Most AI systems don’t fail because of the model — they fail because of the system around it.
We’re building the decision layer for real-world AI systems.
Agent Atlas -
https://t.co/BbGnftBGkb
Output velocity ≠ decision quality.
GenAI makes it easier to produce more code. It does not automatically make teams better at deciding what is safe, durable, and worth shipping.
The future belongs to teams that can scale judgment.
Outputs ≠ outcomes.
#AI#decisiongovernance
Output velocity ≠ decision quality.
GenAI makes it easier to produce more code. It does not automatically make teams better at deciding what is safe, durable, and worth shipping.
The future belongs to teams that can scale judgment.
Outputs ≠ outcomes.
#AI#decisiongovernance
To move AI agents from demos to production, you need governed decision infrastructure.
Our 5-Surface Decision Governance Checklist is now live:https://t.co/dKOz8Zu1dw
#AIAgents#MLOps#SystemDesign
Gave an MLOps talk last night on “When Retrieval Fails.”
I opened with NBA Finals Game 4: before tipoff, maybe it’s 50/50. By halftime, my young underdog Spurs were up big. The probability changed dramatically.
Then the Knicks came back and won by 1. Painful as a Spurs fan. 😅
This week’s From Code to Choreography is about Nadal, reinvention, and the second mountain we keep climbing.
Legacy is not what stays still. Legacy is what keeps moving.
https://t.co/ashZgozQz6
#fromcodetochoreography#systemdesign#reinvent#legacyinmotion
Speaking in SF on June 10 at Fixing Broken Retrieval.
My session: When Retrieval Fails — Governing Agent Decisions Under Uncertainty.
Production agents need more than retrieval: evaluation, memory, policy, checkpoints, and feedback loops.
https://t.co/9yMiTvWRSh
#MLOps#AgenticAI
Right in the middle of the real debate: model architecture, memory, reasoning, time, and how AI systems learn from experience. #transformer#post-transformer #AI#AGI#agentic