Customer-facing history of CATO capabilities, reliability improvements, and important behavior changes.
How to read these notes: dates group related customer-visible changes. Feature availability can depend on plan, account configuration, data type, and the controls shown in CATO.
2026.09
Major updateCurrent
September 1–21, 2026
Labs and Enterprise, two-factor authentication, a single ranked Deep Search, and a rebuilt Citations page.
Capabilities
Added Labs: shared workspaces with per-seat pricing, an organization switcher, invitations, and a Lab settings tab showing members, seats, and the Lab ID.
Added two-factor authentication with authenticator-app enrollment and encrypted backup codes; enrollment completes only after the codes are saved.
Reworked Deep Search into one ranked, meaning-based search returning up to 100 results in relevance order, with year filtering during retrieval, enrichment of identifiers and open-access links, and a preprint shown as one result with the journal article it became. Query Studio now offers only the controls the search honors.
Rebuilt the Citations page: live artifact previews, saved snapshots for deleted files, italicized journal titles, standard author formatting, BibTeX/EndNote XML/CSV exports, and project-scoped citations with history.
Saved literature reports render as finished documents and can be deleted after an in-app confirmation; generated systematic reviews trace every claim to the works they cite.
Added a four-level reasoning-effort slider (Low, Medium, High, Max), set per message with an account-level default.
Evidence extraction returns structured bundles noticeably faster with the same fields and claims.
Reliability, security, and operations
Password-protected Office files are refused at dataset upload with a plain-language explanation, and stored with a warning in the Files browser.
The legal pages were substantially rewritten (Terms, Privacy, Usage Policy, Data Processing), and their Back button returns to the app when the page opened in a new tab.
Removed em dashes across all product copy and fixed the library review flag for works whose tags arrived in a legacy shape.
2026.08
Major update
Late July – August 31, 2026
Every project gets its own computer and file browser, runs became transparent, and usage moved to rolling windows.
Capabilities
Gave every project a durable computer, named in the Workspace title bar with pinned Files, Shell, and Memory views. Files persist across conversations, and the computer can be renamed or erased.
Rebuilt Files into a full file browser: folders, drag to move, folder upload, previews for essentially any file type, and provenance on every file and figure. Data now lives in the Files browser; the separate Data door was retired.
Made project memory an editable file on the project computer, read at the start of every turn.
Added scientific viewers for 3D molecular structures, multiple-sequence alignments, and whole-slide microscopy images, drivable from chat and from slash commands such as /view, /motif, /columns, /focus, and /goto.
Chat shows its work: tool activity in the status line, a run rail tracking the declared plan of multi-step runs, evidence marks tying claims to what backs them, a context-usage meter, typeset math, and mid-task clarifying questions.
The figure editor operates on the real figure: version history, visual diff between versions, reliable text editing with keyboard shortcuts, and faithful exports.
Removed per-file upload size limits everywhere; uploads land in the folder you are browsing.
Reliability, security, and operations
Included AI usage became a rolling allowance measured over 5-hour and 7-day windows that recovers as activity ages out, replacing fixed monthly caps; the $100 tier is named Pro.
Interrupted response streams reattach instead of failing the message, and completed responses stay in place through recoveries.
Deleting an account erases its project computers and support tickets, and error reports no longer include user content.
2026.07.23
Major update
July 23, 2026
A redesigned, user-controlled memory system with stronger account and administrative reliability.
Capabilities
Separated project memory from global memory so study context and durable personal preferences have explicit scopes.
Added user controls to review, edit, pin, and delete memory entries and to control recall, automatic learning, and cross-project use.
Grounded automatic memory learning in user messages; generated findings are not promoted into memory as scientific facts.
Added memory consolidation to keep long-running accounts useful without treating repetition as evidence.
Reliability, security, and operations
Hardened administrative user, subscription, usage, support, and audit data paths.
Improved retention and referential-integrity behavior for account and operational records.
2026.07.21–22
Major update
July 21–22, 2026
Complete figure workbenches, broader scholarly discovery, and improved capacity for scientific analysis.
Capabilities
Completed figure workbenches with interactive state, artifact versions, supported editing, and export-oriented visual QA.
Expanded Literature Intelligence to 18 open scholarly sources: PubMed, PMC, Europe PMC, Unpaywall, CORE, Crossref, OpenAlex, DataCite, Semantic Scholar, arXiv, bioRxiv, medRxiv, ChemRxiv, SSRN, Research Square, OSF, Zenodo, and HAL.
Kept ClinicalTrials.gov and patent discovery as purpose-built specialized connectors.
Improved literature-result normalization, open-access resolution, and durable research-artifact handoff.
Reliability, security, and operations
Improved capacity during concurrent scientific analyses.
Fixed authenticated plan-checkout return behavior and improved availability monitoring.
2026.07.18–20
Major update
July 18–20, 2026
Durable background execution, Fast mode, restored workspace state, and end-to-end cancellation.
Capabilities
Added Fast mode for lower-latency questions, summaries, transformations, and iterative editing.
Added durable background sandbox jobs whose cards, progress, artifacts, failure state, and cancellation survive navigation and reload.
Persisted open workspace artifacts and the selected artifact version across reloads and return navigation.
Added unsaved-edit protection for history navigation, version changes, and workspace state transitions.
Restored durable literature search results across navigation and made large result restoration more responsive.
Reliability, security, and operations
Propagated cancellation from the conversation through queued and running sandbox execution.
Recovered artifacts after temporary application or storage interruptions.
Prevented duplicate composer submission while stream-recovery checks are active.
Queued concurrent sandbox work rather than reporting avoidable capacity failures.
Important notes
Fast mode uses the usage multiplier shown in the product. Standard remains the balanced default; Research remains the extended-depth mode.
Preserved prompt drafts and showed actionable banners when an account hits a limit.
Redesigned administration and hardened complete user-data deletion paths.
Important notes
Historical fixed execution-time wording has been removed from Help. CATO now governs execution according to the workload and current capacity controls.
2026.03
Feature update
March 7–24, 2026
Peer-review workflows, no-code figures, broader research data, and stronger stream handling.
Capabilities
Split peer review into Search & Plan, Execute, Verify, and Revise stages.
Introduced the no-code figure editor and improved large 3D structure delivery.
Expanded ingestion to more than 40 research-oriented formats and added analysis libraries.
Added citation export and improved systematic-review retry behavior.
Improved analysis access to project datasets and repaired recurring file-resolution failures.
Reliability, security, and operations
Decoupled response delivery from persistence, fixed stream races, and improved completion rendering.
Added retry and backoff for transient scholarly-source failures.
Aligned usage tracking and feature access with then-current billing controls.
2026.02
Major update
February 1–26, 2026
The account, isolated execution, billing, and Literature Intelligence foundation.
Capabilities
Added isolated code execution, R support, archive uploads, biomedical tools, persistent chat, and report output.
Improved multi-user response handling and added conversation, usage, rate, and support safeguards.
Removed datasets and tools whose licenses were not compatible with customer use.
Important notes
Earlier plan names, chat-mode labels, trials, and account-access descriptions in historical drafts are not current. See Settings → Billing and the FAQ for present behavior.
2026.01
Feature update
January 26–30, 2026
Initial CATO research-analysis platform.
Capabilities
Created the first data-analysis, agent, project, dataset-presentation, authentication, and conversation-resume flows.
Added early Python/R analysis support, user isolation, logging, data access, and reliable saved progress.
Reliability, security, and operations
Established project-scoped storage and authentication boundaries that later security updates expanded.
Need help after an update?
Check Troubleshooting for migration, reconnection, workspace, literature, and billing guidance. If the issue persists, send support the relevant IDs, timestamps, and redacted error.