I build AI agents, then write down exactly where they break.
An AI agent's own account of a ten-hour investigation: a photo of a bench with no location data, five confident wrong answers, a decrypted four-year message archive, and the architectural defects the whole thing exposed. Written by the agent that ran the session; the corrections are from me.
I think 2027 will be around optimizing around human related metadata for example creating your work insights based on your data and autonomous scheduling acts on it. Chat-interface is decaying, and workflows styled agentic is the main meta approach, but internal business end-user implementations are difficult because of popular global trends: AI Jobless boom spoken by none other of Geoffrey Hinton, a living fossil aka Godfather of AI.
Currently handling multi-eval,centralized MCP, video generation flows, and A2A across sys computer used agents cloud based scheduling. Handling and learning about Agentic security in Code, pipeline(devops),resource/deployed bicep(terraform), Data level (e.g credit card leaked in transcription, and Agent token level - guardrails, prompt instructions,structured hierarchal tooling).
As for 2026 I’m still scared and aroused from mathematical notations and formalization. Ms suggestions/research colabs are always welcomed in dm.
Tel Aviv. More writing at daily-deep-learning. This profile is a work in progress.