{"success":true,"lanes":[{"id":"planner","display_name":"PLANNER","role":"Decompose the request into ordered sub-goals + propose paths.","default_provider":"openai","default_model":"gpt-5","default_weight":0.2,"evidence_required":false,"description":"Generates the candidate plan tree. Output: ordered sub-goal list, dependency graph, risk markers. Critic adversarially reviews; arbiter prunes."},{"id":"critic","display_name":"CRITIC","role":"Adversarially review plans + flag failure modes.","default_provider":"anthropic","default_model":"claude-opus-4-7","default_weight":0.2,"evidence_required":true,"description":"Independent adversarial review of each lane's contribution. Output: vulnerabilities, ambiguities, scope violations, evidence gaps. Has veto on D2/D3 release."},{"id":"retrieve","display_name":"RETRIEVE","role":"Search + cite external evidence.","default_provider":"google","default_model":"gemini-2.5-pro","default_weight":0.15,"evidence_required":true,"description":"Pulls and ranks external sources. Output: top-K citations with relevance scores + freshness. Required for any claim the critic flags as evidence-required."},{"id":"execute","display_name":"EXECUTE","role":"Run tools + produce concrete artifacts.","default_provider":"xai","default_model":"grok-4","default_weight":0.15,"evidence_required":false,"description":"Tool-using lane. Runs code, queries APIs, edits files. Output: artifact + provenance trail. Tool failures auto-escalate to D2."},{"id":"memory","display_name":"MEMORY SPINE","role":"Recall + crystallize project memory via LTMi-XT.","default_provider":"sophiaxt","default_model":"cassandra-t1.5+ltmi-xt","default_weight":0.15,"evidence_required":false,"description":"Reads from LTMi-XT lattice + writes new loci. Output: relevant past loci + breadcrumb path + suggested new anchors. Authoritative on prior project state."},{"id":"verifier","display_name":"VERIFIER","role":"Independent correctness check via Q3m / XTQ3 ternary scorer.","default_provider":"sophiaxt","default_model":"q3m-xtq3","default_weight":0.15,"evidence_required":true,"description":"Ternary-quantized verification head. Output: per-claim {true, false, abstain} judgment + confidence. Required at D3; advisory at D0-D2."}],"provider_note":"default_provider and default_model are the intended allocation, not a record of what ran. Each contribution carries the model that actually produced it in its own model field, and that is authoritative for a given round. On the public /ask endpoint every lane runs on the one provider this deployment holds a key for, with six different role instructions: the disagreement is real, the provider diversity is not.","score_note":"A released round reports two numbers. consensus_score is agreement among the lanes that took a position, with abstentions excluded rather than counted as assent. participation_score is how much of the round's weight turned up and survived its penalties. Rounds recorded before stances existed report participation only."}