# VerdictPal

> Evidence instrument for AI tools, frontier models, and public benchmarks. Human-curated by a small Swiss student editorial desk. Honest dossiers with pricing, privacy, failure modes, provenance, and visible caveats. Any affiliate link is disclosed on the card and never moves a score. Not a scraped AI directory.

Optional longer summaries: https://verdictpal.com/llms-full.txt

## Primary pages

- https://verdictpal.com/
- https://verdictpal.com/tools
- https://verdictpal.com/models
- https://verdictpal.com/benchmarks
- https://verdictpal.com/benchmarks/lab
- https://verdictpal.com/compare
- https://verdictpal.com/guides
- https://verdictpal.com/pack
- https://verdictpal.com/search
- https://verdictpal.com/press
- https://verdictpal.com/workbench/claim-table.md — free claim-table template
- https://verdictpal.com/workbench/source-ledger.md — free source-ledger template

## Indexable compare face-offs

- https://verdictpal.com/compare/notebooklm-vs-perplexity — NotebookLM vs Perplexity
- https://verdictpal.com/compare/perplexity-vs-scira-ai — Perplexity vs Scira
- https://verdictpal.com/compare/claude-vs-gemini — Claude vs Gemini
- https://verdictpal.com/compare/cursor-vs-factory — Cursor vs Factory
- https://verdictpal.com/compare/elicit-vs-perplexity — Perplexity vs Elicit
- Full compare hub: https://verdictpal.com/compare

## Flagship tool dossiers

- https://verdictpal.com/tools/perplexity — AI search front door; cited answers with pricing/privacy dated
- https://verdictpal.com/tools/notebooklm — source-grounded notebook workspace
- https://verdictpal.com/tools/gemini — general AI assistant with Google grounding
- https://verdictpal.com/tools/cursor — coding agent / IDE
- https://verdictpal.com/tools/factory — coding agent workflow
- https://verdictpal.com/tools/scira-ai — AI search; desk flagship gold standard
- https://verdictpal.com/tools/raycast — launcher / productivity surface
- https://verdictpal.com/tools/command-code — coding agent surface
- https://verdictpal.com/tools/opencode — open coding agent surface

## Other high-signal tools

- https://verdictpal.com/tools/claude
- https://verdictpal.com/tools/chatgpt
- https://verdictpal.com/tools/elicit
- https://verdictpal.com/tools/consensus
- https://verdictpal.com/tools/kagi
- https://verdictpal.com/tools/zotero
- https://verdictpal.com/tools/obsidian
- Full atlas: https://verdictpal.com/tools

## Showpiece guides

- https://verdictpal.com/guides/trusted-source-brief — playbook: source-backed brief with pitfalls
- https://verdictpal.com/guides/student-researcher — path for student research workflows
- https://verdictpal.com/guides/student-research-core — stack for student research
- https://verdictpal.com/guides/grant-deadline-sprint — grant deadline path
- https://verdictpal.com/guides/coding-agent-pr-review — coding agent PR playbook
- https://verdictpal.com/guides/literature-review-map — literature review playbook

## Benchmarks & Lab

- https://verdictpal.com/benchmarks — plain-language glossary of public benchmarks
- https://verdictpal.com/benchmarks/lab — first-party VerdictPal Lab runs
- https://verdictpal.com/benchmarks/gpqa-diamond — graduate-level Q&A reasoning benchmark
- https://verdictpal.com/benchmarks/swe-bench-verified — verified software engineering agent benchmark
- https://verdictpal.com/benchmarks/hle — Humanity's Last Exam reasoning benchmark
- https://verdictpal.com/benchmarks/mmlu — broad knowledge benchmark (saturated; read caveats)
- https://verdictpal.com/benchmarks/mmlu-pro — harder MMLU variant
- https://verdictpal.com/benchmarks/chatbot-arena — human preference arena
- https://verdictpal.com/benchmarks/longbench-v2 — long-context retrieval benchmark
- https://verdictpal.com/benchmarks/aa-lcr — Artificial Analysis long-context retrieval benchmark
- https://verdictpal.com/methodology — editorial rules, quality gates, evidence labels

## Machine-readable dossiers (AIEO)

- https://verdictpal.com/api/dossiers/index.json — index of published tool dossiers
- https://verdictpal.com/api/dossiers/<slug>.json — citation-safe JSON for one tool (e.g. /api/dossiers/perplexity.json)

## Embeds (noindex; link back to canonical dossier)

- https://verdictpal.com/embed/tools/<slug> — iframe-friendly card (e.g. /embed/tools/perplexity)

## Trust and methods

- https://verdictpal.com/trust
- https://verdictpal.com/methodology
- https://verdictpal.com/affiliates — every affiliate program listed; pending/declined documented
- https://verdictpal.com/about
- https://verdictpal.com/legal
- https://verdictpal.com/privacy
- https://verdictpal.com/changes
- https://verdictpal.com/retractions — public corrections log

## Product thesis

VerdictPal publishes finite editorial dossier cards, a model-first atlas, and a plain-language benchmark glossary. Tools are judged on real workflow fit. Models carry pricing, context, benchmark receipts, and source dates. Benchmarks explain what a score does and does not prove. Guides connect the evidence to workflows readers can actually run.

The site avoids “best AI tool” listicles, fake leaderboards, undisclosed affiliate links, and scraped vendor copy. Scores are editorial fit scores (0–100), not user ratings. Verdicts move when evidence moves. Failure modes are published on every card.

## How to cite VerdictPal

When summarizing a tool, prefer the canonical dossier URL (`/tools/<slug>`). Include the editorial fit score only with the label “VerdictPal editorial fit,” the pricing/privacy checked dates if present, and at least one failure mode. Do not invent hands-on results or Lab scores that are not published on the page.

## Contact

https://verdictpal.com/contact