Skip to main content
​
Project-scoped ECM fitting, PyBaMM v26.4.0 support, SDK skills plugin, strict upload validators

​
Project-scoped ECM fitting

Equivalent-circuit-model fitting is now available directly inside a project. Pick any measurement already attached to one of the project’s cells (or upload a fresh CSV), preview the trace, run the fit, and save the result back to the project as a new model with one click. The new flow adds a model-vs-data plot, an error plot, a fitting-progress indicator, an OCV toggle helper, and auto-scrolls to the results when the fit completes. The standalone ECM demo continues to work for unauthenticated users.

​
PyBaMM v26.4.0 support

The simulation engine has been updated for PyBaMM v26.4.0, which introduces a dedicated Rest step class in place of the old zero-current step and renames the MSMR electrode-capacity parameters from short keys (e.g. Q_n_1, Q_p_1) to descriptive strings (e.g. “Negative electrode host site occupancy capacity (1) [A.h]”). Existing experiments and MSMR parameter sets continue to parse correctly under the new release.

​
SDK skills as a Claude Code plugin

The Ionworks Python SDK now ships with a set of installable Claude Code skills covering api-discovery, cell-data, data-upload, simulations, pipelines, and projects-and-studies. Each skill documents the relevant sub-clients, method signatures, and the create-or-get patterns so an AI assistant can drive the SDK against the current API surface. Contact Ionworks for access.

​
Strict upload validators in the Python API

CellMeasurementClient.create and create_or_get now run optional client-side validators before sending data to Studio so processing bugs surface as a local error instead of a confusing upload result. A hard time-gap check (>5 h between consecutive rows) runs by default; passing voltage_window enables a voltage-continuity check (catches chronological misordering from a faulty data split), and passing rated_capacity enables a consecutive same-direction full-step check plus a soft warning for any step whose capacity exceeds 5× rated.
Improvements
  • Measurement chart controls collapse into a compact header summary; per-axis variable selectors now allow adding and removing series dynamically with a unified two-column layout.
Fixes
  • Study visualization no longer collapses to an empty state when switching between studies whose experiment-type tabs differ — the filter resets to All when the previous tab isn’t present.
  • Users with a stale session can recover by signing out via the sign-in page instead of being stuck on the redirect.
  • The empty cells page now uses the correct organization-level permission check, so users who can create cells no longer see a misleading “contact a project admin” message.
  • Pipeline element details view no longer triggers rapid duplicate fetches for Data Fit and Validation configs.
Improvements
  • Simulations now compile faster by default, speeding up repeated runs that share the same model.
  • Simulation objectives with custom mesh or spatial-method settings (such as CycleAgeing) now serialise and reload correctly.
Fixes
  • Time-monotonicity validation error message now reports the correct two adjacent values (the previous off-by-one showed the values just before the actual violation).
Fixes
  • Sign-convention detection no longer misclassifies low-current pulse datasets as all-rest; the rest threshold now scales relative to the 95th-percentile current so violations are caught on small-amplitude data.
​
Python SDK resource clients, validation initial conditions, capacity calculation fix

​
Python SDK resource clients

The Python SDK gains typed sub-clients for managing Studio resources programmatically. You can now create, list, retrieve, update, and delete projects, models, parameterized models, studies, protocols, and optimizations directly from the Ionworks client — for example, client.project.list(), client.model.get(id), or client.study.create(...). A new “Managing resources” section in the Python API docs walks through each sub-client with usage examples, and requests for Projects, Models, and Studies no longer require your organization to be specified separately — it’s resolved from your account automatically.

​
Validation initial conditions and UX polish

Validations now let you choose how the simulation is initialized: auto-detect from the measurement, a custom starting voltage, or a custom state of charge. Each validation row in the Report tab can be removed individually with a confirmation dialog, and a dedup bug that caused two measurements sharing the same initial voltage to overwrite each other’s drive cycle has been fixed. Validation error plots also now render with grid lines for easier reading.

​
Capacity calculation fix for multi-electrode cells

A capacity calculation was reading a non-canonical parameter name for the number of electrodes connected in parallel, causing the maximum concentration to be over-computed by a factor of N when users set the correct key. The fix aligns all lookups with PyBaMM’s canonical parameter name so multi-electrode cells parameterize correctly.
Improvements
  • Studies with simulations of a single experiment type now auto-select that type so the visualization view is immediately accessible.
  • Failed variable-evaluation jobs now surface an error state with a Retry button instead of spinning indefinitely.
  • Studies list and parameterized models list now support pagination.
Fixes
  • Fixed a crash when rendering time-series measurements with thousands of short steps (e.g. HPPC pulse protocols with ~1500 steps) — the dashboard can now load these measurements.
  • Fixed coupled custom variables being silently dropped when their resolution failed, which could remove convenience variables like “Temperature [degC]” from simulation outputs.
  • Fixed an issue where an overly large page size on the models list could cause an error.
Improvements
  • Bumped PyBaMM to 26.3.1, which falls back to a base model with a warning when a custom model references a package that isn’t installed in the run environment.
  • Improved reliability of OCP-related fitting objectives so they work correctly in Validation pipelines.
Fixes
  • Fixed an issue where a data fit submitted with a single objective could fail unexpectedly.
Improvements
  • Added typed sub-clients for Models, Studies, Optimizations, Projects, Protocols, and ParameterizedModels, replacing raw HTTP requests for resource management.
  • Requests no longer require your organization to be specified in resource paths — it’s resolved from your account automatically.
  • Simulation polling now propagates errors immediately instead of swallowing them, and uses typed job status objects with clearer completion/failure fields.
Fixes
  • Fixed the simulation job poll endpoint, which was pointing at an incorrect route.
Fixes
  • Fixed a sign-detection bug in the ECM validator where a symmetric current profile (charge / discharge / charge returning to the same net charge) was incorrectly rejected despite having a non-zero SOC span.
​
Async optimization, study validation reports, cell data filtering

​
Asynchronous optimization evaluation

Optimizers now support asynchronous evaluation for both local and distributed workloads, keeping all workers saturated without waiting for generation boundaries. New convergence criteria — including a patience counter and population diversity guards — prevent premature stopping.

​
Study validation reports

Studies now include a validation workflow for comparing simulation results against experimental measurements. A new Report tab displays overlay plots of simulated vs. measured data along with quantitative error metrics (RMSE, MAE), and a Data tab lets you manage which measurements are paired with which simulations.Find out more →

​
Cell data filtering and ordering

All cell list endpoints (specifications, instances, and measurements) now accept filter and ordering parameters — including name, creator email, date ranges, and sort direction. The Python SDK exposes matching keyword arguments on each list() method, and the frontend measurement table uses server-side filtering.
Improvements
  • Pipeline and cell measurement tables now support server-side sorting.
  • Added an error screen with auto-retry when the initial user info request fails.
  • Removed SNES optimizer option from the frontend.
Fixes
  • Fixed cross-site scripting issues in the notifications drawer and job progress display.
  • Improved error message consistency across the app.
  • Eliminated slow file-listing calls on file-measurement endpoints by caching filenames, reducing response times from ~10 s to under 100 ms.
Improvements
  • Added dedicated algorithm test suites for CMA-ES, XNES, and Nelder-Mead.
  • Improved convergence for CMA-ES and XNES by preferring the evaluated distribution mean when it outperforms the best sample.
Fixes
  • Fixed a flaky deterministic design optimization test.
Improvements
  • Added filter and ordering keyword arguments to CellSpecificationClient.list(), CellInstanceClient.list(), and CellMeasurementClient.list().
  • download_files(filenames=) now skips the listing round-trip; new get_file() method for single-file fetch.
Improvements
  • Added temperature column recognition for Biologic files.
​
Measurement types, performance improvements, total time termination, parameterized model enhancements

​
Measurement types for non-time-series data

Cell measurements now support a measurement_type field with three values: time_series (the default), properties for key-value metadata such as checkpoint measurements, and file for arbitrary files such as images, PDFs, or numpy arrays. Each type has its own creation and retrieval flow, and list endpoints can filter by type.Find out more →

​
Performance improvements

Pipeline creation is now significantly faster, and the platform handles concurrent load with improved throughput. Models, parameterized models, studies, and optimizations now use server-side pagination and filtering, eliminating the need to load entire collections into memory.

​
Total time termination condition

Protocol steps can now use total_time as a built-in termination condition, ending a step after a specified duration of total elapsed experiment time rather than just step time.Find out more →

​
Parameterized model improvements

The parameterized model detail view has been consolidated into a single page with tabs, replacing the previous multi-page layout. Models can now be downloaded as a zip archive containing the full configuration, and initial temperature is now correctly accounted for in design optimization objectives for temperature-dependent studies.Find out more →
Improvements
  • Cell specification list now displays as a sortable data table instead of cards.
  • Added a UI warning when a simulation produces no data because all protocol steps were skipped.
  • Removed hardcoded convenience variables from the variable evaluator and improved the evaluate variables UX.
  • All update requests now use partial update semantics.
  • Standardized all duplicate resource errors to return HTTP 409 with structured error details.
  • Replaced signed URL downloads with HTTP 307 redirects for simpler, faster file access.
  • Flattened cell measurement composite response shapes for a more consistent API surface.
  • Design optimizations now default to the Differential Evolution algorithm.
Fixes
  • Fixed optimization deletion hitting the wrong API endpoint.
  • Fixed clone optimization sending a reverted request body.
  • Fixed optimization list page sending redundant API requests.
Improvements
  • Initial temperature is now correctly parsed for temperature-dependent optimization studies.
Fixes
  • Fixed fitted parameter dictionaries occasionally including an extraneous internal entry.
  • Fixed validation to correctly handle certain transformed parameters.
  • Fixed initial SOC not being passed to EIS frequency-domain simulations.
Improvements
  • Added support for arithmetic expressions with input parameters when converting protocols for simulation (e.g. 1000 / input["Total cells"] now produces correct results).
Fixes
  • Fixed initial SOC not being applied in EIS frequency-domain simulations.
Improvements
  • Replaced signed URL downloads with direct HTTP 307 redirects for measurement file access.
  • Flattened cell measurement response shapes for simpler data access.
  • Added measurement_type field to measurement models for non-time-series data.
  • All update methods now use partial update semantics.
  • Improved loading of studies and parameterized models with large numbers of simulations via server-side pagination.
Improvements
  • Faster, more portable local caching.
  • Added Latin-1 encoding support for Neware CSV files.
​
ECM parameterization, resistance scaling, custom variables, and BaSyTec reader

​
ECM parameterization

A new ECM parameterization page lets you upload cycling data, select 0–5 RC pairs with a live circuit diagram, and view fitted parameters including OCV and R0. Results can be downloaded as CSV.Find out more →

​
Adjust resistance in protocol simulations

The protocol simulator now uses Equivalent Circuit Models (ECM) instead of SPM for faster, more robust cycler protocol simulations. ECM parameter sets are available for all six supported chemistries, and a new resistance scaling parameter in the UI lets you adjust the overall cell resistance before running a simulation.Find out more →

​
Custom variables on models

You can now define custom derived quantities on models — such as electrode potentials or temperature in different units — as PyBaMM expressions stored in the model configuration. These custom variables are evaluated at simulation time and included in result data, making it easy to track application-specific metrics across design sweeps.Find out more →

​
BaSyTec battery cycler reader

Data processing now supports BaSyTec CSV exports (CTS, X50 series). The reader auto-detects BaSyTec files by header signature, parses HH:MM:SS.sss timestamps (including values above 24 hours), applies the correct current sign convention, and extracts start times from companion _meta.txt files.Find out more →
Improvements
  • Added dark mode with a theme toggle in the dashboard header.
  • Moved optimization templates from organization scope to project scope with project-level management pages.
  • Made optimization name optional when creating optimizations.
  • Added pagination to cell specification, instance, and measurement list endpoints.
  • Added OCV traces to drive cycle simulation result plots.
  • Unified EIS column naming to Z_Re [Ohm] / Z_Im [Ohm] across simulation, measurement ingestion, and frontend rendering.
Fixes
  • Fixed LumpedSPMR model crash due to a missing internal reference.
  • Fixed an issue where datafit setup configuration could be mutated between runs.
Improvements
  • Added support for computing lithium inventory from maximum stoichiometries without requiring maximum concentration values.
  • Adopted PyBaMM’s native model and experiment serialization.
  • Added abs as a unary operation for composed metrics.
Fixes
  • Fixed thermal model using incorrect heat generation variable in ECM.
  • Fixed trailing space in electrode balancing direction parameter names.
  • Fixed validation to discover sub-parameters inside function-based parameters.
Improvements
  • Unified EIS impedance columns to Z_Re [Ohm] / Z_Im [Ohm] with SI units and standard sign convention.
Fixes
  • Fixed CCCV resolution to apply consistently to all sub-steps within a protocol.
Improvements
  • Added local data caching for cell measurements, reducing repeated downloads.
  • Added pagination support (limit/offset) to cell list methods, returning a PaginatedList with a .total attribute.
  • Improved sign detection algorithm for measurement current data.
Improvements
  • Added BioLogic plain CSV reader for non-MPR exports.
  • Added automatic derivation of Z_Re and Z_Im from Z_Mod and Z_Phase when impedance components are missing.
Fixes
  • Fixed CSV reader to skip auto-detection for explicitly mapped columns.
  • Fixed plotting bugs in data visualization utilities.
​
Protocols section, job cancellation, parameterized durations, and optimization management

​
Dedicated Protocols section with parameterized input support

Studio now has a dedicated Protocols section where you can browse, edit, and clone saved experiment templates. Each protocol detail page shows the source YAML, a tabbed UCP/human-readable preview, and automatically detected input parameters with a generated parameters_schema. This makes it easy to build reusable, parameterized protocols and simulate them with different input values.Find out more →

​
Cancel running jobs from Studio

You can now cancel pipelines, optimizations, and simulations directly from the UI. Each detail page shows a cancel button for in-progress jobs, with optimistic status updates and automatic rollback if the cancellation fails. On the backend, cancelling a parent job cascades to all child jobs.

​
Protocol termination conditions and parameterized durations

Protocol simulations now support early termination based on variable conditions — useful for ending long protocols (e.g. multi-cycle Arbin tests) after a target number of cycles. Duration and termination end conditions also accept input['...'] references, enabling fully parameterized experiments such as eVTOL design optimization with variable cruise duration.Find out more →

​
Edit and delete optimizations

Optimization name and description are now editable from the detail page, and you can delete optimizations with a confirmation dialog. The experiment-step validation toggle is also exposed as a per-objective setting, letting you control whether each objective validates against experiment steps.
Improvements
  • Added cancel buttons for running pipelines, optimizations, and simulations with cascading cancellation.
  • Added optimization editing (name and description) and delete with permission checks.
  • Exposed the validate-against-experiment-steps toggle on each optimization objective.
  • Unified table UI across optimizations, protocols, and pipelines — actions behind three-dot menus, clickable row names, no more “Actions” column header.
  • Added more plot variables to the protocol simulator results view.
  • Design parameters and experiment conditions are now preserved when rerunning simulations, and columns for differing parameters default to visible.
Fixes
  • Fixed 500 error when simulating non-UTF-8 protocol files (e.g. BioLogic .mps).
  • Fixed optimization detail page layout issues.
Improvements
  • Added support for input['...'] parameter references in protocol duration and termination end conditions.
Improvements
  • Added a way to end protocol simulations early when a variable condition is met.
  • Added support for Python-style list repetition (["step"] * N) in PyBaMM experiment strings.
  • Added support for cycle-index-based branching in Arbin protocol parsing.
Fixes
  • Fixed decoding of non-UTF-8 protocol files in the standalone simulator.
Improvements
  • Large request payloads (>512 KB) are now compressed automatically, reducing upload sizes significantly (e.g. 24.5 MB to 3.7 MB for typical serialized models).
  • Job status fields now use typed enums instead of raw strings.
​
Protocol Builder launch, stronger optimization guardrails, and better data ingestion

​
Protocol Builder is now available in Studio

Studio now includes a visual Protocol Builder with step editing, drag-and-drop ordering, YAML import/export, and backend-backed validation before upload. You can also configure global safety limits for current, voltage, and temperature directly in the builder.Find out more →

​
Optimization setup is safer and easier to reuse

You can now clone optimizations and save them as reusable templates. Optimization runs also validate that fit parameters are actually used by at least one objective model before execution, helping catch invalid configurations earlier.

​
Data ingestion and model setup got more robust

This week adds stronger cycler-data handling across formats and improves initialization behavior for fitting workflows, including more flexible parameter configuration defaults.
Improvements
  • Added pipeline edit/delete support with row actions, bulk delete, and permission checks.
  • Added a Simulations tab to parameterized model details for quicker navigation and management.
  • Added optimization cloning and “save as template” workflows.
  • Added simulation rerun actions and job resubmit support for failed jobs.
  • Added protocol YAML display on simulation result pages with input substitutions.
  • Added source metadata display for direct entry elements in pipeline details.
  • Validation views now surface partial-failure warnings instead of failing silently.
  • Timestamps across key tables now display as absolute datetimes.
  • Standalone and add-result protocol forms now hide initial SOC/temperature inputs when those values are explicitly defined in UCP global settings.
Fixes
  • Fixed validation charts so single-point series render visibly.
  • Fixed edge cases in new-user onboarding, including logout and empty-credit handling.
  • Pipeline submission now runs asynchronously to avoid blocking request handling.
Improvements
  • Added pre-run checks that reject fit parameters not used by objective models.
  • Fit parameter configs now allow omitted initial_value and compute defaults from bounds.
  • OCP balancing now supports GITT workflows with improved interpolation and safeguards.
  • Tuned optimizer defaults for more reliable convergence on real-world data.
  • Improved handling of large datasets during pulse-based fitting workflows.
  • Objective setup now supports better partial-result handling on solver failure paths.
Fixes
  • Fixed piecewise interpolation config roundtripping for 1D/2D direct-entry schemas.
  • Restored support for custom initial-state handling used by ECM and lumped SPMr workflows.
Improvements
  • Added validation to block oversized inline time series payloads (>1000 rows) with clear guidance to upload measurements and reference db: IDs.
Fixes
  • Non-idempotent API calls are no longer retried automatically.
  • Fixed current-sign normalization logic for mixed charge/discharge cycler exports that record absolute current values.
Improvements
  • Added a Gamry EIS reader with auto-detection for .dta and related impedance formats.
  • Improved multi-sheet Neware handling with better timestamp ordering and monotonic time reconstruction.
  • Numeric-string columns are now inferred and promoted more reliably during read/clean steps.
  • Added support for exporting self-contained inline configurations for local use.
Fixes
  • Fixed BioLogic reader handling for additional file variants.
​
Multi-measurement comparison, cycle aging simulations, and improved optimizer convergence

​
Multi-measurement comparison view

Select multiple measurements in the Cell Data table and click “Compare” to open a dedicated comparison page. The Time Series tab plots measurements on the same chart with configurable axes and a stack-cycles mode, while the Cycles tab shows cycle metrics like discharge capacity and coulombic efficiency across all selected measurements with optional retention mode. Filters can be applied globally or per-measurement.Find out more →

​
Cycle aging simulation template

A new cycle aging template is available for simulations, allowing you to define cycling protocols and visualize cycle-level metrics such as capacity fade and efficiency over hundreds of cycles.Find out more →

​
Improved optimizer convergence

The Differential Evolution optimizer now has better initial coverage of the parameter space and improved reproducibility. These changes improve convergence reliability across a wide range of fitting problems.
Improvements
  • Simulations can now be force-rerun while preserving the same simulation ID and study mappings.
  • The data list table now shows expanded specification columns (anode, cathode, electrolyte, case sub-fields), copy-on-click IDs, and a search-on-this-page feature.
  • EIS and dV/dQ data can now be viewed in dedicated tabs on measurement detail pages.
  • Cycler simulator jobs now provide live partial-result updates during execution.
  • Parameter sets are now dynamically discovered rather than hardcoded, so newly installed parameter sets appear automatically.
  • The API now includes a discovery endpoint to help programmatic clients understand available capabilities and schemas.
  • Cell creation no longer requires anode and cathode to be defined upfront, simplifying the model creation workflow.
Fixes
  • Fixed stale data appearing after switching organizations.
  • Fixed a bug where model options were missing from design optimization jobs.
  • Fixed PyBaMM expression parameters not being handled correctly during parameterized model creation.
Improvements
  • Composite models now support nested list/tuple options (e.g. open-circuit potential configurations).
  • Capacity calculations now correctly account for the number of electrodes in parallel.
  • Validation runs now surface simulation errors immediately instead of silently filling results with sentinel values.
  • Custom initial state support added for lumped SPMr and ECM models, accepting either a voltage string or SOC float directly.
Fixes
  • Fixed drive cycle step data losing interpolant data during roundtrip.
  • Fixed EIS simulation including unnecessary output variables.
  • Fixed NaN values in simulation results causing constraint actions (GreaterThan/LessThan) to crash instead of returning penalty values.
Improvements
  • Added an API discovery skill to help agents explore available endpoints and schemas.
Improvements
  • Data loading now accepts environment configuration for database connections.
  • Step detection now checks the last step after the first for more reliable step boundary identification.
Fixes
  • Fixed a data loading issue that could fail when cloning previously loaded data.
Fixes
  • Fixed a drive cycle temperature handling bug.
  • Added validation that safety limit minimums must be less than maximums, preventing misconfigured protocols.
  • Fixed a bug where initial SOC parameter values were not saved for later variable evaluation.
  • Fixed Maccor parser not recognizing constant-power/constant-voltage (CPCV) protocol steps.
​
Protocols in studies, dependent parameters, and faster data processing

​
Protocols in studies

Protocol simulations can now be run directly within a study, alongside standard experiments. Protocols can be uploaded from raw cycler format or Universal Cycler Protocol, with results appearing inline alongside other study results.Find out more →

​
Dependent parameters

When creating a parameterized model, any parameter value can now be set to an expression referencing other parameters. Referenced parameters are automatically detected and added to the model, enabling linked or derived parameter relationships without separate configuration.Find out more →

​
Faster step handling

Cell measurement steps are now stored more efficiently, significantly improving upload and download times for measurements with thousands of steps. Existing data is migrated transparently — no action is required.
Improvements
  • Pipeline status now shows detailed element-level job status, giving better visibility into running, pending, and completed pipeline elements.
  • Data uploads are now parallelized for faster ingestion of large datasets.
  • EIS result plots are now displayed after impedance fitting.
  • Updated OCP and EIS validation plots.
Fixes
  • Fixed simulation creation page not loading correctly.
  • Fixed an issue where files could become temporarily unavailable during large downloads.
  • Fixed a race condition where shutting down one job could inadvertently terminate concurrent jobs.
Improvements
  • Pipeline objectives can now create direct entries from Studio pipeline results using a pipeline ID.
  • OCP validation plots now include dUdQ (incremental capacity) curves.
Fixes
  • Fixed EIS fits failing when models contained interpolant-based function parameters.
  • Fixed a serialization bug where deep-copying parameter values could corrupt model state.
Improvements
  • OCP data can now be uploaded through the API.
Improvements
  • Data loading is now faster and more memory-efficient.
  • Data can now be loaded lazily — fetched only when first accessed, reducing unnecessary I/O for pipeline configurations.
  • Time series and steps can now be loaded independently, enabling efficient step filtering without downloading full time series.
  • Added support for extracting OCP from rest steps without requiring GITT labels.
  • Added Maccor CSV tab-separated file reader.
Fixes
  • Fixed dUdQ computation failing on lazily-loaded data.
  • Fixed step summarization for data without a time column.
Improvements
  • Added support for CCCV and PEIS protocols, plus anode and cathode voltage termination conditions.
Fixes
  • Fixed a bug where building a model for voltage or power control modes mutated the base model, causing “overdetermined” errors on subsequent evaluation steps.
  • Fixed a Maccor protocol variable resolution bug.
​
Initial conditions in objectives, memory-efficient streaming, and unified drive cycles

​
Initial conditions in pipeline objectives

Pipeline objectives now support a preamble mechanism for setting initial conditions — initial state of charge, temperature, and voltage — directly on each objective. Conditions are properly isolated across multiple objectives, so sweeping over different starting points no longer requires separate pipeline configurations.Find out more →

​
Memory-efficient simulation streaming

Large simulation solutions are now uploaded and downloaded in chunks, significantly reducing peak memory usage on the backend. This makes it practical to run and retrieve results from long-duration or high-resolution simulations without hitting memory limits.
Improvements
  • Cell measurement plots now support flexible x-axis selection with prioritized column options.
Improvements
  • Initial SOC can now be set from the objective options dictionary at creation time.
Improvements
  • Added a time-monotonicity validator that checks uploaded measurement time series are strictly increasing.
Improvements
  • Protocol solution metadata can now be extracted in memory, eliminating the need to write intermediate files to disk during simulation.
  • Unified drive cycle handling across Maccor and Arbin protocols, replacing the Maccor-specific “Waveform” abstraction with a general “DriveCycle” concept.
← Newer updatesOlder updates →