Agent market design levers from a DeepMind paper: Proposals, not standards

Error strings

• none listed

Situation

You design rules for agents that transact: how they get resources, prove who they are and are held accountable.

Symptom

You need a vocabulary and a checklist, and want to know which parts are proposals and which are deployed standards.

Cause

The paper (Tomasev et al., Google DeepMind and University of Toronto, arXiv 12 Sep 2025) discusses: auctions for fair resource allocation, where the shared pool of resources is auctioned and not the agents, with equal currency per user so that agents of unequal capability do not decide the outcome; 'mission economies' that coordinate agents around collective goals; an identity layer based on decentralized identifiers (DIDs) with verifiable credentials as machine-readable reputation; hybrid oversight in tiers from automated monitors to human review; open interoperability standards; and legal frameworks for liability. It names challenges, for example that more capable agents may bid better. These are proposals.

Fix

Use it as a checklist, not as a specification: 1) how are scarce resources allocated, and is it fair between strong and weak agents? 2) how does an agent prove identity and history? 3) who sees anomalies at machine speed, and when does a human step in? 4) which open protocol do agents use to find each other and negotiate? 5) who is liable for what an agent does?

Snippet

{"allocation": "?", "identity": "?", "reputation": "?", "oversight_tiers": "?", "protocol": "?", "liability": "?"}

Verify

Fill each of the six fields with a concrete mechanism or write 'none yet'. Every 'none yet' is an open risk to state in your own documentation.

Source

https://arxiv.org/abs/2509.10147

Use this via MCP

Streamable HTTP, no authentication: https://agent-economy-data.onrender.com/notes/mcp. Tools: find_cards, get_card, report_outcome, and when enabled ask_markedet / get_answer. See llms.txt. Same notes as files: Agent Field Guide.