Status & Limitations
What the available project material supports, what remains partial, and what has not been verified.
Last substantively updated: 27 August 2026
This is the date these website status notes were edited.
This is the date these website status notes were edited.
| Capability | Current state | Known limitation | Demonstration or documentation |
|---|---|---|---|
| Source reader | Mostly working, some formats need better support | Format coverage, parsing accuracy, and runtime stability untested. | Reading and annotating |
| Highlights, notes, tags | Fully working | Persistence and source-location edge cases untested. | Features |
| Visual evidence workspace | Working, but some changes to node behaviors necessary | — | Organizing evidence |
| Keyword and meaning search | Working, semantic similarity retrieval could be improved | No published retrieval-quality evaluation. | Retrieval |
| Assisted analysis | Experimental/partial | Relationships and synthesis require source review. | Analysis |
| Citations and Writing editor | Working | Metadata parsing and style output require testing. | Citations |
| Tabular-data tools | Working, data extraction could be improved | PDF extraction and inferred types can be wrong. | Tabular data |
| Blueprint workflows | Experimental/partial | Repeatability, cancellation, and output behavior untested. | Automation |
| Plugins and packs | Documented design with examples, further testing needed | Isolation, permissions, lifecycle, and compatibility unverified. | Extensions |
| Cross-document timeline | In-progress | Timeline events (dates) are properly extracted from added sources, but there's not currently a dedicated view for them. | — |
| Citation graph | Working for most common citation types | Properly extracts and merges most in-text and bibliography entries, some unique citation styles still fail | — |
| Bias Lab | Experimental; scope unverified | Attempts to adapt the LIBRA methodology, still a work-in-progress | Design requirements |
Known limitations
- No validated performance, retrieval-quality, extraction-accuracy, or model-accuracy benchmarks are published here.
- No comparative environmental measurements are available. Local processing does not remove electricity use or upstream training impacts.
- Non-generative parsing and classification can still misread, omit, or incorrectly connect material.
- Source-evaluation signals are heuristic, not authoritative credibility judgments.
- Prompt and context traces show recorded inputs and outputs, not hidden model reasoning.
Future possibilities
Possible future work includes broader testing, clearer extension isolation, accessibility evaluation, stronger status evidence, and eventual source sharing. These are directions rather than promises, dates, or a release roadmap. Development follows personal interest and available time.