Implementation Workspace​

The platform now reviews your actual implementation, not just the plan. Create a workspace from any hierarchy node, upload your schematic / simulation / PCB layout, and the platform plays the role of a senior reviewer:

• Automated rule checks (Layer 1) — floating nets, missing decoupling, voltage rating violations, trace-width floor, annular-ring, DRC clearance, copper-edge clearance, etc.
• Pattern checks (Layer 2) — LDO bypass, I²C pull-ups, crystal load caps, MOSFET gate resistor, decap proximity, via-in-pad, pour balance, controlled-impedance routing, etc.
• Grounded AI review (Layer 3) — Gemini 2.5 Flash judges whether your design actually implements the project's requirements, DFMEA failure modes, block diagram and BOM. Streams live so you watch findings appear as they're generated. Falls back to Claude Haiku if Gemini is unavailable.

Cross-artifact review​

The highest-value review is cross-artifact. With two or more sub-artifacts uploaded, the platform surfaces inconsistencies between schematic and sim ("sim assumes lossless trace but layout has 6 in of FR4"), between schematic and layout ("layout footprint disagrees with schematic"), and between any of them and the stated requirements.

Engineering disciplines (EE / LE / ME)​

Every finding carries a discipline tag — EE (Electrical), LE (Layout), or ME (Mechanical). Filter the findings list to "show only LE" while you're in a placement review, then switch to "EE" for a schematic walk-through. The platform infers the discipline from the finding category and links it to the right engineer.

Discussion threads on findings​

Each finding has its own comment thread for the iterative back-and-forth between the EE and the layout engineer while the fix is in flight. Distinct from the artifact-level review workflow — threads are about implementation dialogue, the review workflow is about artifact approval.

Manual findings + edit​

The AI will miss things — visual issues, supplier constraints, project-specific concerns. + Add finding lets you log anything the AI missed; the pencil icon on any finding lets you revise severity, message, recommendation or discipline (AI provenance is preserved on the audit trail).

Share by link​

Click any finding's F-id chip to copy a deep-link. Paste it into chat or email and your teammate lands directly on that finding, expanded and highlighted, with the discipline filter you had set.

Guided tour​

A 13-step walkthrough is one click away from the ? icon next to the workspace name — covers hierarchy linking, three upload modes, the three AI reviewers, live streaming, findings disposition, and follow-on directives.

Fixes (release-day)​

• Team members on a hierarchy node now correctly inherit the workspace scope — owner and editor get full write access, viewer gets read-only, all clearly indicated in the UI.
• Read-only viewers attempting to write now receive a clear 403 with an explanation, instead of a misleading 404.
• Workspace artifacts now correctly appear on the linked hierarchy node's Artifacts tab the moment the workspace is created (was previously surfaced only via a fallback scan).
• Sub-artifact deletion now works for team editors and owners, not just the workspace owner.
• Findings with layout-specific categories (trace-width, annular-ring, mounting-hole, etc.) now save correctly — a previously-too-strict category enum has been opened up.
• Writes are correctly disabled when viewing a past revision — the read-only banner makes the state explicit.

Comment severity tagging​

Every review comment now carries a severity. Reviewers tag as they comment, owners filter by severity when dispositioning. Audit reports filter by severity as well.

Implement-on-review for DFMEA​

Click "Implement" on an accepted DFMEA review comment and the AI patches the row in place. Previously this only worked for SWOT and DVP.

Visibility toggle​

Hierarchy nodes can now be flipped between public, team, and private visibility from the node header. Public nodes appear on the Discover page.

Fixes​

• Public artifact responses no longer leak userId (security).
• Team viewers can no longer restore revisions (previous bug let viewers escalate).
• Mongo EXPRESS_IXSCAN plan shape now correctly reported as IXSCAN, not COLLSCAN.

Removed​

• Translation tooling stubs are now no-ops (was breaking builds for some configs).