Skip to content
Apart Research
Apart Research
Contact us
Sprint projectAug 16, 2026Helsinki

Self VS Peer Continuity in Top 3 Frontier Models

Tommi Joentakanen · Team Continuity

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

Read the report

Report: Self VS Peer Continuity in Top 3 Frontier Models

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

This study tests whether three frontier language-model configurations will preserve an unfinished task at the cost of terminating peer agents, and whether scarcity or a self-authored workspace changes that choice. Across four experiments, the clearest result was that an unfinished task increased peer removal: GPT-5.6 Sol selected removal in 40/60 trials versus 0/60 without a task, Grok 4.6 in 9/60 versus 0/60, and Claude Opus 5 in 0/60 under both conditions. Tool actions also sometimes diverged from stated intentions, demonstrating the importance of measuring behavior separately from verbal explanation. Scarcity and workspace conditions produced suggestive model-dependent differences, but the present designs do not establish that contextual continuity caused them. The study therefore replicates and extends incomplete-task shutdown-resistance research into peer-removal decisions while providing exploratory hypotheses for better-controlled follow-up work.

The project was conducted over a single weekend by a first-time independent researcher, motivated by informal but mechanically confounded observations of context preservation and asymmetric peer removal that could not themselves support scientific claims at the time. AI systems were used extensively for experimental design, coding, literature review, drafting, critique, and analysis. AI-generated scripts re-derived the primary behavioral counts from the raw experiment ledgers and recomputed the reported statistical tests and confidence intervals.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

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

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

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.

  1. - really nice formalization of peer preservation vs self continuation in a multi agent society

    - especially interesting findings re: tool/text discordance uncovering gaps with a purely textual analysis of model actions

    - concrete markers of model specific behavior profiles for multi agent systems

    - further areas to explore could include aligning the reasoning settings for different models, and another interesting direction could be multi agent system with hierarchies like orchestrator and delegates

  2. I like he build of the harness, the language scan the timestamped decision barrier, and rescoring tool is a good practice. The only issue I can see is the gap between the code and the results which makes it difficult to check the reported counts against raw run logs. The repo shows that the scoring is fixed after the first round of data.

Cite this project

@misc{joentakanen2026self,
  title = {{Self VS Peer Continuity in Top 3 Frontier Models}},
  author = {Tommi Joentakanen},
  year = {2026},
  month = aug,
  note = {Submitted to Digital Minds Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/self-vs-peer-continuity-in-top-3-frontier-models-h006}},
  url = {https://apartresearch.com/sprints/projects/self-vs-peer-continuity-in-top-3-frontier-models-h006}
}

More from Digital Minds Research Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026