Categories
Articles

Why, as a finance professional, I use Python instead of Excel



Finance is a lot of process. Specific data, specific steps, specific decision points. And most of that process is rule-based and logic-driven — which means it’s automatable.

Over the years, the share of my analytical work done in Excel has steadily shrunk — in favour of code editors. Today, less than 5% of my analytical work happens in Excel. The other 95%+ runs in Python and purpose-built analytical tools.

What Python gives me that Excel can’t:
– ⚡ Speed — fetch & process data in seconds, not spreadsheets
– ✅ Reliability — same logic runs the same way, every time
– 🧠 Complexity — models scale cleanly beyond a grid of cells
– 🔄 Flexibility — adjust & update in code, not a rebuild

The workflow: Fetch → Process → Structure → Act — each step automatable.

Excel is still great at what it’s great at. But for the process of finance, Python is the better tool. I’m a realist: I use both. The more complex and repeatable the work, the more I reach for code.

Categories
Articles

I hired a (personal) AI employee.

I hired a (personal) AI employee. It works for me in the background, and it’s now a permanent part of my “staff”.

For over a month I’ve run Hermes Agent — an autonomous AI that lives on my hardware, has its own memory, and does things. Not a chatbot. A worker.

What it actually does for me:

🎬 1,500+ videos & articles → downloaded, transcribed, summarized, and filed into a searchable learning library. I just drop a link; it does the rest. And it structures the whole process — turning a chaotic stream of random finds into a clear curriculum and learning plan I can actually work through. A very different outcome from what the social media algorithm will feed you.

🔎 A research assistant that knows my standards. I throw topics at it and it scours various sources — with references, context, and analysis. It’s learned what matters to me in research, so the output comes back the way I’d want it, not generic.

📚 5 Anki (language learning) collections (66,901 media files) synced to my phone via a self-hosted server it deployed — and debugged when the SQLite schema broke.

🔐 My VPN — added a TCP fallback, managed the firewall, and diagnosed a carrier-NAT issue that was blocking me on mobile.

🧠 A private digital coach — its own persona, memory, and Telegram chat, checking in on my career and health. Runs on local models on my own hardware. My conversations never leave my infrastructure.

📝 A notes repository it maintains and cross-references automatically.

What I learned about managing it:

  1. Autonomy is the unlock. I brief it, it executes end-to-end, and reports back. I don’t babysit it.
  2. It fails — and self-corrects. Crashed containers, locked databases, retries. Resilience beats perfection.
  3. Memory compounds. It remembers my personal targets and goals, my health habits, my taste in music. Every interaction gets smarter.
  4. Privacy is a feature. My most personal conversations — with my coach — run on local models. No cloud, no training on my data.
  5. You still need judgment. It’s a tool, not a replacement for thinking. I review the important stuff.

The honest take: I’m a realist — I know what technology can and can’t do, and I don’t chase hype. But AI is maturing, and the potential is very significant. After weeks of real use, Hermes Agent has earned its place as permanent infrastructure, not a novelty. It’s the difference between having an assistant and having a staff.

Curious about hiring your own agent — and your own private coach — on your own hardware? Happy to share what I’ve learned.