Jmix migrates a CUBA Platform Timesheets app to Jmix with Codex in 25 hours against a 215-hour estimate
The Jmix team migrated its Timesheets sample app (14 entities, 35 screens, about 8,000 lines of Java plus XML and SQL) from the legacy CUBA Platform to Jmix using OpenAI Codex with GPT-5.3-Codex, AGENTS.md and agent skills, JetBrains MCP, Context7 MCP and Playwright CLI. Codex migrated entities, fetch plans and business logic from short prompts, then the UI screen by screen, starting with simple reference screens and using a plan-first prompt for complex ones such as the calendar, command-line input and weekly timesheets. Some work stayed manual (project creation, Liquibase changelog, styles, security policies). The project-specific estimate from Jmix's Migration Advisor was 215 man-hours; the actual effort was 25.
The numbers (as reported)
| Actual effort | 25 man-hours |
|---|---|
| Migration Advisor estimate | 315 man-hours (215 project-specific after a fixed 100-hour baseline) |
| Source code size | 8,062 Java, 2,252 XML, 2,151 SQL lines |
| Target code size | 10,916 Java, 3,513 XML lines |
| Screens | 35 screens, 1 frame |
Facts
- Company / subject
- Jmix (Haulmont)
- Tools
- OpenAI Codex
- Models
- GPT-5.3-Codex
- Codebase
- brownfield
- Work type
- Upgrade / modernisation
- Stack
- Java, Jmix, CUBA Platform, Liquibase, Playwright
- Published
- 2026-03-25
- Last checked
- 2026-11-11
What the source claims Vendor-published
Claimed by the vendor that sells the tool. Not checked independently.
- The migration took 25 man-hours against a 215-hour project-specific estimate 25 vs 215 man-hours
- About 10 times less effort than expected ~10x less effort
“we succeeded with the Timesheets migration and spent about 10 times less effort than expected”
Caveats
Published by Jmix, which sells the target framework and the migration tooling, about its own sample application, so this is vendor-published. The baseline is Jmix's own Migration Advisor estimate, not a measured manual migration. Some behaviour was changed rather than reproduced (Ctrl+Enter instead of Enter for command input). The author says other projects may not see a similar reduction.
More stories by tool: OpenAI Codex
Related topics: Upgrades and modernisationBrownfield codebasesVendor case studies