Possible is Everything: How AI Brought Me Back to Building

Building a production system in a weekend, and rediscovering why I loved engineering in the first place

From Code to Capital

After 27 years in software engineering—BlackBerry, Amazon, various startups—I made a deliberate choice to step away from building. The progression from engineer to manager to director felt inevitable, but somewhere along the way, the work stopped being about creation and became about coordination.

So I pivoted to full-time investing. The transition made sense: analyze companies instead of building them, read 10-Ks instead of writing code. The early wins validated the decision—Palantir at $9, Google at $90, and several other multi-baggers.

But I kept a mental list of tools I wished existed. Portfolio trackers that actually understood cost basis. Earnings calendars that filtered intelligently. Market analysis dashboards that didn't require subscriptions to expensive terminals.

The Weekend Experiment

While doing due diligence on AI development tools, I tried Kiro—not as an investor, but as someone who missed building. I had one weekend and one idea: a market intelligence platform that could track 200+ stocks, generate AI commentary, and send daily emails.

Here's what happened:

Saturday: Set up DynamoDB tables, built a Lambda function to fetch data from Yahoo Finance for 214 tickers, and deployed it. Then added Claude (via AWS Bedrock) to generate market analysis. Iterated on the prompt until the output was institutional-grade.

Sunday: Implemented Monte Carlo simulations for 90-day price projections using pure Python (no external libraries). Built the email delivery system with SES. Created the HTML templates. Set up EventBridge to run everything twice daily.

By Sunday evening, the system was live. Real data, real analysis, real emails going out to subscribers.

What Made This Possible

I still wrote code. But the friction was gone. When I said "create a DynamoDB table," it knew I needed composite keys and on-demand billing. When I said "deploy this Lambda," it handled packaging, IAM roles, and configuration. When something broke, I could describe the error and get actionable fixes.

The technical decisions were still mine—serverless architecture, which indicators to calculate, how to structure the prompts. But the thousand small decisions that usually drain your energy? Automated.

The Economics

The entire system costs $0.50/month for 100 users:

• Lambda: $0 (free tier)
• DynamoDB: $0.09
• Bedrock: $0.14
• SES: $0.20
• Everything else: $0.07

This isn't a prototype. It's a production system with 214 tickers, 13,500+ historical records, 20+ technical indicators, AI analysis twice daily, and 99.9% uptime.

What Changed

I'm building again. Not because I have to, but because I want to. That list of ideas I've been carrying? I'm working through it. Portfolio tracker, earnings calendar, sector rotation dashboard—all in progress.

The gap between "I wish this existed" and "here's the URL" has collapsed. Not because AI writes perfect code (it doesn't), but because it removes the friction that makes starting feel impossible.

I left engineering because the process became tedious. I'm back because building became fun again.

Possible is Everything

If you have an idea you've been sitting on—a tool you wish existed, a dashboard you've sketched, a system you know would be useful—the barrier isn't skill or time or money anymore.

It's just starting.

I built a production system in a weekend after years away from code. If that's possible, what else is?