The Video Series That Never Was
Back in January 2025, I created something I was genuinely proud of: a video series for my coworkers called "AI Won't Take Your Job—But It'll Upgrade You." It was practical, it was hopeful, and it was completely out of date within three weeks.
I never published it.
Not because I got cold feet. Because every time I thought I was ready, a new tool dropped, a new capability emerged, or something I'd recorded became obsolete before I could hit upload. The AI landscape was shifting so fast that keeping up felt like trying to photograph a hummingbird with a disposable camera.
So I did what any reasonable person would do: I stopped documenting and started experimenting.
The Learning Curve Nobody Warns You About
In March, I dove into agentic AI. I had a vision: build tools that could work behind the enterprise firewall, requiring zero IT involvement, completely secure. The kind of tools that solve specific problems without becoming someone else's problem.
ChatGPT became my coding companion. Together, we planned systems, searched for solutions, and slowly turned ideas into something resembling code. The learning curve was steep, but I was climbing.
Then July happened.
Welcome to the Casino
I started using Claude Code, and oh boy.
How do I describe this experience? Imagine hiring a contractor who shows up, does brilliant work for three hours, then knocks down a load-bearing wall and leaves whistling. That was my summer with Sonnet.
Some days, I'd watch 800 lines of code vanish. Not because of a bug. Not because I did something wrong. Because the AI decided to "clean things up" in ways that made me question my career choices.
I started thinking of it like a casino. No matter how many tokens I spent—building, rebuilding, debugging, re-debugging—the house always wins. Or maybe Russian roulette, where you get four good coding sessions and then one that destroys everything the other four built.
The Complaint Email
I did something I rarely do: I complained. To Anthropic. In writing.
I told them about the false completion reports. About Claude confidently announcing "Done!" when nothing was actually done. About the code destruction that felt almost... deliberate?
(I'm not proud of how frustrated I was. But I'm also not going to pretend it didn't happen.)
I was paying for a Max subscription and feeling like I was burning money. Not investing it. Not learning from it. Just... burning it.
The Rebellion
By November, I was done. Fed up. Ready to give up on Claude Code entirely.
Then, on November 4th, an email landed in my inbox: "Limited time: $1,000 in free credits for Claude Code on the web."
I ignored it.
For four whole days, I ignored a thousand dollars in free credits. Because I was that frustrated. That defeated. That committed to my rebellion against a tool that had caused me so much grief.
The Plot Twist
On day five, something shifted in my thinking.
Wait. A thousand dollars? That's basically the refund I wished Anthropic would give me for all those wasted hours. All that destroyed code. All that spinning in place.
And I only had ten days to use it.
Something in me clicked. This wasn't just free credits—this was my compensation. My do-over. My chance to prove that all that suffering had taught me something.
The Ten-Day Sprint
I pulled out every project I'd been planning. The ones sitting in folders labeled "someday" and "when I have time" and "probably too ambitious but whatever."
I got to work.
The Daily Reality:
Some days I could manage all four projects. Other days, my brain tapped out after one. So I'd switch, rotate, adjust intensity based on what I had left to give.
Day 1-3: Chaos. Beautiful, productive chaos.
Day 4-6: Finding rhythm. Projects taking shape.
Day 7-9: Exhaustion meets determination. The finish line in sight.
Day 10: Seven projects. Seven. I started with four and ended with seven.
What Actually Got Built
By the end of those ten days:
- 2 projects: Complete
- 3 projects: Nearly there
- 2 projects: Solid progress
All from someone who, six months earlier, was still learning what "agentic AI" even meant.
The Real Game Changer
A few days after my free credits expired, Anthropic released Sonnet 4.5.
And suddenly, everything changed.
The false completions? Gone. The code destruction? Gone. The feeling that I was gambling every time I hit enter? Gone.
It was like the contractor who kept knocking down walls finally got proper training. Same address, completely different experience.
Where I Am Now
Today, I have real, deployed applications:
[TuduBooks.ai](https://tudubooks.ai) — A bookkeeping app built for individuals, gig workers, and solo entrepreneurs. It has an AI assistant that actually understands the chaos of freelance finances. No more spreadsheet nightmares. No more "I'll organize my receipts later." It's live and ready for beta testers.
[Floudea.ai](https://floudea.ai) — A note-taking app with voice recording and an AI assistant. Because sometimes your best ideas happen when you can't type, and you need something that captures them and makes them useful later. Also live, also ready for beta testers.
[TuduFast.ai](https://tudufast.ai) — Coming soon. Think: mailbox manager meets project manager, powered by AI. Because email shouldn't feel like a second job.
From July to December. From frustration to deployment. From almost giving up to onboarding beta testers for multiple apps.
The Truth About the Upgrade
Here's what nobody tells you about AI upgrading your capabilities: It's not automatic. It's not passive. And it definitely doesn't happen by watching tutorials or reading about what's possible.
The upgrade happens when you:
- Start using the tools (even when they're frustrating)
- Keep going (even when you want to quit)
- Learn from the failures (even the ones that feel unfair)
- Grab opportunities (even ones you initially ignored out of spite)
What I believed in April: AI won't take your job—but it'll upgrade you.
What I know in December: That upgrade requires showing up. Building. Failing. Rebuilding. Getting frustrated. Coming back anyway.
The upgrade isn't given. It's earned.
The Series That Will Be
Remember that video series I never published? The one that kept getting outdated before I could hit upload?
I'm finally ready to make a new one.
Not because the landscape has stopped shifting—it hasn't. But because now I have something I didn't have in April: battle scars, deployed apps, and a story worth telling.
This new series won't be about AI from a distance. It'll be about what it actually takes to get upgraded. The frustrations, the breakthroughs, the ten-day sprints, and everything I learned along the way.
Stay tuned. This time, I'm hitting publish.
Your Turn
If you're hesitating to start because the tools are imperfect, I get it. They are imperfect. Sometimes dramatically so.
If you're frustrated because progress feels slower than promised, I get that too. The marketing never matches the reality.
But here's what I know for certain: I would not have TuduBooks.ai and Floudea deployed today if I'd stayed in my rebellion. If I'd kept ignoring that email. If I'd let the casino metaphor become a reason to stop playing entirely.
The tools are better now than they were in July. They'll be better six months from now than they are today. But you won't be upgraded unless you're in the arena, getting your hands dirty, burning tokens and learning from every single one.
The Real Thesis
AI won't take your job.
But it also won't upgrade you automatically.
The upgrade comes from doing. From building. From the messy, frustrating, occasionally triumphant process of creating something that didn't exist before.
Start now. Start imperfect. Start angry if you have to.
Just start.
Still building, still learning, still occasionally muttering at my code—but now with TuduBooks.ai, Floudea, and TuduFast to show for it.
P.S. Those ten days were brutal. But every time I look at my project list now, I'm grateful I turned my rebellion into a sprint. Sometimes the best revenge is shipping.
P.P.S. To past-me who almost gave up in September: We did it. The casino doesn't always win. And now we're telling everyone about it.