humanish 0.108.0 → 0.109.0

package/dist/routes/computer-use/bundle-parts.js

@@ -8,11 +8,12 @@ import {} from "../../run/bundle.js";
 import { participantResourceIds } from "../../run/participant-records.js";
 import { resolveSelfReportedFriction } from "./self-report.js";
 import {} from "./types.js";
+import { plural } from "../../run/text.js";
 /** The human-readable state story appended to the provenance event (and review.md via it). */
 export function describeSubjectState(state, dryRun) {
     switch (state.provenance) {
         case "seeded":
-            return `seeded (${state.seed?.length ?? 0} step(s): ${(state.seed ?? []).map((record) => record.name).join(", ")})`;
+            return `seeded (${plural(state.seed?.length ?? 0, "step")}: ${(state.seed ?? []).map((record) => record.name).join(", ")})`;
         case "unpinned":
             return `unpinned (external: ${(state.externalEnvNames ?? []).join(", ")})`;
         case "declared-not-run":
@@ -103,7 +104,7 @@ export function participantFeedbackCandidates(args) {
             proposed_next_state: "study-quality-review",
             acceptance_proof: [
                 feedbackProofCommands(args.runId).verify,
-                feedbackProofCommands(args.runId).watch,
+                feedbackProofCommands(args.runId).observe,
             ],
         });
     }

package/dist/routes/computer-use/bundle.d.ts

@@ -1,5 +1,5 @@
 import type { CuaActorDescriptor } from "../../actors/registry.js";
-import type { ComputerUsePlan } from "../../lab/plan-types.js";
+import type { ComputerUsePlan } from "../../study/plan-types.js";
 import type { BundleRun, RunBundle, RunRerunLineage } from "../../run/bundle.js";
 import { type Judgment } from "../../run/judge.js";
 import type { CuaParticipantPlan, DesktopParticipantRun, CuaSubjectProjection, CuaSubjectProvenanceArg, ParticipantRunOutcome } from "./types.js";
@@ -20,7 +20,7 @@ export interface CuaRunBundleBase {
 /**
  * The judgment for the current state of a computer-use run, by the rule for its shape: one
  * participant (its tallied status is the verdict) or several (every participant must pass). The bundle's
- * verdict and the lab result's ok both come from it.
+ * verdict and the study result's ok both come from it.
  */
 export declare function judgeComputerUseRun(base: Pick<CuaRunBundleBase, "participantRuns" | "rerun">, state: {
     dryRun: boolean;

package/dist/routes/computer-use/bundle.js

@@ -9,6 +9,7 @@ import { providerResourcesForOutcome } from "./bundle-parts.js";
 import { participantFactsOf } from "./participant-facts.js";
 import { buildCuaFanoutBundle } from "./fanout-bundle.js";
 import { buildSingleParticipantBundle } from "./single-bundle.js";
+import { plural } from "../../run/text.js";
 /** One participant without a rerun keeps the single-participant bundle shape and its rule. */
 function isOneParticipantRun(base) {
     return base.participantRuns.length === 1 && base.rerun === undefined;
@@ -16,7 +17,7 @@ function isOneParticipantRun(base) {
 /**
  * The judgment for the current state of a computer-use run, by the rule for its shape: one
  * participant (its tallied status is the verdict) or several (every participant must pass). The bundle's
- * verdict and the lab result's ok both come from it.
+ * verdict and the study result's ok both come from it.
  */
 export function judgeComputerUseRun(base, state) {
     const inProgress = state.inProgress === true;
@@ -84,8 +85,8 @@ function singleParticipantArgs(base, state) {
             ? {}
             : { caseGroup: spec.planned.labels.caseGroup }),
         dryRun: state.dryRun,
-        labId: plan.labId,
-        ...(plan.title ? { labTitle: plan.title } : {}),
+        studyId: plan.studyId,
+        ...(plan.title ? { studyTitle: plan.title } : {}),
         mission: spec.evidenceInstructions ?? spec.instructions,
         ...(spec.evidenceAssignment === undefined ? {} : { assignment: spec.evidenceAssignment }),
         persona: spec.persona,
@@ -165,9 +166,9 @@ export function renderCuaReviewMarkdown(bundle) {
                 `- actor: ${trace.provider} (${trace.lane}/${trace.protocol})`,
                 // Name the trace's actual screenshot mode ("raw" | "blurred"); say
                 // nothing when no frames exist ("n/a") rather than claim a redaction that never ran.
-                `- evidence: ${trace.items.length} trace item(s), ${trace.counts.screenshots ?? 0} ${trace.redaction.screenshots === "raw" || trace.redaction.screenshots === "blurred"
-                    ? `${trace.redaction.screenshots} screenshot(s)`
-                    : "screenshot(s)"}`,
+                `- evidence: ${plural(trace.items.length, "trace item")}, ${plural(trace.counts.screenshots ?? 0, trace.redaction.screenshots === "raw" || trace.redaction.screenshots === "blurred"
+                    ? `${trace.redaction.screenshots} screenshot`
+                    : "screenshot")}`,
             ]
             : []),
         ...(bundle.review.gaps.length > 0

package/dist/routes/computer-use/e2b-desktop/comms.d.ts

@@ -1,12 +1,12 @@
 import { type OriginMap } from "../../../comms/capture-surface.js";
 import { type DeployedCommsCatch } from "../../../comms/sandbox-catch.js";
-import type { LabCommsEmail, LabConfig } from "../../../lab/types.js";
+import type { StudyCommsEmail, StudyConfig } from "../../../study/types.js";
 import type { Shell } from "../../../substrates/shell.js";
 import type { ReadyParticipantDesktop } from "../participant-desktop.js";
 import { type E2BDesktopDeps, type DesktopParticipantRun } from "../types.js";
 /** A participant's captured-email wiring, decided from config before the sandbox exists. */
 export interface ParticipantComms {
-    readonly email: LabCommsEmail;
+    readonly email: StudyCommsEmail;
     readonly port: number;
     readonly smtpPort: number | undefined;
     /** The loopback URL the persona opens to read captured mail. */
@@ -20,9 +20,9 @@ export interface ParticipantComms {
 /**
  * On an in-sandbox subject route, redirect the app's email-API sends into an in-sandbox catch
  * (loopback) so its verification mail is captured instead of sent to the internet. Undefined when the
- * lab declares no fake email, which leaves the participant unchanged.
+ * study declares no fake email, which leaves the participant unchanged.
  */
-export declare function planParticipantComms(comms: LabConfig["comms"], serveUrl: string | undefined, targetUrl: string, inSandboxSubject: boolean): ParticipantComms | undefined;
+export declare function planParticipantComms(comms: StudyConfig["comms"], serveUrl: string | undefined, targetUrl: string, inSandboxSubject: boolean): ParticipantComms | undefined;
 /** The env the subject app reads to reach the catch, injected at sandbox create so it boots with it. */
 export declare function participantCommsEnv(comms: ParticipantComms | undefined): Record<string, string>;
 /** A started catch and the stop for its inbox-surface loop. */
@@ -33,7 +33,7 @@ export interface RunningCommsCatch {
 }
 /**
  * Start the in-sandbox email catch before the subject serve, so the app's send-API base URL
- * resolves the moment it boots. A comms-declared lab that cannot stand the catch up fails closed
+ * resolves the moment it boots. A comms-declared study that cannot stand the catch up fails closed
  * rather than silently sending real mail.
  */
 export declare function startCommsCatch(shell: Shell, comms: ParticipantComms, requestTimeoutMs: number): Promise<RunningCommsCatch>;

package/dist/routes/computer-use/e2b-desktop/comms.js

@@ -9,14 +9,15 @@ import { redactText, toErrorMessage } from "../../../evidence/redaction.js";
 import { writeContainedOutputFile } from "../../../run/contained-output.js";
 import { inboxRecipientFor, participantHasInboxRecipient } from "../participant-desktop.js";
 import { participantServeUrl } from "../types.js";
-import { addressedRecipients } from "../../../lab/parse/comms.js";
+import { addressedRecipients } from "../../../study/parse/comms.js";
+import { plural } from "../../../run/text.js";
 /** Mid-run inbox-surface render cadence (ms). Coarse enough that the per-tick `cat` + file writes stay
  *  cheap; fine enough that a verification email is visible seconds after the app sends it. */
 const INBOX_SURFACE_CADENCE_MS = 2500;
 /**
  * On an in-sandbox subject route, redirect the app's email-API sends into an in-sandbox catch
  * (loopback) so its verification mail is captured instead of sent to the internet. Undefined when the
- * lab declares no fake email, which leaves the participant unchanged.
+ * study declares no fake email, which leaves the participant unchanged.
  */
 export function planParticipantComms(comms, serveUrl, targetUrl, inSandboxSubject) {
     const email = inSandboxSubject && comms?.email?.kind === "fake" ? comms.email : undefined;
@@ -63,7 +64,7 @@ export function participantCommsEnv(comms) {
 }
 /**
  * Start the in-sandbox email catch before the subject serve, so the app's send-API base URL
- * resolves the moment it boots. A comms-declared lab that cannot stand the catch up fails closed
+ * resolves the moment it boots. A comms-declared study that cannot stand the catch up fails closed
  * rather than silently sending real mail.
  */
 export async function startCommsCatch(shell, comms, requestTimeoutMs) {
@@ -203,7 +204,7 @@ export async function drainCommsEvidence(args) {
             // Captured mail that matched no declared recipient must not vanish silently: tell the
             // operator to declare comms.email.recipients[].address to match
             // the address the app actually sends to (e.g. the one the persona surface will sign up with).
-            warnings.push(`The email catch captured ${collected.captured} email send(s), but none matched a declared recipient inbox, so no email evidence was written. Set comms.email.recipients[].address to the address the app sends to.`);
+            warnings.push(`The email catch captured ${plural(collected.captured, "email send")}, but none matched a declared recipient inbox, so no email evidence was written. Set comms.email.recipients[].address to the address the app sends to.`);
         }
         else {
             // Zero captures is the silent-broken shape: the app never posted to the catch at

package/dist/routes/computer-use/e2b-desktop/fidelity.d.ts

@@ -1,4 +1,4 @@
-import type { ResidualConfig } from "../../../lab/plan-types.js";
+import type { ResidualConfig } from "../../../study/plan-types.js";
 import type { RunDesktopGeometry } from "../../../run/streams.js";
 import type { DesktopBrowserFamily, DesktopBrowserLaunchIdentity } from "../../../substrates/e2b/desktop-browser.js";
 import { makeChromeBrowserStateObserver } from "../../../substrates/e2b/desktop-cdp.js";

package/dist/routes/computer-use/e2b-desktop/prepare.d.ts

@@ -1,6 +1,6 @@
 import { type E2BDesktopSandbox } from "../../../substrates/e2b/sdk.js";
 import type { E2BParticipantContext, E2BParticipantState } from "./state.js";
-export declare const CUA_ACTOR_LAB_PROVIDER_METADATA: {
+export declare const CUA_ACTOR_STUDY_PROVIDER_METADATA: {
     readonly mode: "cua-actor-lab";
     readonly tool: "humanish";
 };

package/dist/routes/computer-use/e2b-desktop/prepare.js

@@ -11,7 +11,7 @@ import { loadE2BDesktopModule } from "../../../substrates/e2b/sdk.js";
 import { e2bShell } from "../../../substrates/e2b/shell.js";
 import { attachReceivingInbox, participantCommsEnv, startCommsCatch } from "./comms.js";
 import { participantSubjectEnv } from "../types.js";
-export const CUA_ACTOR_LAB_PROVIDER_METADATA = {
+export const CUA_ACTOR_STUDY_PROVIDER_METADATA = {
     mode: "cua-actor-lab",
     tool: "humanish",
 };
@@ -37,8 +37,8 @@ export async function acquireParticipantDesktop(ctx, state) {
             requestTimeoutMs: deps.requestTimeoutMs,
             timeoutMs: deps.sandboxMs,
             metadata: {
-                ...CUA_ACTOR_LAB_PROVIDER_METADATA,
-                labId: deps.labId,
+                ...CUA_ACTOR_STUDY_PROVIDER_METADATA,
+                labId: deps.studyId,
                 recordId: spec.recordId,
                 participantId: spec.planned.id,
                 participantIndex: String(spec.planned.index),

package/dist/routes/computer-use/e2b-desktop/start.js

@@ -7,7 +7,6 @@ import { captureDesktopBrowserGeometry } from "../../../substrates/e2b/desktop-g
 import { prepareDesktopMedia, startE2BDesktopMedia, } from "../../../substrates/e2b/desktop-media.js";
 import { startE2BDesktopRecording } from "../../../substrates/e2b/desktop-recording.js";
 import { applyParticipantMobileFidelity, mobileLaunchFlags } from "./fidelity.js";
-import { streamEvent } from "../../../lab/run-lab-homes.js";
 /** Start the declared speech worker and screen recording. A recording that cannot start is a warning. */
 export async function startParticipantMedia(ctx, state, desktop) {
     const { spec, deps, warnings } = ctx;
@@ -58,7 +57,7 @@ export async function openParticipantSurface(ctx, state, desktop) {
     // uploaded first, and a feed that cannot be produced fails the participant closed here.
     const mediaEvidence = requestedMedia === undefined
         ? undefined
-        : await prepareDesktopMedia(desktop, requestedMedia, residual.policies?.mediaPermission ?? "prompt", deps.labCwd, deps.requestTimeoutMs);
+        : await prepareDesktopMedia(desktop, requestedMedia, residual.policies?.mediaPermission ?? "prompt", deps.studyCwd, deps.requestTimeoutMs);
     const emulationFlags = mobileLaunchFlags(residual, spec);
     const browserLaunch = await openDesktopBrowserTarget(desktop, targetUrl, deps.requestTimeoutMs, residual.execution?.desktop?.browser, [...emulationFlags, ...(mediaEvidence?.flags ?? [])], state.speech?.env ?? state.recording?.env);
     state.desktopBrowser =
@@ -169,14 +168,14 @@ export async function startParticipantStream(ctx, state, desktop) {
         });
         if (typeof candidateStreamUrl === "string" && candidateStreamUrl.trim().length > 0) {
             state.streamUrl = candidateStreamUrl;
-            await deps.onStream(streamEvent({
+            await deps.onStream({
                 type: "ready",
                 participantId: spec.planned.id,
                 sandboxId: desktop.sandboxId,
                 recordId: spec.recordId,
                 streamId: spec.streamId,
                 url: candidateStreamUrl,
-            }));
+            });
         }
         else {
             warnings.push("Live desktop stream started but did not return a usable watch URL; Observer will fall back to screenshots.");

package/dist/routes/computer-use/e2b-desktop/state.d.ts

@@ -12,7 +12,7 @@ import type { E2BDesktopSandbox } from "../../../substrates/e2b/sdk.js";
 import type { ParticipantDesktopEvidence } from "../participant-desktop.js";
 import type { ParticipantComms, RunningCommsCatch } from "./comms.js";
 import type { ParticipantFidelity } from "./fidelity.js";
-import type { CuaActorLabErrorCode, E2BDesktopDeps, DesktopParticipantRun, SandboxReleaseFact } from "../types.js";
+import type { CuaActorStudyErrorCode, E2BDesktopDeps, DesktopParticipantRun, SandboxReleaseFact } from "../types.js";
 /** What every step of one participant reads: its spec, the run's dependencies and where it points. */
 export interface E2BParticipantContext {
     readonly spec: DesktopParticipantRun;
@@ -38,7 +38,7 @@ export interface E2BParticipantState {
     desktopResources: DesktopResourceObservation | undefined;
     released: boolean;
     sandboxRelease: SandboxReleaseFact | undefined;
-    failureCode: CuaActorLabErrorCode | undefined;
+    failureCode: CuaActorStudyErrorCode | undefined;
     commsCatch: RunningCommsCatch | undefined;
     commsArtifactPath: string | undefined;
     receivingInboxUrl: string | undefined;

package/dist/routes/computer-use/e2b-desktop/teardown.js

@@ -7,7 +7,6 @@ import { readE2BRelease } from "../../../substrates/e2b/sandbox.js";
 import { e2bShell } from "../../../substrates/e2b/shell.js";
 import { drainCommsEvidence } from "./comms.js";
 import { finalParticipantGeometry } from "./fidelity.js";
-import { streamEvent } from "../../../lab/run-lab-homes.js";
 /**
  * Each route's own keep flag gates its own participants only: a clone.keep can never leak into a
  * local-tree participant's teardown decision, and vice versa.
@@ -150,12 +149,12 @@ export async function finishE2BDesktop(ctx, state, failed) {
         // break teardown.
         if (state.streamUrl !== undefined) {
             try {
-                await deps.onStream(streamEvent({
+                await deps.onStream({
                     type: "ended",
                     participantId: spec.planned.id,
                     recordId: spec.recordId,
                     streamId: spec.streamId,
-                }));
+                });
             }
             catch {
                 // viewer-side only; nothing to record

package/dist/routes/computer-use/external-comms.d.ts

@@ -1,4 +1,4 @@
-import type { LabCommsEmail, LabCommsExternal } from "../../lab/types.js";
+import type { StudyCommsEmail, StudyCommsExternal } from "../../study/types.js";
 import type { PreparedRunArtifactPaths } from "../../run/paths.js";
 import type { DesktopParticipantRun, ParticipantRunOutcome } from "./types.js";
 /**
@@ -10,8 +10,8 @@ import type { DesktopParticipantRun, ParticipantRunOutcome } from "./types.js";
  * Returns the warnings the drain produced; participants that received mail get `commsArtifactPath`.
  */
 export declare function drainExternalComms(args: {
-    externalCommsConfig: LabCommsExternal;
-    externalCommsEmail: LabCommsEmail;
+    externalCommsConfig: StudyCommsExternal;
+    externalCommsEmail: StudyCommsEmail;
     env: Record<string, string | undefined>;
     runPaths: PreparedRunArtifactPaths;
     participantRuns: readonly DesktopParticipantRun[];

package/dist/routes/computer-use/fanout-bundle.js

@@ -10,6 +10,7 @@ import { participantPassed, participantStatus } from "../../run/judge.js";
 import { buildRunCostSummary, desktopSpanToMinutes } from "../../run/cost-summary.js";
 import { formatParticipantPlanEntry } from "./participant-runs.js";
 import { fanoutParticipantRecords } from "./fanout-records.js";
+import { plural } from "../../run/text.js";
 /** The run's first two events: its creation and the fan-out plan. */
 function fanoutPlanEvents(args) {
     const { specs, plan } = args;
@@ -19,14 +20,14 @@ function fanoutPlanEvents(args) {
         at: args.run.createdAt,
         level: "info",
         type: "cua-lab.run.created",
-        message: `Created computer-use fan-out run for ${plan.labId} (actor ${args.descriptor.id}, ${specs.length} participants, one world each).`,
+        message: `Created computer-use fan-out run for ${plan.studyId} (actor ${args.descriptor.id}, ${specs.length} participants, one world each).`,
     });
     events.push({
         id: "event-001-fanout-plan",
         at: args.run.createdAt,
         level: "info",
         type: "cua-lab.fanout.plan",
-        message: `Fan-out plan: ${args.participantPlan.laneCount} participant(s) (${args.participantPlan.strategy}), concurrency ${args.participantPlan.concurrency}, ${args.participantPlan.waves} wave(s); session budget ${Math.round(args.participantPlan.perLaneSessionBudgetMs / 1000)}s per participant; worst-case ~${args.participantPlan.worstCaseSandboxMinutes} sandbox-minutes${args.dryRun ? " (dry-run: $0)" : ""}. Participants: ${args.participantPlan.lanes.map(formatParticipantPlanEntry).join(", ")}.`,
+        message: `Fan-out plan: ${plural(args.participantPlan.laneCount, "participant")} (${args.participantPlan.strategy}), concurrency ${args.participantPlan.concurrency}, ${plural(args.participantPlan.waves, "wave")}; session budget ${Math.round(args.participantPlan.perLaneSessionBudgetMs / 1000)}s per participant; worst-case ~${args.participantPlan.worstCaseSandboxMinutes} sandbox-minutes${args.dryRun ? " (dry-run: $0)" : ""}. Participants: ${args.participantPlan.lanes.map(formatParticipantPlanEntry).join(", ")}.`,
     });
     return events;
 }
@@ -75,7 +76,7 @@ function fanoutReview(args, streams) {
             ? `Live computer-use fan-out is running (${specs.length} participants, one world each); terminal participant evidence has not been written yet.`
             : args.dryRun
                 ? `${args.rerun ? `Rerun from ${args.rerun.sourceRunId}: ` : ""}Dry-run fan-out: ${specs.length} participants composed for ${args.descriptor.id} against ${args.appUrl}, one world each; no desktops launched, $0 spend.`
-                : `${args.rerun ? `Rerun from ${args.rerun.sourceRunId}: ` : ""}Computer-use fan-out (${specs.length} participants, one world each): ${passedParticipants}/${specs.length} participant(s) reached a terminal, engaged verdict${participants ? `: ${formatParticipantOutcomes(participants, participantEndings)}` : ""}${runTasks ? `; tasks: ${formatRunTaskFunnel(runTasks)}` : ""}.`,
+                : `${args.rerun ? `Rerun from ${args.rerun.sourceRunId}: ` : ""}Computer-use fan-out (${specs.length} participants, one world each): ${passedParticipants}/${plural(specs.length, "participant")} reached a terminal, engaged verdict${participants ? `: ${formatParticipantOutcomes(participants, participantEndings)}` : ""}${runTasks ? `; tasks: ${formatRunTaskFunnel(runTasks)}` : ""}.`,
         gaps: args.inProgress === true
             ? ["Live fan-out session is still running."]
             : args.dryRun
@@ -136,8 +137,8 @@ function fanoutFeedbackCandidates(args) {
         ? []
         : participantFeedbackCandidates({
             runId: args.run.runId,
-            scenarioId: `cua-${plan.labId}`,
-            adapterId: plan.labId,
+            scenarioId: `cua-${plan.studyId}`,
+            adapterId: plan.studyId,
             goal: redactText(specs[0].evidenceInstructions ?? specs[0].instructions),
             substrate: plan.residual.execution?.target === "local" ? "local-desktop" : "e2b-desktop",
             participants: specs.map((spec, index) => {
@@ -179,7 +180,7 @@ export function buildCuaFanoutBundle(args) {
             at: args.run.createdAt,
             level: "info",
             type: "cua-lab.fanout.rerun",
-            message: `Rerun selected ${args.rerun.selectedLaneIds.length} participant(s) from ${args.rerun.sourceRunId}: ${args.rerun.previous.map((prior) => `${prior.laneId} was ${prior.status}${prior.completionReason ? `/${prior.completionReason}` : ""}`).join(", ")}. This is a new linked run; the source run verdict is unchanged.`,
+            message: `Rerun selected ${plural(args.rerun.selectedLaneIds.length, "participant")} from ${args.rerun.sourceRunId}: ${args.rerun.previous.map((prior) => `${prior.laneId} was ${prior.status}${prior.completionReason ? `/${prior.completionReason}` : ""}`).join(", ")}. This is a new linked run; the source run verdict is unchanged.`,
         });
     }
     specs.forEach((spec, index) => {
@@ -218,12 +219,12 @@ export function buildCuaFanoutBundle(args) {
         persona: {
             id: specs[0].persona.id,
             name: `Computer-use fan-out (${specs.length} participants)`,
-            source: `study:${plan.labId}`,
+            source: `study:${plan.studyId}`,
             sourceDigest: specs[0].persona.promptDigest,
         },
         scenario: {
-            id: `cua-${plan.labId}`,
-            title: plan.title ?? `Computer-use fan-out: ${plan.labId}`,
+            id: `cua-${plan.studyId}`,
+            title: plan.title ?? `Computer-use fan-out: ${plan.studyId}`,
             // Redacted at write time, like every other raw-text surface in the bundle. Participant records are
             // digest-only by design, but scenario.goal keeps one participant's composed instructions verbatim,
             // and an adopter whose authored participant text must name a runtime world URL (an inbox on a route
@@ -233,7 +234,7 @@ export function buildCuaFanoutBundle(args) {
             //
             // The instructions the model actually receives are untouched; only the persisted copy changes.
             goal: redactText(specs[0].evidenceInstructions ?? specs[0].instructions),
-            source: `study:${plan.labId}`,
+            source: `study:${plan.studyId}`,
             sourceDigest: specs[0].persona.promptDigest,
         },
         lifecycle: [

package/dist/routes/computer-use/fanout-records.js

@@ -1,6 +1,6 @@
 // Each participant's records in a computer-use fan-out bundle: its simulation, its stream, its
 // subject provenance event and the events of its outcome (session, geometry warnings, phase trail).
-import { participantAssignment } from "../../lab/participant-assignment.js";
+import { participantAssignment } from "../../study/participant-assignment.js";
 import { participantEvent, participantRecord, participantStream, } from "../../run/participant-records.js";
 import { declaredScreenForRender } from "../../substrates/e2b/desktop-geometry.js";
 import { describeSubjectState, phaseEventIdSuffix, publicSafeAppUrlLabel } from "./bundle-parts.js";
@@ -66,7 +66,7 @@ function fanoutParticipantRecord(args, spec, index, view) {
     const { outcome, publicTargetUrl, session, status, reason } = view;
     return participantRecord(spec, index + 1, {
         personaId: spec.persona.id,
-        scenarioId: `cua-${plan.labId}`,
+        scenarioId: `cua-${plan.studyId}`,
         status,
         streamKind: "browser",
         mode: "browser-sim",

package/dist/routes/computer-use/live-flush.d.ts

@@ -2,7 +2,7 @@ import type { ActorTokenUsage, ActorTraceItem } from "../../actors/contract.js";
 import type { CuaLiveMetadata } from "../../actors/computer-use/loop.js";
 import { type ObserverResult } from "../../observer/render.js";
 import type { RunBundle } from "../../run/bundle.js";
-import type { RunLabHomes } from "../../lab/run-lab-homes.js";
+import type { RunStudyHomes } from "../../study/run-study-homes.js";
 import type { DesktopParticipantRun } from "./types.js";
 export interface LiveTraceFlush {
     /** Record a participant's recorded-so-far items; the bundle rewrite follows on the flush
@@ -32,9 +32,9 @@ export declare function startLiveTraceFlush(args: {
  * the run. The URLs carry auth, so they live in memory and on the Observer result, never in run
  * artifacts.
  */
-export declare function trackRuntimeStreams(onStream: RunLabHomes["onStream"] | undefined): {
+export declare function trackRuntimeStreams(onStream: RunStudyHomes["onStream"] | undefined): {
     /** Records each stream for the Observer, after the caller's onStream. */
-    onStream: NonNullable<RunLabHomes["onStream"]>;
+    onStream: NonNullable<RunStudyHomes["onStream"]>;
     /** From now on, keep this live Observer's stream list current. */
     showIn: (observer: ObserverResult & {
         ok: true;

package/dist/routes/computer-use/live-phase.d.ts

@@ -5,7 +5,7 @@ import type { CuaParticipantsSetup, CuaRunSetup } from "./setup.js";
  * Runs the participants (a dry run runs none). Returns the refusal when real email setup fails, or
  * the participant outcomes and the warnings the finish records.
  */
-export declare function runLabParticipants(setup: CuaRunSetup, participants: CuaParticipantsSetup): Promise<{
+export declare function runStudyParticipants(setup: CuaRunSetup, participants: CuaParticipantsSetup): Promise<{
     ok: false;
     result: {
         automaticAnalysis?: import("../../analysis/job.js").AutomaticAnalysisOutcome;
@@ -13,7 +13,6 @@ export declare function runLabParticipants(setup: CuaRunSetup, participants: Cua
         schema: typeof import("../../run/study-result.js").STUDY_RESULT_SCHEMA;
         route: "computer-use";
         studyId: string;
-        labId: string;
         ok: boolean;
         cwd: string;
         actor: string;
@@ -40,7 +39,7 @@ export declare function runLabParticipants(setup: CuaRunSetup, participants: Cua
         diagnostics?: import("./diagnostics.js").CuaDiagnostics;
         warnings: string[];
         error?: {
-            code: import("./types.js").CuaActorLabErrorCode;
+            code: import("./types.js").CuaActorStudyErrorCode;
             message: string;
         };
         runId: string;

package/dist/routes/computer-use/live-phase.js

@@ -2,7 +2,7 @@
 // receiving, the participants themselves, and the drain of an operator-hosted inbox.
 import { prepareReceivingRun } from "../../comms/receiving-runtime.js";
 import { liveObserverResult } from "../../observer/live.js";
-import { pricedModel } from "../../lab/plan-base.js";
+import { pricedModel } from "../../study/plan-base.js";
 import { runAllCuaParticipants } from "./participant-execution.js";
 import { startLiveTraceFlush } from "./live-flush.js";
 import { drainExternalComms } from "./external-comms.js";
@@ -12,7 +12,7 @@ import { participantSubjectEnv } from "./types.js";
  * Runs the participants (a dry run runs none). Returns the refusal when real email setup fails, or
  * the participant outcomes and the warnings the finish records.
  */
-export async function runLabParticipants(setup, participants) {
+export async function runStudyParticipants(setup, participants) {
     const { plan, input, cwd, streams, descriptor, run } = setup;
     const { participantRuns, participantPlan, bundleBase } = setup;
     const { env, knownSecretValues, deps, liveTrace, externalComms, inProgress, fail } = participants;
@@ -22,7 +22,7 @@ export async function runLabParticipants(setup, participants) {
     const { runId, paths: runPaths } = run;
     // A live run writes what it is doing as it does it, whether or not anyone is currently watching.
     // This is not gated on `options.onObserverReady` (the interactive Observer callback): gated, a
-    // run launched by an agent (`lab run --json`), detached, or from the terminal surface records
+    // run launched by an agent (`run --json`), detached, or from the terminal surface records
     // nothing at all until it completes, and anything asking "what is this participant doing right
     // now" got silence for the whole run. Who reads the evidence is not the run's business; the
     // callback below stays conditional, the writing does not.

package/dist/routes/computer-use/local-tree-pack.d.ts

@@ -1,12 +1,12 @@
-import type { LabConfig } from "../../lab/types.js";
+import type { StudyConfig } from "../../study/types.js";
 import { type LocalTreeArchive } from "../../subject/local-tree-archive.js";
-import type { LabDeps } from "../../lab/lab-deps.js";
+import type { StudyDeps } from "../../study/study-deps.js";
 /**
  * Pack the working tree for a local-tree run and report what left the host on stderr, by counts
  * and digest only, never paths or file names.
  */
-export declare function packRunLocalTree(deps: LabDeps, config: {
-    readonly subject: Pick<LabConfig["subject"], "localTree">;
+export declare function packRunLocalTree(deps: StudyDeps, config: {
+    readonly subject: Pick<StudyConfig["subject"], "localTree">;
 }, cwd: string): Promise<{
     archive: LocalTreeArchive;
     buffer: ArrayBuffer;
@@ -14,7 +14,7 @@ export declare function packRunLocalTree(deps: LabDeps, config: {
 /**
  * Default local-tree packing implementation: createLocalTreeArchive(root, opts) on the host,
  * then a single read of the produced archive file into an ArrayBuffer for upload. The DI seam
- * (LabDeps.packLocalTree) overrides this in deterministic tests so they never require
+ * (StudyDeps.packLocalTree) overrides this in deterministic tests so they never require
  * tar/git.
  */
 export declare function defaultPackLocalTree(args: {

package/dist/routes/computer-use/local-tree-pack.js

@@ -24,7 +24,7 @@ export async function packRunLocalTree(deps, config, cwd) {
 /**
  * Default local-tree packing implementation: createLocalTreeArchive(root, opts) on the host,
  * then a single read of the produced archive file into an ArrayBuffer for upload. The DI seam
- * (LabDeps.packLocalTree) overrides this in deterministic tests so they never require
+ * (StudyDeps.packLocalTree) overrides this in deterministic tests so they never require
  * tar/git.
  */
 export async function defaultPackLocalTree(args) {

package/dist/routes/computer-use/local-vm.d.ts

@@ -1,23 +1,23 @@
-import type { InternalRunLabOptions } from "../../run-lab.js";
-import type { LabConfig } from "../../lab/types.js";
+import type { InternalRunStudyOptions } from "../../run-study.js";
+import type { StudyConfig } from "../../study/types.js";
 import type { LocalVmInput } from "./types.js";
 import { type LocalFirecrackerAssets } from "../../substrates/local/firecracker-desktop.js";
-type LocalVmRunOptions = InternalRunLabOptions & {
-    config: LabConfig;
+type LocalVmRunOptions = InternalRunStudyOptions & {
+    config: StudyConfig;
     assets?: LocalFirecrackerAssets;
     signal?: AbortSignal;
 };
 /** A local browser study's bindings, and the cleanup of what its participants started. */
 export interface LocalVmRun {
-    /** The caller's runLab options, with the account provider when the study runs one. */
-    readonly options: InternalRunLabOptions;
+    /** The caller's runStudyWith options, with the account provider when the study runs one. */
+    readonly options: InternalRunStudyOptions;
     /** The desktop, the analysis gate and the abort signal the computer-use run takes. */
     readonly localVm: LocalVmInput;
     close(): Promise<void>;
 }
 /**
- * Local desktop/provider composition: the options the lab is planned and run with. It throws before
- * anything starts on a lab the local study cannot run.
+ * Local desktop/provider composition: the options the study is planned and run with. It throws before
+ * anything starts on a study the local VM cannot run.
  */
 export declare function prepareLocalVmRun(options: LocalVmRunOptions): LocalVmRun;
 export {};

package/dist/routes/computer-use/local-vm.js

@@ -152,8 +152,8 @@ function accountProvider(state) {
     };
 }
 /**
- * Local desktop/provider composition: the options the lab is planned and run with. It throws before
- * anything starts on a lab the local study cannot run.
+ * Local desktop/provider composition: the options the study is planned and run with. It throws before
+ * anything starts on a study the local VM cannot run.
  */
 export function prepareLocalVmRun(options) {
     const config = localBrowserDefaults(options.config);

package/dist/routes/computer-use/participant-desktop.d.ts

@@ -1,6 +1,6 @@
 import type { CuaExecutor } from "../../actors/computer-use/loop.js";
 import type { DesktopParticipantRun, ParticipantRunOutcome } from "./types.js";
-import type { LabCommsEmail, LabCommsRecipient } from "../../lab/types.js";
+import type { StudyCommsEmail, StudyCommsRecipient } from "../../study/types.js";
 import type { PreparedOutputRoot } from "../../run/contained-output.js";
 /** A prepared desktop supplies only participant input/observation and its inbox location. */
 export interface ReadyParticipantDesktop {
@@ -41,8 +41,8 @@ export type ParticipantDesktopFactory = (run: DesktopParticipantRun, warnings: s
 /** The participant's addressed comms recipient, when one exists: the gate and the address source for the
  *  inbox instruction. A participant told to check an inbox it can never receive into would stall,
  *  so no addressed recipient means no instruction. */
-export declare function inboxRecipientFor(commsEmail: LabCommsEmail, participantId: string): LabCommsRecipient | undefined;
+export declare function inboxRecipientFor(commsEmail: StudyCommsEmail, participantId: string): StudyCommsRecipient | undefined;
 /** True when a participant has a declared comms recipient with an address, so the drain can actually match the
  *  mail the persona will be told to read. Gates the inbox instruction to participants that can receive
  *  mail; a participant told to check an inbox it can never receive into would just stall. */
-export declare function participantHasInboxRecipient(commsEmail: LabCommsEmail, participantId: string): boolean;
+export declare function participantHasInboxRecipient(commsEmail: StudyCommsEmail, participantId: string): boolean;

package/dist/routes/computer-use/participant-desktop.js

@@ -1,4 +1,4 @@
-import { recipientParticipantId } from "../../lab/parse/comms.js";
+import { recipientParticipantId } from "../../study/parse/comms.js";
 /** The participant's addressed comms recipient, when one exists: the gate and the address source for the
  *  inbox instruction. A participant told to check an inbox it can never receive into would stall,
  *  so no addressed recipient means no instruction. */

package/dist/routes/computer-use/participant-execution.js

@@ -309,7 +309,7 @@ export function toParticipantResult(spec, outcome, subject, dryRun) {
                                     ? `Computer-use session ended with a harness error: ${session.reason}`
                                     : session?.status !== "passed"
                                         ? `Computer-use session ended with ${session?.status ?? "unknown"}: ${session?.reason ?? "no terminal reason"}`
-                                        : "Computer-use lab did not produce a terminal session."),
+                                        : "The computer-use run did not produce a terminal session."),
                 },
             }),
     };

package/dist/routes/computer-use/participant-model.js

@@ -3,7 +3,7 @@
 // finished session. The desktop side stays with the participant runner.
 import { startClaudeSession } from "../../actors/local-agent/claude-session.js";
 import { createLocalAgentProvider } from "../../actors/local-agent/cli.js";
-import { pricedModel } from "../../lab/plan-base.js";
+import { pricedModel } from "../../study/plan-base.js";
 import { estimateActorCostForExecution, round6 } from "../../run/pricing.js";
 import { redactText } from "../../evidence/redaction.js";
 import { createRestrictedCodexParticipant, } from "../../actors/codex/restricted-participant.js";
@@ -54,7 +54,7 @@ export async function startParticipantModel(spec, deps, executor) {
         // started with no memory of the last. HUMANISH_LOCAL_AGENT_ONE_SHOT=1 keeps that path
         // reachable as a measurement switch: MemTrapBench (2026-08) reports memory frameworks
         // degrading agent performance by 10-40% on some tasks, so "remembers" has to be measured
-        // against "does not" on the same lab, not assumed. The trace records which one ran.
+        // against "does not" on the same study, not assumed. The trace records which one ran.
         const oneShot = env.HUMANISH_LOCAL_AGENT_ONE_SHOT !== undefined &&
             env.HUMANISH_LOCAL_AGENT_ONE_SHOT !== "" &&
             env.HUMANISH_LOCAL_AGENT_ONE_SHOT !== "0";
@@ -208,7 +208,7 @@ export async function closeParticipantModel(model, warnings) {
 /** Prices the finished session's tokens onto its trace, writes the trace, and warns on raw screenshots. */
 export async function recordParticipantTrace(spec, deps, session, warnings) {
     // Per-participant model-token cost estimate, attached to the trace before it is persisted (the model
-    // id is authoritative here: provider.version). Kept at the lab boundary so the pure loop
+    // id is authoritative here: provider.version). Kept at the study boundary so the pure loop
     // never depends on the operator rate table. estimateActorCost declares absent (null) for an
     // unknown rate / missing usage rather than guessing.
     session.trace.estimatedCost = estimateActorCostForExecution(session.trace.tokenUsage, session.trace.ids.model, session.trace.executionProfile);