Argument Reviews

• New Review panel. A Review section in the argument side panel lets anyone walk through an argument claim by claim and decide whether each one is true, false, or unknown — and then premise by premise whether the relationship between the claims really holds up. When you finish, we run the logical evaluation and show you the verdict (Valid and Sound, Fails, Logically Invalid, Vacuous, or Indeterminate).
• Breadcrumb navigation. The review lives in the same sidebar as the Claims and Sources libraries. Navigate to Overview → Review → Claims / Relationships / Result, and click any claim or premise to jump straight into its editor. A "Proceed to Result" button appears once everything is decided.
• Pick a reason. Each True / False / Unknown and Accept / Reject decision can be annotated with a plain-language reason (Common knowledge, Non sequitur, Counterexamples exist, etc.) drawn from a curated 35-entry catalogue.
• Skip and come back. Uncertain? Skip an item and keep reviewing; use the TOC to revisit anything later.
• Results overlay on the graph. After evaluation, the main argument graph is painted with your decisions — claim nodes show your assignment, propagated values pinned by constraints are highlighted (semi-green / semi-red pills when an Unknown was inferred), operator edges reflect your accept/reject decisions, and the conclusion premise gets a verdict badge. Click the verdict badge to re-open the Result view; click any assignment pill to jump back to that claim's editor.
• Verdict explainer. The Result view explains each verdict in plain language with a mini worked example (rendered in the same text-tree style as the argument) and links to further reading (IEP, SEP, Wikipedia).
• Reviews persist locally. Your work-in-progress review is saved to your browser automatically and restored the next time you open the argument. Reviews in multiple tabs of the same browser stay in sync.
• Cleaner argument graph. Supporting premises no longer show a "Supporting" chip and use a neutral border; only the conclusion premise stands out visually.

My Sources Library

• New "My Sources" page at /profile/sources shows every source you've added across your arguments, organized by argument and version in a tree, just like My Claims. Sources you've created that aren't attached to any claim appear in a Not Used section at the bottom of the tree (italicized so they're distinguishable from arguments).
• View Sources button in the argument view. A new icon button beside View Claims opens a breadcrumb-navigated sidebar overlay listing every source used in the current argument version. Click a source to drill into its details, then use the breadcrumb to return.
• Source detail panel shows the full IEEE citation, a list of every claim that uses the source, and per-claim View buttons that open the argument with the relevant claim node pre-selected in a new tab.

Citation Schema Update

Citation sources now use structured author names (first/last name fields) instead of plain text strings. Existing citations that don't conform to the updated format have been removed. If you had sources attached to claims, they may need to be re-added.

Argument Forking Improvements

• More accurate fork comparisons. When comparing a forked argument to its original, entities are now matched by their fork provenance rather than by position or title. This means edits and additions are identified correctly even when the argument structure has changed significantly.

• Argument forking now preserves full history. Forked arguments carry detailed provenance records for each premise, expression, variable, claim, and source. This enables precise change tracking across all parts of an argument.

Logic Editing

• Premises automatically create a bound variable. When you add a new premise to an argument, a bound variable is automatically created for it. This makes cross-premise referencing available immediately without an extra step.

• Stronger data integrity checks. The engine now validates the logical structure of an argument after every edit. If an inconsistency is detected, you will see a clear error message describing the problem rather than silent incorrect behavior.