Why I am Writing
I'm Troy Jarvis. I'm 26, I live in Berks County, Pennsylvania, and three weeks ago I started a job I'd been quietly training myself for without knowing I was training for it.
I was put on a Performance Improvement Plan at my last finance job that, as I learned when I quit, HR had no record of. I was working 60+ hours a week there. Six months before I left, my boss asked me to build predictive forecasting models. I had never written a line of Python. I figured it out because I had to. That was the beginning.
Six weeks after I left, the skills I'd quietly stacked up landed me an offer I couldn't turn down at a company I'd been circling for months. This blog is about what happened next.
Here's the fast version:
CubeSmart (2022 to 2024). First job out of college. FP&A in self-storage. Learned the finance craft. A food manufacturer in Reading, PA (2024 to January 2026). Financial Analyst in logistics, then Finance Manager. Where the coding started. Left on the kind of terms that become their own post eventually. EnerSys (January 2026 to March 2026). Short stop where I shipped enough automation fast enough to change their AI strategy. They restructured around the idea of "vibe coders" embedded in functional departments like Legal, HR, and Marketing, building with AI instead of waiting on IT. They offered money and equity to stay. Nakama (March 30, 2026 to now). Said no to EnerSys, yes to this. AI Implementation Manager. Building an AI-native platform for outsourced CFO services. That's the shape. The rest of this post is the why.
What I'm building at Nakama Nakama is a small, fully remote firm that provides outsourced CFO services to venture capital firms. My job, as the AI Implementation Manager, is to replace the collection of tools the team currently stitches together. Spreadsheets, shared inboxes, project trackers, dozens of manual handoffs. I'm building it as a single AI-native platform, from scratch, where every person on the team has a human-in-the-loop AI partner for the workflows they run every day.
It's the kind of high-trust, high-stakes finance work where getting it wrong is expensive. That's part of what makes it interesting. AI has to earn its place in workflows that have real money on the other side.
This is the project I was quietly preparing for during those six months of teaching myself to build. Twelve months ago I was writing forecasting models I didn't know how to write. Today I'm architecting a production platform that a real team is going to depend on. I still don't write most of the code. I direct AI to write it, review what it produces, argue with it when it's wrong. That's the Implementation Manager part. The work is still mine. The output is still mine. The craft is different.
If this works, and I think it will, it changes how the team scales. Which changes how they serve their clients. That's why I'm here. That's also why I'm writing about it.
A little about me I'm from Berks County. I have three dogs that run my life. Harper is a three year old Goldendoodle named after Bryce Harper. Rosie is a five year old Basset Hound who runs the house. Nebby is an 18 week old Lab mix puppy who is currently both the worst and the best thing about my day.
I'm a degenerate Phillies fan. I love the Eagles. If I'm home and a Philly team is playing anything, the TV is on. My first car was a '92 Miata, which explains a lot about me. I golf on weekends. I'm a normal guy who happens to be obsessed with what AI is about to do to the world, and to me, over the next five years.
Why I'm writing Six reasons, honestly:
My head is full. I have AI thoughts all day with nowhere to go. They don't belong in a Slack channel and they don't all belong in Obsidian alone. Writing them in public forces me to sharpen them.
I want to prove change can be the best thing that happens to you, even when it feels like getting pushed off a cliff. Fourteen months ago I was stressed, overworked, and convinced I was in trouble. Today I do the work I should have been doing my whole career. If you're reading this from a job you hate, quietly suspecting there's more out there: there is.
I want to show people what AI actually makes possible right now. Most people in my old world think AI is ChatGPT and autocomplete. That's about 2% of what it is. A non-coder with the right setup can ship production software at a pace that would have required a five person engineering team two years ago. I am proof of this, and I want more proof on the pile.
I want to become one of the most knowledgeable people in the world on AI. That's a large goal. It's the only goal worth having for me right now. Public writing is the shortest path. The feedback loop is brutal and fast, and the smartest people only show up when you put real work in front of them.
I want to document the failures, not just the wins. Most AI content online is inflated nonsense. I'll post my wins and the things I tried that didn't work. Harness changes that blew up. Tools I was wrong about. Projects I abandoned. The interesting stuff is almost always in the failures.
I'm trying to get rich. I'm not writing this for charity. I want to build a platform, get my name known in a space I'm obsessed with, and compound both over years. Writing publicly is the cheapest, highest leverage bet I can make on that outcome.
What this blog will be Short posts, long posts, whatever the week calls for. Some weeks a setup tour. Some weeks a take on a new tool. Some weeks a failure I want to work through out loud.
A few I already know are coming:
The setup I actually run. Claude Code, hooks, subagents, Obsidian knowledge graph. The harness I've built to do real work without writing most of the code myself. The review process I use to trust AI-written code in production. What happened at the food manufacturer. The full story, once it's more useful as a lesson than a vent. It's coming. You'll want this one. Updates from the Nakama build as it moves toward launch. Weekly-ish. Some weeks I'll miss. I'm not trying to be a content farm.
If you're reading this Three things I'm looking for from anyone who sticks around:
Finance or ops people quietly suspecting you could be building instead of reviewing. Yes. You can. I want to help prove it. People in AI who'll push back on what I get wrong. I'm learning in public. Tell me where I'm off. Anyone building something. I like connecting with builders. Reach out. Follow on X: @jarvistroy00. Connect on LinkedIn. Subscribe to this blog if bearblog lets you, or just check back.
I'm new at writing publicly. The prose will get tighter, the takes will get sharper, and the dogs are going to keep showing up in posts whether they're strictly relevant or not.
Thanks for reading. Back to prompting.
Troy
"The people who are crazy enough to think they can change the world are the ones who do." —Steve Jobs