Back to Blogging: AI is Moving Fast

It has been a while. Life got busy — new roles, a country move, and the general entropy that accumulates when you stop doing something deliberately. But I’ve been meaning to come back to writing for a while, and now feels like the right time.

Why now?

Honestly, because AI development is moving so fast that I feel compelled to think out loud about it. The field has changed more in the past three years than in the previous decade. We went from GPT-3 being a curiosity to large language models being embedded in almost every software tool I use daily. Reasoning models, multimodal systems, coding assistants, agents — it’s a lot to keep track of, and writing is how I process things.

I work with data and statistics for a living, so I have a particular lens on this: what does it actually mean for quantitative work? What parts of the job change, what parts don’t, and where are the genuinely hard open problems?

What to expect

I’m not planning to turn this into a news feed or a hype machine. I’d rather write less frequently but with more substance — working through an idea, sharing something I built, or documenting something I learned that took me longer than it should have.

Topics I’m thinking about:

• LLM internals — how these models actually work, from attention to RLHF to inference-time scaling
• AI in science — what tools like AlphaFold and the wave of AI-assisted research mean for fields like genetics
• Practical tooling — what’s actually useful day-to-day versus what’s a demo
• Statistics meets ML — the old tensions and the new syntheses

Getting started again

The blog itself got a makeover too — migrated from Hugo to Astro, cleaner design, dark mode that actually works. Small things, but they make it more pleasant to write.

If you want to follow along, there’s a subscribe link in the nav. Or find me on Twitter/X at @gumgumeo.

Let’s see where this goes.