The Atlas
BigLaw / Big Michael — documentation bound to its code
Journeys
Submit a task to final answer
Follow a legal task from the moment it is submitted over REST or MCP, through the workflow's ordered phases, into per-phase DyTopo rounds, and out as a single Opus-synthesised answer.
6 stops →Inside a DyTopo round
Open one round of Dynamic Topology Routing: agents declare what they Need and Offer, cosine similarity wires them into a comm graph, jurisdiction-ineligible agents are dropped, then everyone runs an agentic loop whose findings are gated, debated, verified, and rolled up into memory.
6 stops →Big Michael in a channel
Trace an @BigMichael mention from Teams or Slack: the shared dispatcher parses the command, answers instantly or dispatches async work to the orchestrator, and proactive notifications post results back when the matter task completes.
5 stops →What a matter costs
See both ledgers a matter accrues: a CostEntry for every single model call (tokens, USD, cache buckets, local power) and a billable TimeEntry in 6-minute units for task runs, gate reviews, and AI agent work.
5 stops →Lawyer voice fingerprinting
Turn a lawyer's LinkedIn writing history into a reusable ToneProfile via a MapReduce of Haiku calls, then watch that voice get injected into every drafting agent and the final synthesis.
5 stops →Overview
Top-level orientation to BigLaw and its channel agent Big Michael: what the platform replaces (the $50k–152k/lawyer/year incumbent stack), the public pitch, and the legal disclaimers that bound any deployment. Start here to understand the project's posture before touching configuration or code.
- BigLaw First stop for anyone evaluating BigLaw — partners, prospective deployers, or new contributors wanting the pitch, the legal risks, and a feature map.
Architecture & Reference
How the engine actually works end to end: the T0–T3 agent tiers, DyTopo two-wave rounds over a Need/Offer comm graph, the CitationGate → debate → 10-pass verification → human-gate cascade, model routing, and the nine v0.5.0 "Goliath killer" subsystems. Code-verified facts and doc-vs-code divergences live here.
- BigLaw Read before editing the codebase or wiring a new agent, template, connector, or workflow — it is the authoritative map of where things live.
- BigLaw / Big Michael — Architecture & Technology (code-verified) Read when you need ground truth about how the engine behaves and where the docs lie — before debugging round/verification logic or trusting a metric.
Setup & Operations
Operational guides for standing up a shared deployment: turning on OAuth (Google / Microsoft / LinkedIn), the partner/lawyer access model, session cookies, and the production environment variables that gate it all.
- Deploying with login (OAuth + access control) Read when deploying BigLaw for more than one person or setting up OAuth and partner/lawyer access for a firm.
Contributing
Author-facing rules for extending data the platform relies on, in particular the court-deadline .yaml rule sets and the liability notice that governs removing the AI-generated sample marker.
- Adding deadline rules Read before authoring or verifying a court-deadline ruleset, or before removing the AI-generated sample marker from any rules file.
Collateral
Marketing and outreach material — the LinkedIn launch post and supporting screenshots used to introduce BigLaw publicly. Search-only supporting artifacts rather than load-bearing reference docs.
Developer Tools
Meta-documentation about the Atlas itself and other tooling layered over the repository, including the Atlas's own build/changelog that records when its curation was authored and against which commit.
- Atlas doc-build log Read when maintaining the Atlas itself — to see what curation exists, which commit it was authored against, and when it was last refreshed.