The conference circuit is not expertise.
Three years running self-hosted models in production teaches you things no keynote does. Most people selling AI strategy have never owned the consequences of a deployment.
Essay in progress.
What we actually believe.
Five stances we'll defend in public. Each one comes from owning deployments, not attending keynotes. The full essays are in progress; the claims are not going to soften.
Headcount is a failure mode, not a plan.
If your AI roadmap needs twenty new hires to work, you bought the wrong stack. The point is to bend the cost curve down, not staff up to feed a system.
Essay in progress.
Pilots are where ambition goes to die.
Almost everyone has a proof of concept. Almost no one has it running across the business. The gap is execution discipline, and it's the only thing that matters.
Essay in progress.
Owning your models is leverage, not overhead.
Self-hosting keeps your institutional knowledge compounding inside the building instead of leaking into someone else's API logs.
Essay in progress.
Most "AI" is a prompt wrapped around a borrowed API.
The real work is orchestration, memory, and the feedback loop that makes a system improve on its own. That's where we live.
Essay in progress.
Disagree with one of these? Good.
A direct conversation, no deck, no discovery theater. You leave knowing what's worth building and what it takes.