I found Michael Matloka’s path really interesting.

He joined PostHog as employee number five straight out of high school and later started and led PostHog AI.

His main lesson from building agents: the model is rarely the hardest part. Context, tools, permissions and reliability are what determine whether an agent can actually do useful work.

That explains why he is now joining @viktor_com, which is tackling exactly these problems across more than 3,200 tools.

I’m curious to see what he builds there.