sammatuba

ESSAY · PROJECT 0003

An agent is a loop, not a model

Dataset the ARENA 3.0 curriculum (TARA vault submodule, pinned 81533b2) + the TARA Learning Plan v2 · Charts 1 · Written SEP 2026

Entry 9 of the ARENA guide. Part 3.4, LLM agents. Route position: program week 8, tagged [core] (§1–3 core, §4 elicitation [light]; ARENA budgets the whole set "up to 2 days"). Reading thread: MacDiarmid et al., natural emergent misalignment and reward hacking.

The mental model

An agent is a model with tools and a loop. The model emits a tool call instead of a

final answer; the harness executes the call and hands back the observation; the model

goes again. Everything ARENA builds this week is that loop wearing different clothes:

an arithmetic agent whose tool is a calculator, then a Wikipedia-game agent whose

tools are link navigation, then chat history management so the loop survives more

than a handful of turns.

Two ideas deserve the week's attention. First, the loop is where control lives: the

harness decides which tools exist, what permissions they carry, and when the episode

ends, which is why agent evaluations can be safety-relevant at all. Second, agents

fail in loop-shaped ways: they revisit, they loop on the same failed call, they

burn tokens. The scaffolding choices (history truncation, max iterations, tool

design) are the difference between an eval and a fortune-burning session.

The picture

The exercises

Quoted from the pinned 3.4 page, with ARENA's own budgets:

What it unlocks

The capstone's episodes (plotted for week 10 as smoke tests) are exactly this loop

around a market environment. The route's W44 gate assumes you can read an agent

scaffold and name its permissions, and the control chapter next week assumes the

loop exists to be attacked.

Budget and receipts

ARENA's whole-set budget is "up to 2 days"; the listed exercises sum to roughly

1.5–2 hours. The route takes §1–3 as core and defers §4 elicitation to [light].

[TODO: receipts — hours vs budget, WikiGame agent behavior, what broke (fill from the TARA vault when the week closes)]

PREVIOUS Inspect turns an eval into a record you can argue with · NEXT Control assumes the model is trying to fool you · GUIDE HOME