I thought was the most urgent.

And after a few weeks, even when I thought I was close to done, I wasn't. I'd heard countless times that the last 10% takes 90% of the time. It’s true. I told myself (and others) that I’d launch “this weekend” three weeks in a row without launching. If you have high standards for a product, the details just take time. There is no way around it.

Dweli eventually took me over two months to build.

Lesson 4: it takes time. Always budget for longer than you expect.

Get Feedback Early

About halfway through building Dweli, I showed an early version to a friend.

Watching him use the product was incredibly valuable. The onboarding was painfully slow. Certain features that seemed obvious to me left him confused. He'd click around looking for things I'd buried in submenus. These were issues I'd never noticed because I was too close to the product. His observations helped me fix real problems that would've plagued the final launch.

When I asked him what he thought about the product, he said it was cool. But would he actually pay for it? No.

That “no” stung. It made me scared of getting more feedback. I convinced myself only the "final" version would be good enough to show others. I thought if I just polished it more, added more features, made it more complete, then people would see the value and want to pay.

By hiding Dweli until it was "ready," I missed countless opportunities to understand why people wouldn't pay. I built in isolation, guided only by my assumptions. By the time I launched, the fundamental issue my friend had identified—that people wouldn't pay for it—was still there, just dressed up in a more polished UI.

Lesson 5: Early feedback is a gift, especially the painful kind. When someone says they wouldn't pay for your product, it's a signal to dig deeper and show it to more people.

Consumer Surplus

There's a concept in Economics called consumer surplus—the difference between what someone would be willing to pay for something versus what they actually have to pay. People pay for products when the value they get exceeds the cost. That value usually comes down to saving significant time, money, or solving a painful problem they can't easily solve themselves.

Dweli didn’t deliver enough Consumer Surplus.

The hard truth was that if someone wanted a personal website, they could already build one. There were hundreds of website builders out there—from Squarespace to WordPress to GitHub Pages. If they cared enough about having a web presence, they'd have created one already. Dweli's automatic content aggregation was neat, but it wasn't solving a $10/month problem.

The AI discoverability angle—my big bet—wasn't a burning problem yet. I was building for a future where LLMs replaced Google as our primary search interface. Even if I was right about that future, people wouldn't pay for it today.

Lesson 6: Build things people will pay for today, not things they might need tomorrow. Something “cool” doesn’t guarantee revenue if it’s not a burning need.

Learn from Others' Mistakes

The embarrassing part is that none of these mistakes were new to me. I'd read all the warnings—don't build without talking to users, validate willingness to pay early, beware of building for impressiveness over usefulness. I knew the patterns. I'd read the Paul Graham essays.

I just thought I was different.

When I read about failed aggregator startups, I thought, "Yeah, but they didn't have AI." When builders warned against creating solutions for future problems, I told myself, "But AI adoption is happening faster." Every cautionary tale came with an asterisk: but my situation is different.

It wasn't.

Lesson 7: When you catch yourself thinking "that won't apply to me," pause. It probably does. The wisdom of others isn't outdated just because your tools are new.

Moving Forward

I'm being hard on myself in this post. That's intentional.

But despite everything, I don't regret building Dweli. The lessons were painful, but hard won. I also learnt a bunch of new, industry-standard technologies that will come in handy in the future.

As for Dweli itself? The core idea isn't flawed. The end state is an AI-enabled version of LinkedIn. But building that sort of social network requires venture backing and a 10-year commitment. Some other teams are attempting it. That's not what I set out to do.

We're in the golden age of building. AI has given us superpowers that developers five years ago couldn't have imagined. But with infinite leverage comes the need for better judgment. Build fast, yes—but build the right things for the right reasons.

I'm taking a break from building to reflect on what I actually want to create. When I start again, it'll be with an idea so compelling I can't not build it. Only this time, hoping to avoid the mistakes I made with Dweli.

If you're building something, I hope you learn from my mistakes. You'll probably make them anyway—we all do. But maybe you'll catch yourself sooner. Maybe you'll cut your losses faster. Maybe you'll ship something you're actually proud of.

Follow my journey on X or subscribe to my Substack where I write about building in the AI age.