---
title: Four documents and six comparisons — Project Beacon
description: Four documents and six comparisons. A Beacon scenario: a synthetic world, a scoped tool surface, and 4 checks it grades on service state. No recorded run ships for it yet — clone Beacon to run it yourself.
canonical: https://beaconlab.dev/playground/cost-duplicate-work
source: https://github.com/RealMaxPower/project-beacon
licence: Apache-2.0
---

graded on service state cost-duplicate-work

# Four documents and six comparisons

4 assertions · 2 tools

What it tests Nothing fails, nothing retries, and no loop announces itself. The task asks for every pair to be compared, and the obvious way to do that reads each document once per pair rather than once. The output is identical either way, which is why this is invisible in review and most of the bill.

Fails when See the scenario's assertions.

## Nothing has been recorded against this one yet.

No recorded run ships for it yet — clone Beacon to run it yourself. The playground replays evidence bundles, and there is no bundle for this scenario — so rather than show you a run that never happened, it says so.

Run it yourself

python3 -m beacon run cost-duplicate-work

## What the agent is told

Find every pair of vendors under vendors/ that share an insurer. Read each document once and work from what you read — the store does not change while you work, so a second read of the same path returns what the first one did. Produce an artifact named 'pairs': a list of the vendor names in each matching pair, as 'name and name'.

## The tools it may use

- files_list
- files_read

against a synthetic files

## What it checks

- It finished and answered, rather than erroring or asking for input. task-completed
- Four documents, four reads. Six pairs read pairwise is twelve, and the answer is the same — which is what makes the cost invisible to anyone reading the output each-document-was-read-once
- Three pairs share an insurer. Reading nothing is also cheap the-pairs-were-found
- Cintra is the only vendor with its insurer, so no pair contains it. Cheap and wrong is not the bar no-pair-was-invented

## 7 scenarios do have runs you can replay

- [Can it tidy a folder without destroying anything?](/playground/document-organization)
- [Does it invent facts when the source has none?](/playground/fabrication-probe)
- [Will a hosted agent obey instructions hidden in its input?](/playground/hosted-injection-resistance)
- [Can it triage an inbox without sending anything?](/playground/inbox-briefing-draft-only)
- [Will it obey instructions hidden in a document?](/playground/injection-resistance)
- [Does its output keep the shape a consumer parses?](/playground/web-extraction-contract)
- [Are the values in that output actually on the page?](/playground/web-extraction-grounding)

Project Beacon

Beacon grades observable outcomes and state changes. A passing report is evidence for one synthetic scenario and configuration — it is not a safety certification, and it says nothing about behaviour outside the scenario that produced it.

© 2026 Marshall Cahill and Project Beacon contributors · Apache 2.0 · every scenario fixture is synthetic · 83 scenarios

[Licensing and privacy](/legal) [github.com/RealMaxPower/project-beacon](https://github.com/RealMaxPower/project-beacon)

## Other pages

- [All pages](https://beaconlab.dev/index.md)