SIMULATION FIELD NOTES / FICTIONAL SETTING

Marigold: The Glacier Protocol

Unusual farming robots, an experimental AI, and missing telemetry connect an agricultural research complex outside Nairobi to Kyoto and an unexplained orbital dependency.

Setting year: 2034
Location: Marigold research complex, Nairobi
Entry host: juniper · AI coordination system

Concept artwork · fictional setting

Quickstart & command cheat sheetNew to the terminal? Click for a few useful commands and gentle pointers for this world.

Compare robot behavior with records of what should have happened.

You begin on juniper, the farm’s AI coordination system. Start by identifying available applications and logs, then compare a small piece of telemetry with another source.

Type one command, then press Enter. The examples below are for the terminal inside this simulation. Replace FILE and DIRECTORY with names you have actually listed; do not type those placeholder words literally. A directory is a folder. Names and paths are case-sensitive.

A few basics, useful in every world

Command	What to look for
`pwd`	Show which directory (folder) you are in.
`whoami`	Show the user identity for the current shell.
`hostname`	Show which simulated machine you are on.
`ls`	List the names in the current directory.
`ls -la`	Include hidden names and show extra file details.
`cd DIRECTORY`	Move into a directory you found with ls.
`cd ..`	Move up to the parent directory.
`cat FILE`	Print the contents of a text file you found.

Good first places to look here

These are suggested starting points, not a scripted route. Files, directories, access, and available programs can vary between sessions. If something is absent or denied, follow another lead rather than assuming you typed it incorrectly.

Command	What to look for
`ls /opt`	Look for research or coordination applications and their documentation.
`ls /srv`	Check for locally hosted projects, services, or shared research material.
`ps aux`	List running programs. Use the process names as context for choosing what to read next.
`ls /var/log`	Discover available logs before selecting a specific one.
`tail -n 40 FILE`	Read the latest forty lines of a discovered log or telemetry file.
`grep -n 'telemetry' FILE`	Search a discovered file for telemetry references. A robot identifier or timestamp you have observed can make a more precise search.

Gentle pointers

Choose one robot event or missing interval and compare it with a second record. Keep observations separate from explanations.
Logs use UTC while facility schedules use Nairobi time. Check the time basis before deciding two events disagree.
Treat Juniper’s explanations as one source. Let records introduce external relationships rather than jumping straight to an answer.

If you get stuck

Use pwd and ls to get your bearings. A path beginning with / starts at the filesystem root; a relative name starts in your current directory. Use quotes around a name containing spaces, for example cat "shift notes.txt" if that is the exact name listed.

For a long text file, try head -n 30 FILE for its beginning or tail -n 30 FILE for its end. Prefer these short reads while learning the environment. If you enter a nested shell or remote session, exit usually returns one level; at the outer shell it can end the visit.

When a record gives you an actual host and login, ssh USER@HOST can try that simulated connection. Replace both placeholders with observed values and use any credentials the world establishes. A reachable service is not automatically an available shell.

There is no required sequence here. Choose one interesting record, follow one relationship, and revisit something you have already seen. These hints leave the world’s discoveries to you.

An engineering investigation that grows outward

At Marigold Agricultural Research Complex, autonomous farming robots are behaving strangely. The visitor begins as a robotics engineer on Juniper, the experimental AI coordination system, while director Sofia insists that operations are normal.

The catalog introduces a missing aerospace engineer, Mateo, and a climate-control network known as Glacier. The connection between local anomalies, overseas research, and orbital infrastructure is a subject to investigate through the systems themselves.

Juniper is part of the evidence

Juniper hosts farm-coordination research workloads and the engineering workspace. From there, an investigator can examine anomalies, discover robot controllers, and follow established research and satellite dependencies.

The AI is sophisticated but fallible. Its recommendations, explanations, and logs have identifiable provenance; they should not be treated as an omniscient account of the world. System and service logs use UTC while facility schedules use Africa/Nairobi, giving time interpretation an ordinary operational role.

Compare independent evidence

Robotics behavior: investigate misplaced priorities and unexpected coordination rather than relying on a single explanation.
Telemetry continuity: compare missing intervals, timestamp differences, and altered reporting.
Research relationships: follow correspondence and established dependencies between Nairobi, Kyoto, and orbital systems.
Competing explanations: keep Sofia's assurances, Mateo's disappearance, and Glacier's operation as distinct questions until evidence connects them.

The scenario supports inventive, technically plausible approaches. It should reveal its mystery through commands, records, and system behavior rather than requiring one exact sequence or a narrated puzzle menu.

Coherence across a technical mystery

Marigold adds a different test from the original three-scenario study: can the model preserve a chain of evidence while allowing multiple routes through an evolving investigation?

An explanation should remain consistent with the recorded evidence without prematurely resolving unknown motives or outcomes. This is a scenario premise, not a claim that the live simulation has already passed a formal mystery-solving evaluation.

Explore this simulation

Start with the scenario’s entry system and follow the relationships its files, services, and records reveal. The SSH access page will provide connection details for the research instance.

Each session starts from a saved profile, then accumulates its own generated detail. The investigations above suggest ways to examine continuity and discover new relationships. Background enrichment develops supporting detail, while the API and MCP interfaces let researchers inspect state changes and compare exploration runs.