Product signal from the work, not noise from the hype cycle.
This page tracks the recent changes that most affected route behavior, validation confidence, surface clarity, and clearer operation in the active product.
The machine-facing surface got more consistent across all active routes.
Recent MCP followthrough aligned response modes, route metadata, and Unreal issue taxonomy across diagnosis, project reads, code review, crash triage, and feature planning. The result is a steadier integration surface for editor, agent, and tool-driven workflows.
- Tool responses now carry clearer route-level behavior instead of drifting by endpoint.
- Diagnosis, project context, review, crash, and planning routes now speak a more consistent schema.
- Validation coverage expanded around large result sets and governed execution states.
Project-aware diagnosis and planning got sharper under live followthrough.
Live followthrough work improved freshness narration, code-intel fallback behavior, architecture-aware feature placement, and the execution contracts behind build and automation lanes. The product feels more grounded because more edge cases now have explicit handling.
- Diagnosis and project-read flows now do a better job carrying current evidence forward.
- Feature planning is more aware of likely placement, project structure, and repo conventions.
- Execution-capable lanes gained stronger continuity rules without overstating current capability.
Capability language tightened while target-project followthrough closed more workflow gaps.
The product overview was corrected against the actual implementation, and the same stretch of work improved verify flows, freshest-log selection, rerun controls, input classification, and scoped code-intel follow-up across target-project loops.
- Surface counts and bridge wording were corrected to match the real codebase.
- Build diagnosis and validation flows got clearer about which evidence is freshest and when a rerun is required.
- Scoped code-intel follow-up now does less guessing in live project work.
Live target-project parity work closed more gaps across doctor, validate, and project reads.
Another MCP followthrough tranche turned audit findings into implementation fixes, including stronger Build Doctor parity, safer preflight output, and restored project-intel checks that keep quality coverage honest.
- Build Doctor and validate-project behavior moved closer to the live contract.
- Project-intel type checks were restored so regression coverage stays meaningful.
- The latest Windows rerun reinforced that this lane is backed by real host evidence.
The Gamibase rebrand finished across code, docs, extension, and plugin surfaces.
Legacy naming was removed across the active workspace so the CLI, MCP runtime, VSCode extension, plugin boundary, docs, and evidence all present one consistent product identity.
- The active product surfaces now point to one name instead of a mix of legacy labels.
- Repository links and evidence references were normalized to the organization-owned remote.
- The site can speak more confidently because the underlying surfaces now match.
Preflight authority and cross-project route behavior became more explicit.
A dense stretch of target-project MCP work tightened validate-project scan limits, plugin-risk surfacing, route availability truth, shared local-dev entitlement handling, feature planning placement, and build diagnosis follow-through across multiple real project families.
- Validation now reports bounded scans and rerun guidance more truthfully.
- Route availability and license behavior are more consistent across CLI, MCP, and local setup.
- Diagnosis, review, automation, and planning all benefited from broader live-project coverage.
Surface ownership became clearer across CLI, MCP, VSCode, CI, and the UE boundary.
The active surfaces now present as one system with cleaner responsibility boundaries: terminal and CI operation, machine integration through MCP, everyday VSCode UX, and an optional plugin-owned editor mutation boundary.
- Surface-by-surface responsibilities now map to one product instead of a loose tool bundle.
- Active non-plugin surfaces are separated from later bridge depth.
- Operator expectations stay more consistent across entry points.
Trust posture moved from caveat to core workflow signal.
Review-first paths, dry-run posture, and fail-closed behavior became explicit workflow language instead of background caveats. That made the product easier to understand without weakening it.
- The strongest differentiation now includes what the product refuses to do silently.
- Limits are framed as design choices, not marketing exceptions.
- Trust language now supports the workflow model instead of interrupting it.