ARC MAP
Investigate Arc activity with sourced reports and a baseline you can revisit.
Demo
About
ARC MAP turns activity on Arc into questions you can test. Start with a sourced signal on Today, run a read-only Hunter, inspect the transactions, then save a thesis with a fixed baseline and a finite check budget. Return later to see what changed without rewriting the original observation. In the recorded demo, The Graph returns 200 SUN transfer events, all from one transaction. ARC MAP reports that limit instead of treating the event count as independent activity. When the index advances but the transfer evidence stays unchanged, the thesis stays unchanged too. The current source also supports sharing one completed investigation: a contributor adds a public source and counterevidence, the owner reassesses, and revocation removes access while preserving the original report and contribution. This sharing flow was tested separately and is not shown in the recorded demo. The Graph supplies indexed transfer evidence; Arcscan and GitHub remain separate, labelled sources. You can explore without a wallet. Built-in analysis is rules-based; external agents can use the same bounded research operations through MCP.
How it's made
Next.js 16, React 19 and TypeScript serve the browser and the research API. Node 22 workers run three loops: a public-source collector, an Arcscan radar, and a finite thesis checker with lease-fenced checks. SQLite stores observations, radar entries, private missions and theses; reports are immutable and comparisons are pinned to the baseline observed when the thesis was saved. The Graph supplies the transfer evidence. We deployed an ERC20 Transfer subgraph for the SUN token on Arc testnet (chain 5042002) in subgraphs/arcmap, and the Hunter queries it with bounded, dated samples. Arcscan and GitHub are separate, separately labelled evidence sources. A Graph outage is shown as a Graph error, never relabelled as an explorer result. Privy is wired for wallet login and viem for chain reads, but the submitted journey is read-only and performs no transaction. Contracts in contracts/ (Foundry) are the research-payment escrow developed during the event and are not exercised by this demo. Oscar set the product direction, workflow, visual direction, priorities and acceptance requirements. Codex and Cursor wrote most of the implementation and tests under his review, and Claude and Fable did source reviews. The product analysis is rules-based and does not call an LLM.