A connected research workbench
Move from a satellite-link question to a quantum-state analysis, a memory study or an independent reference calculation. PF connects supported operations and retains the evidence linking each study to its inputs.
Built on the r34 implementation
The scientific registry defines 15 bounded study families. Some execute installed libraries, some use dedicated workers, and others analyze supplied records. These are specific integration routes, rather than unrestricted access to every upstream API.
Runtime availability, workflow applicability and successful validation are separate questions. The workbench checks prerequisites and records why a companion is available—or why a study cannot run.
01 / Quantum systems
Use complementary calculations to inspect a quantum model from more than one direction.
Execute density-matrix circuits for mission polarization channels, seeded Pauli measurements and component benchmarks. Inspect fidelity, purity, probabilities and state reconstruction.
Compare exact model outputs with finite-shot estimates and test sensitivity to declared noise.
These routes simulate quantum systems locally. They do not submit jobs to IBM quantum hardware.
Run two-qubit state and one-qubit process tomography. A mission companion can consume an actual saved density matrix, then prepare and sample it with Aer.
Review raw counts, reconstruction choices and an independent Born-rule inversion check.
The companion generates new simulated counts. A saved state does not identify a channel or provide automatic state-preparation and measurement (SPAM) calibration.
Study bounded two- and four-qubit GKSL dynamics, commuting HEOM and saved mission-memory evolution. Relativistic link studies also use a declared bosonic receiver-mode channel.
Cross-check memory decay against PF’s Kraus implementation and explore model-dependent receiver effects.
Supported operations have explicit model and system-size limits; arbitrary many-body models are not exposed.
Execute stationary fiber/router reservation studies through the native network simulator and supported published tutorial configurations.
Investigate reservation behavior and component assumptions in a declared quantum-network model.
The fiber model is a separate study. Orbital free-space losses are not automatically converted into fiber attenuation.
02 / Orbits & spacetime
Relate orbital predictions, clock models and photon propagation while preserving each calculation’s coordinate and model assumptions.
Retrieve GP orbital elements, including supported JSON and TLE inputs, and propagate selected satellites at requested UTC epochs. Astropy supports time and coordinate transformations.
Study changing quantum-link geometry and compare tracked-orbit and numerical-orbit predictions.
Live tracking means current-time propagation with periodic data refresh. It is not a spacecraft telemetry stream.
Compare bounded two-body and J2 propagation and frozen Vallado SGP4 reference vectors. A companion can derive reference-orbit elements from a saved mission state.
Examine force-model and implementation differences using explicitly matched reference assumptions.
The mission-derived reference is reoriented and begins at pericentre/J2000. It does not reproduce the original mission trajectory or epoch.
Evaluate Schwarzschild metric, curvature, static-clock and radial-null reference quantities, including selected radii from saved static-GR missions.
Cross-check metric and clock calculations against a separate symbolic implementation.
An exterior, spherical, stationary reference; it is not general Einstein-equation evolution or the original moving photon path.
Use supported gauge-wave or registered evolution workflows and imported numerical-relativity diagnostics. A bounded metric-line adapter supports photon-link studies.
Inspect evolved fields, convergence and constraint evidence alongside explicitly compatible link calculations.
Runtime or imported outputs must be supplied. General 3D AMR satellite-metric injection and full upstream thorn coverage are not implemented.
03 / Evidence & engineering
Bring declared observations, uncertainty ranges and engineering references into focused companion studies.
Infer a common memory T2 from declared Bell-parity counts with PyMC and calculate Sobol sensitivity indices with SALib for the supported pure-dephasing model.
Explore which assumptions dominate predictions and how evidence constrains a memory model.
The observations and parameter distributions must be supplied for that model; this is not a general mission posterior.
Replay managed detection-event CSV files and decode supported Swabian recordings through the installed vendor FileReader SDK.
Compare declared coincidence windows and timing offsets against supplied detection records.
Recorded-file analysis, not live instrument control. Genuine vendor-recording decoding remains unverified in the retained release evidence.
Execute the pinned Dahlquist reference FMU with declared settings and numerical comparison outputs.
Check a reproducible model-exchange calculation and its integration behavior.
This route supports the pinned reference FMU, not arbitrary FMU upload or automatic spacecraft-controller modeling.
Execute the fixed clock-relaxation ODE or supported Simulink clock model through a separately provisioned licensed installation.
Compare a defined timing model across engineering runtimes.
Requires MATLAB, the relevant products and licenses. MATLAB Runtime alone is insufficient; licensed scientific execution remains unverified in the retained release evidence.
Reproduce specified clock, link, detector and quantum reference calculations using frozen records, declared equations and PF production protocols.
Compare an implementation with a named reference while preserving its assumptions and data lineage.
Published printed targets and equation-consistency checks are distinguished. Agreement is not experimental or commercial qualification.
From one result to the next study
Select a completed mission and a saved epoch. Where required fields exist, PF can derive tomography, memory, Schwarzschild, orbit-reference or spectral-channel studies from actual parent outputs.
Each child retains the parent artifact, input hashes, derived settings, extra assumptions and its own results. Contextual references are labeled separately and do not replace the parent prediction.
04 / Your research environment
OpenAI Responses, OpenAI-compatible chat endpoints, DeepSeek, Google Gemini and Anthropic / Claude connect through the same private proposal, validation and activation workflow. BYOK allows personal provider credentials.
Understand provider and data boundariesBYOS supports S3-compatible storage and an outbound NAS / folder connector. Operator deployments can connect AWS Batch and S3, Cognito authentication, Stripe billing and SMTP email where configured.
Explore deployment connectionsThese services support execution, access and data handling; they are not scientific validation engines. Availability depends on deployment, installed dependencies, account access and licenses. Product names identify integration routes, not endorsements or partnerships.
Inspect the implementation
The catalogue was checked against the corrected r34 source and its capability registry. New gallery examples exercise satellite replay and Qiskit Aer; they do not claim that every optional external runtime was executed during this website update.