Retraining after 15 years in tech

A friend said to me on a call the other day that he sees the writing on the wall. His consulting work is getting too easy because of Claude code.

Too easy? Why's that a problem?

Because it'll be easy for everyone. His work, coding plugins for a niche platform, won't be profitable forever.

This resonated with me because around 7 months ago I was thinking the same thing.

15 years in tech has given me a lot of experience. I've shipped products and grown them.

But my analytics expertise, that has been my consulting lifestyle business since 2020, what parts of that are still valuable?

Amplitude and Posthog have agents now. They're not perfect - and I really believe an experienced consultant can bring a lot on top. But people will sit in those tools and try them, instead of looking for a consultant. Companies will also need fewer data analysts - flooding the freelancer market and forcing prices down.

I sat down to think about what to do. Fortunately I had one SaaS app making enough to pay the rent.

I decided to do 2 things.

Firstly: focus on analyzing agents. Agent observability. That is the new wave, the in-demand skill. Adapt all my playbooks to this new domain.

I read hundreds of research papers. I built agents and analyzed them. I ran free analysis for friends with agents. I got good at optimizing agents.

Secondly: encode all of this in a product. Products are the ultimate interface for delivering expertize.

That product is twotail.ai, and it's my capstone. It does so much more than Langfuse, and as it stands it's much better than just running claude code over your logs!

Would love to hear how anyone else is thinking about retraining in this AI era.