The Day I Finally Believed
In 2025, you really can just do things.
I've been developing and managing software projects for over twenty years. I thought I'd seen plenty of tech come and go. I’ve gotten excited about all sorts of new stuff, but it was always in a certain frame: I knew what it meant to do software engineering. I was wrong.
Bari Weiss was on Honestly this week describing how they had to manually go through thousands of government PDFs to track foreign influence. As far as I could tell, they pulled the documents from fara.gov and manually looked at filing after filing, hunting for the things that mattered.
The site was exactly as awful as described. There was a huge list with thousands of scanned PDFs in an interface from 2004.
So I decided to DOGE it. I imagined what I wanted to see, which was a dashboard of all the dollar amounts rolled up per country.
About four hours later, I had faradash.com: a dashboard showing which countries spend the most on lobbyists or firms registered as foreign agents.
The process was awesome. I scraped 8,000+ government PDFs from 2023-2025, fed them to OpenAI to extract agent names, payments, and foreign principals, setup a dashboard view with Elixir and Phoenix, bought a domain, and deployed to production.
Total OpenAI cost? $3.10.
This whole thing was vibecoded in Cursor with Claude and I went from an empty directory to a live dashboard in well under five hours. And just three measly dollars to automatically parse years of government paperwork that would take humans forever.
The data is interesting, but for me the build process was even cooler than the results.
I know tons of developers way better than me. But as this unfolded in a couple of hours instead of a few days, I felt pretty superhuman. This was my conversion moment: the day I stopped seeing AI as a useful coding assistant and started seeing it as something that changes what's possible.
I'd done some vibecoding before. I thought it was helpful, but I still wanted to control every line of code. I’ve been treating it like a really good autocomplete.
This was different and new. We're at this weird inflection point where you can just think of something and build it. The same day. For three bucks.
The technical details matter because they demonstrate the absurdity. Eight thousand documents processed. Complex AI data extraction. Full stack web app. A good data schema and database. All in just a couple hours. If I hadn't slowed down the data download to respect the FARA website it all would have been faster. To be clear again, I wrote ZERO code.
If building is this easy, it changes who gets to solve which problems. All the engineering we've thought was economically valuable and hidden behind high-walled moats of skill and knowledge are under siege. Problems that used to require teams, budgets, or weeks of development now require curiosity and an afternoon. Someone with an idea and a few free hours can just solve it.
This little project is still just a toy. Production code is a whole different ballgame with different requirements, different scale, and different stakes. But man, this is changing insanely quickly.
So what else will we give up control of?
I've always been someone who wanted to understand. I wanted to know every detail and every abstraction layer. I liked knowing how the code worked, line by line. But this week it felt like I was flying.
We're about to find out what happens when the friction between having an idea and building it goes to zero.
I always believed AI would change how we code, but I had a different idea in mind, a simpler image of an assisted human. I was getting used to my new style of vibecoding, but all of a sudden it's a bigger picture. AI keeps giving us new tools.
Next week someone will build a new application that will blow us all away. A scientist will discover a new technique to solve a disease. A biochemist will find new protein folding mechanisms. A teacher will develop a new tutor that gets the entire class to understand algebra. Now it's about what we choose to manifest into existence with a whisper and a few keystrokes.
Skill and knowledge aren’t the same gatekeepers they once were. Agency and ideas are more important when anyone gets to solve problems. It’s going to take us all awhile to catch up.