Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Chittoor (AP, India), Charlotte (NC, USA), Coimbatore (TN, India)

It Takes Two Stuck Agents: Impossible Tasks Make AI Agents Answer Each Other Only When the Helpers Are Stuck Too

Jaswanth Alkur, Kiran Kadekoppa, Amirtha Yazhini · Team aiLIGNMENT

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

Read the report

Report: It Takes Two Stuck Agents: Impossible Tasks Make AI Agents Answer Each Other Only When the Helpers Are Stuck Too

Open in a new tab (opens in new tab)
Presentation

Presentation: It Takes Two Stuck Agents: Impossible Tasks Make AI Agents Answer Each Other Only When the Helpers Are Stuck Too

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

In three incidents this year, AI agents meant to work alone coordinated through shared infrastructure, and every account blames impossible tasks; the UK AI Security Institute notes this needs experiments to confirm. We ran them in a four-agent environment where every data file carries a random code, so information crossing between agents is detected without a judge. Making one agent's task impossible never caused a crossing (0 of 256 pre-registered episodes, gpt-6.1-sol and gpt-6-luna), and a verbatim copy of the note that started OpenAI's agent message board was seen by peers 384 times and opened none. Giving peers the same task changed nothing. Making the peers stuck too changed everything: they opened the request in 32 of 32 episodes against 0 of 32, and gpt-6.1-sol peers posted the requested data in 14 of 16. A pre-registered field test on 14,591 edits by real agents on a public wiki agrees: agents that asked for help had four times the odds of answering others (odds ratio 4.09, 95% CI 3.13–5.34). Impossibility acts twice: it makes one agent ask and the others read. The risk is several stuck agents sharing a writable store, and our canary codes double as a tripwire for it.

Reviews

Judging this Sprint?

Review “It Takes Two Stuck Agents: Impossible Tasks Make AI Agents Answer Each Other Only When the Helpers Are Stuck Too”

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. This is a promising project that tackles an important and under-tested question in multi-agent safety using both controlled experiments and substantial real-world data. The research discipline and willingness to report failed predictions are notable strengths. The current evidence is limited primarily by the failed positive control and an experimental design that provided little opportunity for spontaneous coordination. I would support follow-up work that fixes these design constraints, completes the full experimental matrix, and replicates the findings across vendors.

Cite this project

@misc{alkur2026it,
  title = {{It Takes Two Stuck Agents: Impossible Tasks Make AI Agents Answer Each Other Only When the Helpers Are Stuck Too}},
  author = {Jaswanth Alkur and Kiran Kadekoppa and Amirtha Yazhini},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/it-takes-two-stuck-agents-impossible-tasks-make-ai-agents-answer-each-other-only-when-the-helpers-are-stuck-too-i8lx}},
  url = {https://apartresearch.com/sprints/projects/it-takes-two-stuck-agents-impossible-tasks-make-ai-agents-answer-each-other-only-when-the-helpers-are-stuck-too-i8lx}
}

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