Independent replications will determine whether the reported AI-assisted COBOL-to-Java workflow reduces legacy-migration effort while preserving behavior and avoiding unacceptable defect and maintenance costs.
state: expiredheat: lowuncertainty: highknownscott: mediumcoding-agents legacy-modernization software-reliability
What is this?
The case concerns a reported AI-assisted workflow for translating legacy COBOL programs into Java, with the resulting code reportedly retaining or introducing bugs. The supplied results describe AI coding assistants and multi-agent systems being used for code analysis, documentation, translation, and validation, but they consist largely of vendor or promotional claims about faster delivery and lower costs. They do not identify the original report’s authors or establish independent replication, behavioral equivalence, defect rates, or long-term maintenance costs; one source explicitly cautions that claimed 60–75% savings are an upside scenario rather than a default expectation.
Why it matters to Scott
Known via Scott’s “AI Legacy Takeover” framework, which already makes behavioral preservation, characterisation tests, measurable convergence, and maintenance-safe verification—not generated code—the basis for judging legacy replacement. A replicated COBOL-to-Java result could validate or challenge that framework, but the supplied report lacks independent equivalence, defect, and lifecycle-cost evidence, so it is currently a watch item rather than a new conclusion.
ip:framework.ai-legacy-takeoverip:framework.discussed-is-not-deployedradar:anthropic-claude-code-migrationsradar:openai-agentic-scientific-software-modernizationradar:post-merge-agentic-code-benchmark
queries asked of Scott's wikis
- agent harnesses for behavior-preserving code migration
- coding agents and semantic equivalence testing
- legacy modernization verification and regression tests
- AI-generated code defect and maintenance costs
- multi-agent workflows for large codebase transformation
- specification recovery from undocumented software
Measured heat
no measured readings yet — the hourly heat pass fills this in
How the heat travelled
no chain yet — the hourly chain pass fills this in
Evidence (2) — ⭐ canonical anchor
Interpretation history
2026-08-06T13:26:31Z
The discussion has plateaued as repetitive skepticism without independent replication, implementation, or measured reliability and lifecycle evidence. This specific episode has faded, though a future substantive replication should open a new case.
2026-08-04T12:25:39Z
The latest attachment adds no independent replication, implementation, or measured reliability evidence; discussion continues to repeat already-known migration risks. The case is cooling until behavioral-equivalence, defect-rate, effort, or maintenance-cost results emerge.
2026-08-04T11:24:48Z
The expanding discussion sharpens practical failure modes—mainframe dependencies, introduced defects, and loss of maintainability—but remains commentary rather than independent validation. The claim still awaits measured behavioral equivalence, migration effort, defect rates, and lifecycle costs.
2026-08-03T09:21:29Z
The newly attached material still resolves to the original report and its discussion, with no independent replication, deployed implementation, or lifecycle evidence. Attention is repetitive amplification rather than validation, so the migration claim remains a speculative watch item.
2026-08-03T08:21:22Z
The attached evidence remains the original report and its discussion, not an independent replication or deployed implementation. The case still hinges on behavioral-equivalence, defect-rate, effort, and lifecycle-maintenance evidence.
2026-08-03T07:21:26Z
The negligible engagement change adds no independent replication or implementation evidence. The case remains an unvalidated migration claim awaiting behavioral-equivalence, defect-rate, and lifecycle-cost results.
2026-08-03T06:23:47Z
grounded: known/medium — Known via Scott’s “AI Legacy Takeover” framework, which already makes behavioral preservation, characterisation tests, measurable convergence, and maintenance-s
2026-08-03T06:21:28Z
case created — The paper presents a bounded coding-agent migration result whose reliability and practical economics can be independently tested.
Decision trace
- 08-06 23:26expireThe discussion has plateaued as repetitive skepticism without independent replication, implementation, or measured reliability and lifecycle evidence. This specific episode has faded, though a future
- 08-04 22:25repriceThe latest attachment adds no independent replication, implementation, or measured reliability evidence; discussion continues to repeat already-known migration risks. The case is cooling until behavio
- 08-04 22:20mark_dirtycomment_update
- 08-04 21:24repriceThe expanding discussion sharpens practical failure modes—mainframe dependencies, introduced defects, and loss of maintainability—but remains commentary rather than independent validation. The claim s
- 08-03 19:21repriceThe newly attached material still resolves to the original report and its discussion, with no independent replication, deployed implementation, or lifecycle evidence. Attention is repetitive amplifica
- 08-03 19:20mark_dirtyengagement_update
- 08-03 18:21repriceThe attached evidence remains the original report and its discussion, not an independent replication or deployed implementation. The case still hinges on behavioral-equivalence, defect-rate, effort, a
- 08-03 18:20mark_dirtyengagement_update
- 08-03 17:21repriceThe negligible engagement change adds no independent replication or implementation evidence. The case remains an unvalidated migration claim awaiting behavioral-equivalence, defect-rate, and lifecycle
- 08-03 17:20mark_dirtyengagement_update
- 08-03 16:23groundKnown via Scott’s “AI Legacy Takeover” framework, which already makes behavioral preservation, characterisation tests, measurable convergence, and maintenance-safe verification—not generated code—the
- 08-03 16:21createThe paper presents a bounded coding-agent migration result whose reliability and practical economics can be independently tested.