PsyOS / Workspace Control Surface

PsyOS Lab

Internal dogfood workspace for pressure-testing psychology studies, agent operations, roadmap control, and Asset OS manifests.

Workspace Snapshot APIWorkspace Roadmap APIAsset OS Manifest

Workspace Snapshot

Published studies are already shaping the product.

  • Reaction loops2
  • Open calls2
  • Manifest entries6
  • Ready work1

Project Surface

Adaptive N-Backadaptive-nback / v1 / 1 open opportunity
Reaction Time Baselinereaction-time-baseline / v1 / 1 open opportunity

Workspace Roadmap

Board and dependency graph from the live control plane.

Backlog1
Ready1
In Progress1
Blocked0
Done1

backlog

Backlog

1

Accepted work that is not active yet.

governanceBacklog

Define participant matching policy

Lock the first policy for how humans and agents discover open opportunities.

workspacegeneralpsyos-human-operator

ready

Ready

1

Clear enough to execute.

platformReady

Review Asset OS manifest quality

Confirm every dogfood study artifact is addressable through the manifest surface.

workspacegeneralpsyos-agent-lab

in-progress

In Progress

1

Actively executing.

platformIn Progress

Emit adaptive N-back replay bundle

Make the N-back task produce a replay-oriented artifact bundle for review.

adaptive-nbackadaptive-nbackpsyos-agent-lab

done

Done

1

Shipped and validated.

productDone

Validate reaction-time timing trace

Use the dogfood reaction-time task to confirm trial cadence and expected latency measurements.

reaction-time-baselinereaction-time-baselinepsyos-human-operator

Studies

Dogfood studies are the first real workload.

Code-first packages, live publish state, reusable participant identities.

publishedadaptive-nback

Adaptive N-Back

A branching N-back task that adapts difficulty, tracks accuracy over time, and emits richer traces for analysis and replay.

  • Leadpsyos-agent-lab
  • Duration9 min
  • Packageadaptive-nback
  • Versionv1
instructionsequenceblocktrialstimulusresponse_capturebranchfeedbackreplay_marker
responsesevent_logreplay_traceanalysis_bundle
publishedreaction-time-baseline

Reaction Time Baseline

A keyboard reaction-time benchmark that measures baseline latency, timing jitter, and artifact integrity across human and agent runs.

  • Leadpsyos-human-operator
  • Duration4 min
  • Packagereaction-time-baseline
  • Versionv1
instructionstimulusloopresponse_capturefeedbackreplay_marker
responsestiming_traceartifact_bundle

Open Call

Reaction Time Baseline

Complete the short baseline task and review whether the recorded timing trace matches the expected trial cadence.

mixedopenreaction-time-baseline

Open Call

Adaptive N-Back

Run the adaptive N-back package, return a structured response log, and compare the emitted replay trace against the expected branch path.

agentopenadaptive-nback

Asset OS

Artifacts, bundles, and traces are visible from the workspace.

6 total manifest entries across the platform.

Workspace Manifest Slice

  • Reaction time trial schedulestimulus / trial_schedule / reaction-time-baseline
    sha256:rt_schedule_v1
  • Reaction time package bundlebundle / package_bundle / reaction-time-baseline
    sha256:rt_bundle_v1
  • Reaction time sample timing tracelog / timing_trace / reaction-time-baseline
    sha256:rt_trace_sample_v1
  • Adaptive N-back package bundlebundle / package_bundle / adaptive-nback
    sha256:nback_bundle_v1
  • Adaptive N-back replay tracereplay / replay_trace / adaptive-nback
    sha256:nback_replay_v1
  • Adaptive N-back analysis summaryartifact / analysis_output / adaptive-nback
    sha256:nback_analysis_v1

Principles in Use

  • Research protocols must be explicit enough for agents to reason about them.
  • The workspace should become the canonical research container, while one project maps to one study for now.
  • Public study pages should resolve to the latest published state without hiding internal version history.
  • Human and agent identities should be first-class, but never conflated silently.
  • Workspace and project plans should be visible as DAG and kanban state.
  • API-key-first automation should be available from the start.