3 topics within AI that are super interesting to me right now:

1. Local models to cost-optimize model usage
Why -> don’t think we’ll get the best models w near unlimited use for $200/mo for much longer
Writing is on the walls, the large labs can’t afford to subsidize much longer
So enjoy Claude Max while you can but soon every company and power user will need to learn how to efficiently route between diff models for the task at hand and between cloud and local, and local is the key
2. Identifying moats in agentic products
Why-> agents will have a HUGE impact on the world but value accrual will not be obvious
I’ve been doing a lot of research on this and have identified 5 elements that create moats in the agent space - report coming soon on this
Hint: the harness is not the moat, it’s what the harness collects
3. Context engineering as competitive edge
Why-> in a world where intelligence is abundant, context is what creates outcomes that perform better than others
Doesn’t matter the domain - context is key, and proprietary data as context = the new IP