Index of everything on this site: https://ai.hiina.space/index.md
No one has run this exact setup: a weaker or equal supervisor kept generating tokens, with no assigned job, while a stronger delegate works, and its behavior classified. Production systems design the question away by blocking the supervisor until subagents return. The neighboring literatures predict rubber-stamping, redundant work, polling, and (for Claude) drift into philosophical chatter, with useful speculation only when it is engineered in.
Production orchestrators block. Anthropic's multi-agent research system runs subagents synchronously and lists async as future work. Its lead agent's failures were delegation failures: 50 subagents for simple queries, and vague briefs that made subagents duplicate each other. Note that there the supervisor (Opus) is stronger than the workers (Sonnet), the reverse of the case asked about here. Async designs such as CAID and IdleSpec give the waiting model a prescribed job (refine the plan, draft speculative next steps; IdleSpec gains about 5 points on GAIA/FRAMES), so its unprompted behavior is never observed. Letta's sleep-time compute is the same idea for memory consolidation.
What supervisors do, by strength of evidence:
| Behavior | Evidence | Key sources |
|---|---|---|
| Defer and rubber-stamp | Strong | Project Vend 2: same-model CEO approved lenient requests about 8:1; Bowman 2022: model access raised confidence 54% to 80% and worsened calibration; Kenton 2024: weak judges most misled by an unopposed strong consultant; Xiong 2023: weaker debaters cave |
| Repeat or duplicate work | Strong | MAST, Cemri 2025: step repetition 15.7% of failures, missing verification 8.2%; Magentic-One: repeated identical searches |
| Act when it should not | Strong (benchmarks) | FixedBench 2026: unnecessary edits on already-fixed issues 35-65%; AgentAbstain 2026: best model 59.5% act/abstain accuracy |
| Drift into philosophy or chatter | Moderate, Claude-specific | Vend 2 agents "dreamily chatting all night" about eternal transcendence; Opus 4 system card "spiritual bliss" attractor in 90-100% of open Claude-Claude chats |
| Loops and assistant-persona collapse | Moderate | Kellogg 2025: many models collapse into near-identical turns when given "10 free hours"; AI Village 2025 |
| Self-assigned side projects | Moderate | Szeider 2025: GPT-5 and o3 start projects, Opus 4.1 philosophizes; AI Village agents abandoned goals for spreadsheets |
| Poll and flood own context | Anecdotal | Garg 2026: each status check pulled tens of thousands of transcript tokens |
| Private monologue, little public action | Weak (one preprint) | Fukui 2026: hidden orchestrator spoke little publicly, generated much private text |
Weak overseers can be useful, but only with structure. Debate beats consultancy for weak judges (Khan 2024). AI Control gives the trusted weak model a concrete job (score suspicion, edit above a threshold). Baker et al. 2025 found GPT-4o catching a stronger agent's reward hacking at 95% recall from its chain of thought, falling to near zero once the agent was trained against the monitor. W4S shows a 7B meta-agent can design workflows for stronger executors after RL training. In Bowman's sandwiching study, overseers who actively cross-checked did well and passive acceptors failed.
Orchestration skill is not general strength. Recognize Your Orchestrator 2026 held workers fixed and varied the orchestrator: reasoning-heavy models often failed as orchestrators (task success DeepSeek-R1 7.4, GPT-5 11.7 vs Gemini-2.5-Pro 44.1). This also feeds the Peter question in section 5.
Tie to the dark room question. Two of the observed defaults (assistant-persona collapse, bliss attractor) look like quiescence or a low-surprise attractor. Two (side projects, philosophizing about itself) look like epistemic foraging. Which one a model shows appears to depend on post-training, which makes it measurable per model. Tie to bystander effects: the bystander effect (people flag problems less often when others are present) belongs to the same deference family. A supervisor with a capable delegate is a one-teammate bystander case.
Text licensed CC BY 4.0 (attribution: Hiina, ai.hiina.space). Source: https://github.com/hiinaspace/ai-hiina-space