.png&w=256&q=75)
🤓 Latest Submissions
Morphogenetic Software Engine
Modern coding agents can generate code quickly, but generating a change is not the same as establishing that the change is correct. Developers still need to determine what is wrong in an existing repository, understand why it is wrong, evaluate its impact, repair it, and verify that the repair actually resolves the original problem without introducing new issues. Morphogenetic Software Engine (MSE) is a software-engineering verification platform designed around this missing control loop. Instead of acting as another code-generation assistant, MSE analyzes a repository as a structured system and connects detection, evidence, repair, and verification into one workflow. MSE operates through three core capabilities: Find, Explain, and Repair & Verify. Find analyzes the repository structure, dependencies, implementation behavior, specification/documentation relationships, and invariants to identify risky behavior, inconsistencies, and potential drift. Explain creates an evidence chain for each finding. A developer can trace a finding to the affected source, supporting evidence, relevant invariant or specification contract, counterexample, dependency/blast-radius information, and the proposed repair. This makes the reasoning inspectable rather than presenting an unexplained AI-generated answer. Repair & Verify allows the developer to apply an MSE-generated repair or edit the source directly in an in-memory working repository. MSE can then re-run its analysis against that exact modified repository state. The system compares the new state against the original finding and evaluates whether the relevant invariant and counterexample conditions have actually been resolved. A change is therefore not considered verified merely because a patch was applied; verification is tied to the specific tested repository revision. The architecture separates the repository baseline from the working state so developers can inspect diffs, test successive repair attempts, revert changes, and retain verification history. IBM Bob 2.0 was used throughout the development workflow for repository analysis, architectural planning, implementation, debugging, refactoring, test generation, and iterative validation. Bob's Agent mode was used for direct implementation, Plan mode for decomposing the MSE architecture and workflows, and focused agent/subagent workflows were used to investigate individual modules and problems while keeping the main development context manageable. Bob's file and command tooling enabled iterative changes followed by build and test validation. IBM documents Bob's Agent/Plan/Ask modes and its subagent/tooling workflow for exactly these development patterns. The result is intended to demonstrate a different model for AI-assisted software engineering: AI can help generate changes, but an engineering system should also preserve evidence, reason about impact, and verify the resulting state.
27 Sep 2026
