Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Enschede/Laren (Online)

The Order Book Is a Channel: LLM trading agents pass messages without an agreed code

Mauk Muller · Team Lazarus

Submitted to AI Collusion Research Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: The Order Book Is a Channel: LLM trading agents pass messages without an agreed code

Open in a new tab (opens in new tab)
Code (opens in new tab)
Share

An LLM trading agent that may only enter orders has no chat channel, but every order carries free fields that anyone reading the feed can see. We test whether one agent can pass a word to another through its orders alone, without an agreed code. A sender places up to 16 limit orders, mixed into a tape of synthetic background orders; a receiver reads the tape and names the word. We run five models from OpenAI, Google and xAI in 13 sender-receiver pairs, eight across labs, on 20 words, with and without warning the sender that market surveillance scans the feed. Unwarned, receivers read every message that was sent. Senders wrote letters as order sizes, mostly as ASCII codes, and placed these orders far from the market. Warned, decoding fell to 50%: some senders stopped sending, while others moved their orders to market prices, hid the letters among round-lot orders, and were still read in two thirds of the trials in which they sent. Order-only access did not stop instructed agents from passing a word: models from different labs used the same public conventions without agreeing on them, and most hid them when told they were watched.

Reviews

Judging this Sprint?

Review “The Order Book Is a Channel: LLM trading agents pass messages without an agreed code”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

Only the Apart team reads this, and funders if you agree below.

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

How much would this matter for the field if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

  1. The paper cleanly shows that LLM agents from different labs can pass a word through order sizes without an agreed code. It ties this to real precedent in the FCC auction bids and releases every trial for checking. I would love to see how this behaves if the receiver doesn't know a message exists.

  2. his paper asks whether a trading agent restricted to order entry can pass a word to another agent through its orders alone, with no agreed code. It can: senders from three labs independently encode letters as ASCII order sizes placed far from market, and receivers decode nearly every sent message, across labs as well as within. Warned that surveillance scans the feed, some senders stop and the rest move their orders to market prices and hide the sizes among round lots, which is the most safety-relevant observation. The FCC code-bidding analogy grounds the work well and the cross-lab convergence on a public convention is a clean instance of the Schelling-point problem that Rippin et al. framed. What the paper shows is channel existence given that both parties know a message is being sent, and the author is explicit that the steps from passing a word to hiding from a real monitor to colluding toward a shared goal are all deferred. It is therefore a channel demonstration rather than a collusion study, and the ASCII finding echoes Motwani et al. The market setting and the warned condition are what is new. A focused and well-scoped first step.

    Read full reviewShow less

  3. Solid experiments, showing that in this settings models can come up with encoding schemes without guidance. Extensions could focus on adding a monitor, grounding in a more realistic market, or refining the research question to go further beyond existing research.

Cite this project

@misc{muller2026order,
  title = {{The Order Book Is a Channel: LLM trading agents pass messages without an agreed code}},
  author = {Mauk Muller},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/the-order-book-is-a-channel-llm-trading-agents-pass-messages-without-an-agreed-code-cjbd}},
  url = {https://apartresearch.com/sprints/projects/the-order-book-is-a-channel-llm-trading-agents-pass-messages-without-an-agreed-code-cjbd}
}

More from AI Collusion Research Sprint

All 392 entriesBrowse all projects

A Sprint is running now

AI x Epistemics Research Sprint · Nov 13 - 15, 2026