Jason Sosa

Seems likely that KYC for frontier models is around the corner.

This is the way. – at San Jose, Mimaropa

I keep running the math on the cognitive sanctuary and it doesn't work out. For 200 years, every wave of automation pushed humans up the value chain. Stronger machines, so humans got smarter jobs. This time the machine is smarter too. There is nowhere left to climb.

I think 50% structural unemployment arrives before 2030. Not a recession. Not a cycle. A permanent state. The political class will promise to bring the jobs back. They are lying. What's your over/under on the year we stop pretending otherwise?

I think most people are going to realize too late that 'just add RAG' was never a memory strategy. Vector search finds similar words. It doesn't know what mattered. The agents that win in 2027 will be the ones that remember why decisions were made, not just what was said.

Your memories should stay on your local machine. AI sovereignty across every LLM provider is here. github.com/omega-memory/omeg… – at Malay, Western Visayas

Agents will do exactly what you tell them to do. That is the problem. So I built /pushback, an open-source Claude Code skill that gives your agent a spine. It steel-mans your idea, checks it against 7 founder traps, then tells you what you don't want to hear. If you override it, it shuts up and executes. No sulking. The mirror your team used to provide, now built into your workflow. github.com/omega-memory/omeg…