{"id":51475363,"url":"https://github.com/asmuelle/claimwatch","last_synced_at":"2026-07-06T20:30:33.235Z","repository":{"id":366008898,"uuid":"1266480041","full_name":"asmuelle/claimwatch","owner":"asmuelle","description":null,"archived":false,"fork":false,"pushed_at":"2026-06-19T20:24:54.000Z","size":247,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-19T22:12:39.269Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/asmuelle.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-06-11T16:51:15.000Z","updated_at":"2026-06-19T20:24:58.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/asmuelle/claimwatch","commit_stats":null,"previous_names":["asmuelle/claimwatch"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/asmuelle/claimwatch","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/asmuelle%2Fclaimwatch","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/asmuelle%2Fclaimwatch/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/asmuelle%2Fclaimwatch/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/asmuelle%2Fclaimwatch/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/asmuelle","download_url":"https://codeload.github.com/asmuelle/claimwatch/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/asmuelle%2Fclaimwatch/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35205739,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-06T02:00:07.184Z","response_time":106,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-07-06T20:30:32.532Z","updated_at":"2026-07-06T20:30:33.223Z","avatar_url":"https://github.com/asmuelle.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"# ClaimWatch\n\n[![CI](https://github.com/asmuelle/claimwatch/actions/workflows/ci.yml/badge.svg)](https://github.com/asmuelle/claimwatch/actions/workflows/ci.yml)\n\n\u003e Patent and IP radar for IP-heavy startups: a weekly agent over USPTO/EPO/WIPO bulk feeds maintains living dossiers per technology area and competitor — alerting on new filings in your claim space, claim amendments, assignments, and lapses, with every assertion checkable against the canonical document.\n\n**Category:** LLM wiki / auto-research (living documents + delta alerts, à la Karpathy) \n\n## Concept\n\nPatent and IP radar for IP-heavy startups: a weekly agent over USPTO/EPO/WIPO bulk feeds maintains living dossiers per technology area and competitor — alerting on new filings in your claim space, claim amendments, assignments, and lapses, with every assertion checkable against the canonical document.\n\n## Target User\n\nFounders, CTOs, and in-house counsel at deeptech, biotech, hardware, and AI startups whose valuation rests on IP (fundraising diligence and freedom-to-operate anxiety are board-level issues), plus boutique patent law firms managing many small clients who want a $400/hr associate's monitoring work turned into a reviewable brief.\n\n## Auto-Research Mechanic (the living document + delta engine)\n\nBuilt on free, structured, canonical government data: USPTO full-text and assignment feeds, EPO OPS, WIPO PATENTSCOPE, litigation dockets — refreshed on the official Tuesday/Thursday publication cycle (cadence matches the data). Watchlists defined as CPC classes plus natural-language claim-space descriptions and named competitors/inventors. Cheap-model screening of new publications and amendments; frontier pass updates living dossiers: per-family claim-evolution timelines, assignment changes, prosecution status, fee lapses and expirations. Weekly brief reasons over prosecution history: 'Competitor X's continuation narrowed claim 1 after rejection — this opens design space in Y.' Every assertion links to document, claim number, and publication date — the highest-citation-trust vertical possible because every citation is mechanically checkable.\n\n## Product Surface\n\nWeb SaaS (dossiers, claim timelines, watchlist management) + weekly email brief as primary touchpoint — patent monitoring is inherently weekly-cadence and the brief gets forwarded to boards and outside counsel. Counsel tier adds PDF brief export formatted for client files.\n\n## Why Now (2026 timing)\n\nPatent watch explicitly identified as an under-disrupted niche with quote-only incumbents and zero cheap challengers. LLMs only recently became capable of meaningfully reading claim language and prosecution histories — and the corpus is free, structured, and crawl-unrestricted, so the access constraints hardening across the general web in 2026 don't apply at all.\n\n## Proposed Monetization\n\n$149/mo Startup (1 technology-area watchlist, 10 competitors/CPC classes), $399/mo Pro (5 watchlists, litigation-docket monitoring, counsel-ready export), $799/mo Firm (multi-client workspaces, white-label briefs). The entire sub-$1K/mo tier is empty under PatSnap/Clarivate/Questel's $20K+ quote-only floor; COGS is a few dollars per watchlist on free data.\n\n## Competition \u0026 Gap\n\nPatSnap/Clarivate/Questel (enterprise, quote-only database access plus analyst tooling), Google Patents (free search, zero monitoring or synthesis), keyword-matching alert services. None maintain a living claim-evolution dossier or explain what an amendment means for your design space.\n\n## Claimed Moat\n\n(1) Claim-language extraction and prosecution-history reasoning is deep vertical NLP on a corpus horizontal chatbots don't index conversationally — a scheduled ChatGPT task cannot diff claim amendments across a patent family. (2) The longitudinal claim-evolution graph is accumulated state: two years of watching how a rival's claims narrowed is retroactively irreproducible. (3) Law-firm tier creates distribution lock-in as firms standardize client deliverables on the brief format. (4) Provable accuracy in the one research domain where every citation is mechanically verifiable builds a trust brand that survives any incumbent feature launch.\n\n---\n\n# Evaluation (multi-agent adversarial review)\n\n\n## Comparables\n\n- PatSnap — quote-only, ~$15K/yr single-seat to $40K-$80K+/yr enterprise; $100M ARR (2023, +20% YoY), $352M raised, $1B valuation\n- Questel — ~$750M annual revenue (2025), enterprise IP software and services, PE-owned\n- Clarivate — $2.56B total revenue (2024); divested its IP product group in 2024, signaling incumbent retreat from the segment\n- Patlytics — AI patent platform, $40M Series B (Apr 2026), ~$65M total raised, 40% of Am Law 100 as customers; nearest well-funded threat\n- Solve Intelligence — AI patent drafting/prosecution, $40M Series B (Dec 2025), $55M total, eight-figure ARR up 10x YoY; 60% law firms / 40% corporate IP\n- IamIP — patent search and monitoring from $99/mo; proves a low-cost tier exists but is keyword-alert grade, no claim-evolution synthesis\n- XLPat — free patent alerts (keyword matching); price-anchor pressure at the bottom of the market\n- Boutique IP firm subscription services — $2K-$10K/mo per client; one firm reports $2.4M/yr from 30 subscription clients (validates Firm-tier white-label economics)\n- Patent attorney hourly rates — $200-$800/hr; the inferior 'alternative' startups currently pay for monitoring/FTO work\n\n## Tech Stack \u0026 Unit Economics\n\nIngestion: USPTO Open Data Portal bulk delta datasets (daily) + Patent File Wrapper API (free), EPO OPS paid tier (~EUR 1-4K/yr at scale; free 4GB/mo fair-use tier insufficient beyond pilot), WIPO PATENTSCOPE via vendor or constrained scraping, CourtListener/RECAP + metered PACER (or Docket Alarm API) for the litigation tier; cron aligned to the Tue (grants) / Thu (pre-grant pubs) cycle, Temporal or pg_cron workers. Storage: Postgres + pgvector (claim embeddings via voyage-3 or text-embedding-3-large), S3 for raw XML/PDF, claim-version tables per family. Critical design rule: claim diffing is deterministic XML text diff — never LLM-generated. Screening: CPC prefilter then Haiku 4.5 / Gemini Flash relevance classification of ~200-800 docs/week per watchlist (~1-3M tokens/wk, $1-3/wk). Synthesis: Sonnet 4.6 dossier updates + weekly brief with cite-or-omit grounding — every sentence must resolve to a stored doc ID + claim number and is mechanically validated post-generation; Opus-class pass only for flagged prosecution-history reasoning over office-action context (50-150K tokens/family, $3-10/wk). App: Next.js + Postgres SaaS, Resend/Postmark email brief, PDF export for counsel tier. Unit economics: Startup tier COGS ~$20-40/user/mo (screening $5-12, frontier synthesis $12-40 across 4 weekly briefs amortized, embeddings/infra $3-5) against $149 = ~70-85% gross margin; Pro litigation-docket tier adds $30-150/mo PACER/docket exposure and must be metered or it eats margin. Team of 2-3 can ship an MVP in roughly one quarter; the hard 20% is claim-space recall evaluation, which needs a hand-labeled ground-truth set per vertical before any retention claim is believable.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fasmuelle%2Fclaimwatch","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fasmuelle%2Fclaimwatch","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fasmuelle%2Fclaimwatch/lists"}