# ObviouslyNot

> AI-powered concept scanner that surfaces strategic technical concepts hidden in your GitHub repositories. First scan free. No API key or credit card required. Patentability determinations are made by patent attorneys, not by ObviouslyNot.

ObviouslyNot scans codebases for distinctive algorithms, architectures, and methods, then generates scored strategic concepts with evidence citations from actual code. The scanner identifies; the patent attorney determines whether what was found may be patentable. Built for software engineers, tech founders, vibe coders, and patent attorneys.

We also publish original research on patent representation gaps, including the largest current data series on the U.S. patent gender gap (9 pages covering the prosecution funnel, the PNAS 2026 "breakthrough penalty" study, AI patenting, international comparisons, and evidence-based interventions).

## Main Pages

- [Homepage](https://obviouslynot.ai/): Product overview, how it works, FAQ, and free scan signup
- [For Patent Attorneys](https://obviouslynot.ai/attorneys): Attorney-specific landing page with AI-structured disclosures, Alice/Mayo aware claims, attorney-first design, and research on prosecution outcomes
- [Discover](https://obviouslynot.ai/discover): Four ways to find distinctive technical concepts in your code (local scanner, GitHub Action, AI agent skills, or hosted service)
- [Patent Skills](https://obviouslynot.ai/patent-skills): Open-source skills for patent analysis and principle extraction; works with Claude Code, Gemini CLI, Codex CLI, Cursor
- [Download Concept Scanner](https://obviouslynot.ai/download-patent-scanner): Privacy-first local scanner, runs on your machine using Ollama, no cloud uploads
- [GitHub Concept Scanner](https://obviouslynot.ai/github-concept-scanner): Run the concept scanner as a GitHub Action or Docker image; scan every pull request in CI, bring your own OpenAI-compatible provider or local Ollama
- [Women in Innovation](https://obviouslynot.ai/women): Audience page for women inventors and the institutions that support them
- [About](https://obviouslynot.ai/about): Mission, team background, and company story
- [Contact](https://obviouslynot.ai/contact): Contact information and social links (YouTube, LinkedIn, GitHub)

## Resources Hub

- [Resources Hub](https://obviouslynot.ai/resources): All frameworks, guides, research series, and tools

## The Patent Gender Gap Series (9 pages)

The patent gender gap series is our flagship research output. The hub summarizes the data; each child page walks one specific mechanism with primary-source citations.

- [Patent Gender Gap (hub)](https://obviouslynot.ai/patent-gender-gap): USPTO + WIPO data on women in U.S. and global patenting. Hub for the 8 child pages below.
- [Where Women's Patents Disappear](https://obviouslynot.ai/patent-gender-gap/where-womens-patents-disappear): The prosecution funnel, stage by stage. Half the gap lives in differential abandonment after rejection.
- [The Breakthrough Penalty](https://obviouslynot.ai/patent-gender-gap/the-breakthrough-penalty): The 2026 PNAS study by Sowrirajan, Whalen, and Uzzi: the gender gap is concentrated on unconventional inventions and mediated by examiner experience, not bias. Flagship article. 14.9 pp gap on unconventional patents; 44.7 pp paired-team gap.
- [AI's Gender Gap Is Becoming an Ownership Gap](https://obviouslynot.ai/patent-gender-gap/ai-ownership-gap): USPTO AI Patent Dataset (15.4M documents, 1976-2023). Women's AI patenting lags the baseline; today's gap becomes tomorrow's wealth gap.
- [Engineering vs Biotech Breakdown](https://obviouslynot.ai/patent-gender-gap/engineering-vs-biotech-breakdown): Field-by-field data. Chemistry/biotech ~18%, mechanical engineering ~6%. Why the spread exists.
- [WIPO International Data](https://obviouslynot.ai/patent-gender-gap/wipo-international-data): 18.0% global PCT inventor share (2024). Türkiye 26%, China 24%, Japan 10%. EPO 2026 + UK IPO 2026 data with termination rates by team type.
- [One Woman Is Not Parity](https://obviouslynot.ai/patent-gender-gap/one-woman-is-not-parity): The three-metric distinction: "at least one woman" 21.9%, share-weighted 12.8%, all-women teams 4%. Why the choice of metric determines the story.
- [Historical Timeline 1900-2025](https://obviouslynot.ai/patent-gender-gap/historical-timeline-1900-2025): 125 years of women in U.S. patenting. Title IX (1972), Bayh-Dole (1980), and the modern wave of institutional interventions.
- [Closing Strategies That Work](https://obviouslynot.ai/patent-gender-gap/closing-strategies-what-works): Four interventions with measured evidence: WashU WIT (+129%), USPTO Pairolero RCT (+12 pp grant rate), Stanford OTL 50-year arc, USIPA pledge.

## The Patent Representation Gap Index

- [Patent Representation Index](https://obviouslynot.ai/patent-innovation-gap-index): Composite score ranking 91 tech companies on how well they translate women's technical work into patent inventorship. Five sub-indices, weighted, normalized 0-100. Built from BigQuery patent data and government EEO-1 workforce filings.
- [Full Company Data (170 companies)](https://obviouslynot.ai/patent-innovation-gap-index/patent-innovation-gap-full): The complete underlying data table with sub-index breakdowns by company.

## Access and Representation

- [Technology Discrimination](https://obviouslynot.ai/technology-discrimination-patent-access): How software inventors face functional inequality in the patent system; the structural argument behind the software-access gap.

## Guides

- [Is it obvious?](https://obviouslynot.ai/is-my-invention-obvious): Hub: obviousness under 35 USC 103, with MPEP 2141 (Graham and KSR), MPEP 2143 (the seven rationales) and MPEP 2131 (anticipation). Describes what the tests require; applies them to nothing.
  - [The seven rationales behind an obviousness rejection](https://obviouslynot.ai/is-my-invention-obvious/mpep-2143-seven-rationales): An obviousness rejection is an argument, and this is the menu it is drawn from. Most of the seven are about combination rather than copying.
  - [When combining things that already exist is obvious](https://obviouslynot.ai/is-my-invention-obvious/combining-things-that-already-exist): Almost every software invention is an arrangement of known components, which makes this the question engineers actually have.
  - [Novelty and obviousness are different tests](https://obviouslynot.ai/is-my-invention-obvious/novelty-and-obviousness-are-different): Two statutes, two questions. An invention can clear the first and fail the second, and nobody has done exactly this answers only one of them.
  - [Anticipation needs one reference to carry the whole claim](https://obviouslynot.ai/is-my-invention-obvious/one-reference-has-to-carry-the-claim): Being told an idea is not new usually means someone found something similar. Anticipation is stricter than that.
- [Why Most Software Patents Start Too Late](https://obviouslynot.ai/software-patents-start-too-late): The entire patent industry assumes someone already knows they have an invention. For software engineers, that assumption is almost always wrong. The gap is not the filing process. It is the discovery step.
- [What counts as prior art?](https://obviouslynot.ai/what-counts-as-prior-art): Hub: prior art under 35 USC 102, with MPEP 2128 (printed publications and the Wayback Machine presumption) and MPEP 2152 (available to the public). Describes the tests; applies them to nothing.
  - [Printed publication means findable](https://obviouslynot.ai/what-counts-as-prior-art/printed-publication-means-findable): The test has almost nothing to do with printing. A forum post qualified because it was directed to skilled readers and sufficiently disseminated.
  - [What available to the public reaches](https://obviouslynot.ai/what-counts-as-prior-art/available-to-the-public): The broadest phrase in the novelty statute, and the statute never defines it. On sale is the exception to it, not an example of it.
  - [The one year you get, and what it does not cover](https://obviouslynot.ai/what-counts-as-prior-art/the-one-year-you-get): Narrower than its reputation. It is tolerance for your own disclosure, not protection against someone who got there independently.
- [What Is Patent Mining?](https://obviouslynot.ai/patent-mining): Patent mining extracts hidden value from data sources, engineering teams, and codebases. Three definitions, the crypto mining parallel, and how AI changes the discovery step.
- [The PPA as a Stock Option on Your IP](https://obviouslynot.ai/provisional-patent-strategy): A provisional patent application costs $65-$325 and buys you 12 months. Think of it like a stock option: a small premium for the right to decide later whether to invest fully.
  - [What a provisional application actually is](https://obviouslynot.ai/provisional-patent-strategy/what-a-provisional-actually-is): A filing date with a description attached. Cheap to file and easy to file badly, and the two facts are related.
  - [What the twelve months actually buy](https://obviouslynot.ai/provisional-patent-strategy/what-the-twelve-months-buy): A provisional on its own confers nothing. Three conditions, each able to fail independently, and failing any one buys nothing at all.
  - [How a later application inherits an earlier one's date](https://obviouslynot.ai/provisional-patent-strategy/how-a-later-application-inherits-a-date): Copendency is the condition that does the structural work. Let the chain break and the date is gone, with no reaching back.
  - [The restoration nobody quotes](https://obviouslynot.ai/provisional-patent-strategy/the-reference-deadline-nobody-quotes): The one documented softening of the deadline. It is a restoration route, not a fourteen-month window, and treating it as the latter is how a date gets lost.
  - [Continuation and divisional are names, not instruments](https://obviouslynot.ai/provisional-patent-strategy/continuation-divisional-and-cip): Knowing they are one provision tells you the section 120 conditions apply to all of them, whatever the paper is called.
  - [Claiming the benefit: what the manual adds](https://obviouslynot.ai/provisional-patent-strategy/claiming-the-benefit): Four requirements that can defeat a benefit claim and none of which appear in section 119(e).
- [Software patents in Europe](https://obviouslynot.ai/software-patents-in-europe): Hub: European and UK software patentability. EPC Article 52 as such, T 0641/00 COMVIK on technical and non-technical features, and the Aerotel four-step approach. Reports the frameworks; applies them to nothing.
  - [COMVIK: features that cannot support inventive step](https://obviouslynot.ai/software-patents-in-europe/comvik-technical-and-non-technical): Europe does not throw out an invention because part of it is a business idea. It does something more consequential.
  - [Aerotel: superseded in February 2026](https://obviouslynot.ai/software-patents-in-europe/aerotel-macrossan-four-steps): The four-step approach governed UK practice for two decades. The Supreme Court set it aside in February 2026 and aligned the UK with the EPO.
  - [The further technical effect](https://obviouslynot.ai/software-patents-in-europe/the-further-technical-effect): Assessed without reference to the prior art, so being faster than an earlier program is not itself a further technical effect.
  - [UK Patents Act section 1, after Emotional Perception](https://obviouslynot.ai/software-patents-in-europe/uk-patents-act-section-1): The statute states the exclusion and has not changed. The method for reading its as-such limit changed in February 2026, when the Supreme Court set Aerotel aside.
- [Invention Disclosure Procedure: A Step-by-Step Guide](https://obviouslynot.ai/invention-disclosure-procedure): Most engineering teams have no procedure for capturing inventions. The gap is not filing. It is recognizing what to file. A 7-step process from recognition through filing.
- [Patent Continuations](https://obviouslynot.ai/patent-continuation-applications): A continuation turns one specification into a patent family: new claims, same priority date, no new matter. Why disclosed-but-unclaimed scope can be dedicated to the public, the copendency and term rules, the double-patenting tradeoff, and an honest read on where AI drafting helps.
- [How your application is put together](https://obviouslynot.ai/how-your-application-is-put-together): Hub: application format. 35 USC 113 (when a drawing is required and why a later one cannot rescue), 37 CFR 1.84 (drawing rules and the colour petition), MPEP 608.02 (flow charts belong in the drawings).
  - [When a drawing is required, and why a later one cannot rescue](https://obviouslynot.ai/how-your-application-is-put-together/when-a-drawing-is-required): A figure added later is not retroactive. It sits in the file and does not reach back to the filing date.
  - [What a patent drawing must look like](https://obviouslynot.ai/how-your-application-is-put-together/what-a-patent-drawing-must-look-like): A coloured architecture diagram is not a small formatting choice.
  - [Flow charts belong in the drawings](https://obviouslynot.ai/how-your-application-is-put-together/flow-charts-belong-in-the-drawings): A location rule, not a prohibition. Engineers get this wrong by default and an examiner will object.
- [Software Patents After Alice](https://obviouslynot.ai/software-patents-after-alice): Alice changed the rules, not the game. 58% of software patents still get approved. A guide to what survives, prosecution strategies, and what developers should document.
  - [Recentive v. Fox: the AI eligibility ceiling](https://obviouslynot.ai/software-patents-after-alice/recentive-v-fox): Applying established machine learning to a new field is an abstract idea. The court expressly left room for claims that improve the machine learning itself.
  - [Contour v. GoPro: what passes Alice step one](https://obviouslynot.ai/software-patents-after-alice/contour-v-gopro): Claims reciting a particular mechanism, two parallel video streams with only the low-quality one transmitted, were eligible at step one without reaching step two.
  - [Ollnova v. ecobee: timing rules survive Alice](https://obviouslynot.ai/software-patents-after-alice/ollnova-v-ecobee): Timing constraints and a conditional transmission trigger were a specific technological improvement. The $11.5 million verdict fell anyway, on procedure.
  - [GoTV v. Netflix: labels are not structure](https://obviouslynot.ai/software-patents-after-alice/gotv-v-netflix): "Algorithm" and "architecture" are highly general labels when nothing structural sits behind them. A $2.5 million verdict was wiped out.
  - [Results-oriented claims still fail Alice](https://obviouslynot.ai/software-patents-after-alice/us-patent-7679637-v-google): Claims saying what applications are "arranged to allow" rather than how they achieve it are result-oriented. The court set out three analytical approaches.
  - [Desjardins: when AI training claims are eligible](https://obviouslynot.ai/software-patents-after-alice/ex-parte-desjardins): An agency adjudication, not a court decision. The claim limitation required learning without forgetting. The claims still stand rejected for obviousness.
  - [Examination guidance on eligibility](https://obviouslynot.ai/software-patents-after-alice/uspto-2024-eligibility-guidance): This is what examiners are told, and it binds no court.
- [Continuation Claims](https://obviouslynot.ai/continuation-claim-drafting): A continuation is a vehicle for adaptive claiming, not adaptive disclosure. What the Federal Circuit permits when claims are drafted with a competitor's product in mind, where the original disclosure caps that reach, and what Kroy did and did not decide about asserting unadjudicated claims after an adverse IPR.
- [Building in Public and Patents](https://obviouslynot.ai/build-in-public-patent-risk): The modern way to build software is, by default, the fastest way to disclose an invention before anyone files on it. How a public commit, Show HN post, or open-source release starts the section 102 clock, why the US grace period is a fallback rather than a plan, and why foreign rights close immediately.
- [Section 101 Declaration Strategy](https://obviouslynot.ai/section-101-declaration-strategy): The USPTO's April 30, 2026 memo clarifies and reinforces the Subject Matter Eligibility Declaration as a voluntary §101 evidence tool. Practitioner adoption is uneven and the §132(a) new-matter rule constrains what a SMED can do. The evidentiary opportunity lives at the drafting stage, not the prosecution stage.
- [Who Owns the Code You Write on Your Own Time?](https://obviouslynot.ai/who-owns-your-side-project): State statutes limiting employee invention assignment, and why a software side project's git history is the evidence that decides it.
- [Co-Founder IP: Who Owns What](https://obviouslynot.ai/cofounder-ip-ownership): Two founders build a product. Who owns the patents? Without an assignment agreement, each co-inventor can independently license the invention to your competitor. Without permission. Without sharing a dollar.
- [AI Patent Discovery and Privilege](https://obviouslynot.ai/ai-patent-discovery-evidence): GAI use in patent drafting may not enjoy attorney-client privilege (Baker Donelson, May 2026). The risk is real for generic LLM use; structured-disclosure tools with audit trails are categorically different in discovery.
- [AI Inventorship and Human Conception](https://obviouslynot.ai/ai-inventorship-who-counts-when-ai-helps): Built it with Cursor, Copilot, or Claude Code? Under Thaler v. Vidal and the USPTO's November 2025 revised guidance, only humans can be inventors, even when AI tools materially contribute. A doctrinal walk-through plus a conception note template and practitioner intake checklist.
- [IP Ownership Clauses for Remote Contractors](https://obviouslynot.ai/ip-ownership-clauses-remote-contractors): A remote contractor can write your next core patent before lunch. Without the right cross-border assignment language, they may still own it.
- [Your Open Source License Is Also a Patent Decision](https://obviouslynot.ai/open-source-licenses-patent-rights): What Apache 2.0, MIT, BSD, GPL, and MPL actually do to patent rights, in both directions, plus the disclosure clock every release starts.
- [IP Protection Checklist for Engineering Teams](https://obviouslynot.ai/ip-protection-checklist-engineering): Your team ships code every day. Some of it is novel enough to patent. Without a checklist, those inventions get published with no filing date and no protection. Five areas, eighteen items.
- [Tight Ownership Protocols for AI Code Teams](https://obviouslynot.ai/tight-ownership-ai-code-dev-teams): When AI generates most of the code, who conceived the invention? A practical protocol for capturing inventorship evidence before the filing scramble.
- [Agentic AI Inventorship: The Open Question](https://obviouslynot.ai/agentic-ai-inventorship): Using Copilot is settled, the human conceives and is the inventor. When an autonomous agent does most of the building on its own, no patent office or court has decided who invents. A frontier map: what's settled globally (human inventor required), what's open, and the conception record to build while it's unsettled.
- [AI Patent Risk: 2026 Firm Consensus](https://obviouslynot.ai/ai-patent-risk-firm-consensus-2026): Baker Donelson, Hogan Lovells, Kirkland, Alston & Bird, and Arnold & Porter have all published 2026 alerts on AI-drafted patent risk. The consensus on discovery, privilege, work product, and validity, and what it means for prosecution practice.
- [Drafting with Local AI: What It Protects, and What It Costs](https://obviouslynot.ai/local-model-patent-drafting): Patent practice has a sharper version of the AI confidentiality problem. We measured what self-hosted drafting costs across 29 models. Whether the weights are open barely matters, about 4 points. How much memory you have matters enormously, about 23.
- [Why Local-First Matters for Patent Discovery](https://obviouslynot.ai/local-first-patent-discovery): Before you file a patent, your invention is a trade secret. Any AI tool that processes unfiled inventions in the cloud introduces risk across three legal frameworks. Local-first scanning eliminates all three.
- [Cloud AI Leaks: Trade Secrets vs Patents](https://obviouslynot.ai/cloud-ai-leaks-trade-secrets-vs-patents): Pasting code into a cloud AI can jeopardize both trade secret protection and patent eligibility before you knew there was a decision to make. ~1,500 federal trade secret cases filed in 2025. The fix starts before your code touches the cloud.
- [Software Invention Disclosure Example: A Completed Walkthrough](https://obviouslynot.ai/software-invention-disclosure-example): Most disclosure guidance tells you what to write. This shows a finished one: a fictional software invention taken from a ten-minute flag through the full technical record, with the evidence and contributor questions filled in.
- [AI Tool Compliance Checklist for Patent Practitioners](https://obviouslynot.ai/ai-tool-compliance-checklist): AI adoption among patent professionals is 15% while general legal AI is at 79%. The gap is confidentiality. A structured checklist for evaluating AI tools before they touch client material.
- [How to Find Inventions in a Codebase](https://obviouslynot.ai/find-inventions-in-a-codebase): Distinctive engineering rarely announces itself. Twelve places it tends to hide, a seven-step process for surfacing it, and a 30-minute review agenda you can run this sprint.
- [Pre-Filing Patent Privacy: Cloud Risk and Trade Secret Protection](https://obviouslynot.ai/patent-cloud-privacy): Three legal frameworks govern how pre-filing patent documents can be stored and processed. The risk isn't cloud itself. It's how long your unfiled invention sits on someone else's server.
- [AI Agents and Your Intellectual Property](https://obviouslynot.ai/ai-agents-intellectual-property): 99% of dev teams use AI coding tools. 29% are prepared to secure them. When your agent reads your codebase, where does that code go? The answer has legal consequences.
- [IP Ownership in Local-First Sync Engines](https://obviouslynot.ai/ip-ownership-local-first-sync-engines): Distributed contributors, open-source CRDTs, and cross-border teams create unique IP ownership risks. Learn how to secure clean patent title before it becomes a deal-breaker.
- [Tesla's .smol Patent: Data Format IP for Counsel](https://obviouslynot.ai/tesla-smol-data-format-patent-analysis): Tesla's WO2024073080 describes a hybrid columnar-row file format for ML training data. This analysis walks in-house counsel through the eligibility, prior art, and disclosure risks.
- [Protecting Creativity in the AI Era](https://obviouslynot.ai/protecting-creativity-ai): A practical guide for artists, developers, and knowledge workers to safeguard their ideas, innovations, and value in an age where AI can absorb everything you create.
  - [AI Innovation Addendum](https://obviouslynot.ai/protecting-creativity-ai/ai-innovation-addendum): A comprehensive legal template to protect human creativity and innovation in AI-assisted development
  - [AI Innovation Addendum (Lightweight)](https://obviouslynot.ai/protecting-creativity-ai/ai-innovation-addendum-lightweight): A simplified 1-page template for protecting creativity in AI-assisted work
- [Patenting AI Agents for Autonomous Payments](https://obviouslynot.ai/patenting-ai-agents-autonomous-payments): Bank of America grew its AI patent portfolio 94% from 2022 to 2024 while most fintech founders are still asking whether their agentic payment system is patentable. A practical guide to what is, what isn't, and where the legal landmines sit.
- [Invention Discovery in AI-Augmented Sprints](https://obviouslynot.ai/invention-discovery-ai-dev-teams): When sprints compress from weeks to days, documentation shrinks with them. Novel solutions get committed, merged, and forgotten. A practical workflow for AI-augmented dev teams to capture inventive work before competitors file first.

## Frameworks

- [How to Write a Patent Disclosure: The Three-Level Framework](https://obviouslynot.ai/problem-first-disclosure): Most developers describe what they built. Attorneys need to know what problem it solves. A practical template that combines problem-first structure with three levels of specificity.
- [AI+Human Innovation Framework](https://obviouslynot.ai/ai-human-innovation-framework): Innovation has always been a dance of imagination and tools. This framework maps the nine dimensions of innovation, showing how human imagination and AI acceleration can align to expand the frontier of what's possible.

## For Patent Practitioners

- [Before the PTAB](https://obviouslynot.ai/ptab-practice): Hub: PTAB institution and review. Bifurcated institution (2025 memo), Apple v. Squires, Arthrex and Director Review, and a 2025 NPRM that remains a proposal.
  - [Discretionary denial comes first, and separately](https://obviouslynot.ai/ptab-practice/ptab-discretionary-institution): The Director decides institution, not a panel, and not in two stages. Nonbinding guidance that has changed twice since March 2025.
  - [Apple v. Squires: no right to institution](https://obviouslynot.ai/ptab-practice/apple-v-squires): Meeting the statute's requirements does not entitle you to an inter partes review.
  - [Arthrex, and the rule that makes you choose](https://obviouslynot.ai/ptab-practice/arthrex-and-director-review): You get Director Review or panel rehearing, not both. The institution-review deadline was waived to thirty days in June 2026.
  - [Proposed, not adopted](https://obviouslynot.ai/ptab-practice/ptab-institution-nprm-2025): A proposal, not a rule, and a final rule under OIRA review is still not a rule either. Status stated as of August 15, 2026.
- [Filing and Fees](https://obviouslynot.ai/uspto-filing-and-fees): Hub: USPTO fees and filing mechanics. UAIA entity discounts, FY2025 IDS size fees, AFCP 2.0 expiration, DOCX surcharge and electronic signatures.
  - [The 60 and 80 percent fee reductions](https://obviouslynot.ai/uspto-filing-and-fees/uaia-small-micro-entity-fees): Bigger discounts mean a wrong entity-status call costs more than it used to.
  - [IDS size fees and the required assertion](https://obviouslynot.ai/uspto-filing-and-fees/uspto-fy2025-fee-rule-non-continuation): Every information disclosure statement now needs a written assertion about its fee.
  - [After-final practice without AFCP 2.0](https://obviouslynot.ai/uspto-filing-and-fees/afcp-termination): The free after-final look is gone; plan the interview earlier.
  - [Filing format and who may sign](https://obviouslynot.ai/uspto-filing-and-fees/docx-and-electronic-signatures): Software may now carry the signature, but the signer's own hand still has to make it.
- [Duties, and Who May Practise](https://obviouslynot.ai/practitioner-duties): Hub: practitioner duties and admission. Duties of disclosure and reasonable inquiry, USPTO AI-tools guidance, foreign-applicant representation, and the 2024 access reforms.
  - [Disclosure, inquiry, and other agencies](https://obviouslynot.ai/practitioner-duties/duties-of-disclosure): Two duties, and the second one reaches documents nobody sent you.
  - [Using AI before the Patent Office](https://obviouslynot.ai/practitioner-duties/ai-tools-in-practice-guidance): No new rules for AI, and no relief from the old ones.
  - [When a foreign applicant must use a US practitioner](https://obviouslynot.ai/practitioner-duties/foreign-applicant-representation): If your applicant lives abroad, a registered practitioner has to sign, and the penalty is not a fee.
  - [Who may practise, and on what terms](https://obviouslynot.ai/practitioner-duties/practitioner-access-reforms): Three 2024 rules changed who may appear before the Office and on what terms.
- [Invention Disclosure Form](https://obviouslynot.ai/invention-disclosure-form): A comprehensive, patent-bar-first invention disclosure form: conception date, prior and planned public disclosure, outside collaboration, inventorship, the technical description, and prior-art context.
- [Software §101 Eligibility Worksheet](https://obviouslynot.ai/software-101-eligibility-worksheet): An Alice/§101 scaffold for software: name the abstract-idea risk, the technical improvement and its evidence, meaningful integration, real-world effect, and hardware integration points. Practitioner work-aid.
- [Non-Obviousness Worksheet](https://obviouslynot.ai/non-obviousness-worksheet): A §103 scaffold: define PHOSITA, map the prior-art landscape, capture the closest reference combinations, articulate why PHOSITA wouldn't combine them, and record secondary indicia. Practitioner work-aid.
- [Claim Scope Ladder](https://obviouslynot.ai/claim-scope-ladder): A graduated-specificity scaffold: capture the invention at narrow, intermediate, and broad rungs, check disclosure support for each, anchor to the commercial embodiment, and run a design-around self-test.

## Legal Templates

- [Idea Disclosure Template](https://obviouslynot.ai/idea-disclosure-template): A free, plain-language form to capture a technical idea, what it is, the problem it solves, how it works, before you take it to a patent attorney. Download the editable Word template.
- [Idea Strength Check](https://obviouslynot.ai/idea-strength-check): A one-page screen before a full disclosure: the problem, your solution, why it's better, whether anyone's built it, your protection path, and public-disclosure status.
- [Three Levels of Your Idea](https://obviouslynot.ai/three-levels-of-your-idea): An articulation exercise: describe your invention at three levels, the specific implementation, what it does, and the underlying principle, so you can talk about it precisely with anyone.
- [Disclosure Quality Check](https://obviouslynot.ai/disclosure-quality-check): A pre-handoff self-check: enablement detail, limiting language, claim-shaped sentences, figures, inventorship, ownership clarity, and public-disclosure dates, the gaps that weaken a disclosure.

## Legal Precedents

- [Who Owns It, and Who Invented It?](https://obviouslynot.ai/who-owns-it-and-who-invented-it): Hub: ownership versus inventorship in US patent law, via Thaler v. Vidal, Omni MedSci v. Apple and Fortress Iron v. Digger Specialties.
  - [Omni MedSci: promise to assign is not assignment](https://obviouslynot.ai/who-owns-it-and-who-invented-it/omni-medsci-v-apple): A university bylaw said inventions "shall be the property of the University." The Federal Circuit held that transferred nothing, over a dissent. The ownership half of the picture.
  - [Fortress Iron: the coinventor you can't find](https://obviouslynot.ai/who-owns-it-and-who-invented-it/fortress-iron-v-digger): Two supplier engineers were coinventors and were not named. One could be added later; the other could not be found. Both patents were held invalid.
  - [Thaler v. Vidal: inventors must be human](https://obviouslynot.ai/who-owns-it-and-who-invented-it/thaler-v-vidal): An AI system cannot be listed as an inventor, because the Patent Act says inventors are individuals. The court expressly did not decide whether AI-assisted inventions are patentable.
  - [Minerva v. Hologic: when founders can fight back](https://obviouslynot.ai/who-owns-it-and-who-invented-it/minerva-v-hologic): The rule reaches only as far as what the inventor actually represented. The Court described three situations involving no contradiction, starting with the ordinary employment agreement.
  - [The 2024 AI inventorship guidance is rescinded](https://obviouslynot.ai/who-owns-it-and-who-invented-it/revised-ai-inventorship-guidance): The 2024 AI inventorship framework is gone, and conception is the whole test again.
- [What recent patent decisions add up to](https://obviouslynot.ai/recent-patent-decisions): Cross-case synthesis of 27 U.S. patent decisions (2021 to 2026) affecting software and AI companies, and the front door to the per-case library. Reports holdings only; contains no legal advice.
- [Patent Term and Double Patenting](https://obviouslynot.ai/patent-term-and-double-patenting): Hub: obviousness-type double patenting and patent term, via In re Cellect and Allergan v. MSN, which must be read as a pair.
  - [In re Cellect: when extra term becomes a liability](https://obviouslynot.ai/patent-term-and-double-patenting/in-re-cellect): Time the Patent Office gives back for its own delay counts toward the expiration date used in a double-patenting comparison. The court also closed every equitable escape route.
  - [Allergan v. MSN: the limit on In re Cellect](https://obviouslynot.ai/patent-term-and-double-patenting/allergan-v-msn): The first patent in a family sets the ceiling on exclusivity, so it cannot be an improper extension of anything. Filing order and issuance order decide it.
  - [Ex parte Baurin: the limit on Allergan's limit](https://obviouslynot.ai/patent-term-and-double-patenting/ex-parte-baurin): The Patent Office will apply Allergan only in a narrow configuration, and will reject for double patenting even when no patent term is being extended. An agency decision, not a court decision.
- [If You Are Sued, or You Sue](https://obviouslynot.ai/if-you-are-sued-or-you-sue): Hub: IPR estoppel via Caltech v. Broadcom and Ingenico v. IOENGINE, and willfulness via SRI v. Cisco and Columbia v. Gen Digital. Each pair must be read whole.
  - [Caltech v. Broadcom: IPR is one shot](https://obviouslynot.ai/if-you-are-sued-or-you-sue/caltech-v-broadcom): The petition is a single, final opportunity to raise every patent and printed-publication argument, so the prior-art search has to be finished before you file.
  - [Ingenico: estoppel follows grounds, not evidence](https://obviouslynot.ai/if-you-are-sued-or-you-sue/ingenico-v-ioengine): The same document can be estopped in one argument and freely usable in another. A launch email and an archived downloads page proved public use.
  - [SRI v. Cisco: what willful infringement means](https://obviouslynot.ai/if-you-are-sued-or-you-sue/sri-v-cisco): Willfulness is not a heightened test requiring egregious conduct. Cisco lost because it had no reasonable defense, with no evidence of copying or concealment.
  - [Columbia v. Gen Digital: claims, not the spec](https://obviouslynot.ai/if-you-are-sued-or-you-sue/columbia-v-gen-digital): Every improvement Columbia identified was real, described in the patent, and absent from the claims. The $185 million judgment was vacated.
  - [Hikma v. Amarin: inducement needs active steps](https://obviouslynot.ai/if-you-are-sued-or-you-sue/hikma-v-amarin): What matters is whether you encouraged the infringing use, not whether a reader could take your words that way. Encouragement can still be implicit.
  - [EcoFactor v. Google: damages evidence tightened](https://obviouslynot.ai/if-you-are-sued-or-you-sue/ecofactor-v-google): A royalty number pulled from a license's preamble, rather than its terms, is not evidence. The lesson lands at signing time, not at trial.
  - [Brumfield v. IBG: foreign damages, U.S. patents](https://obviouslynot.ai/if-you-are-sued-or-you-sue/brumfield-v-ibg): Foreign revenue can be reached by U.S. patent damages if domestic infringement caused it. The framework applies; the causation still has to be proved.
  - [Lynk Labs: prior art dated from filing](https://obviouslynot.ai/if-you-are-sued-or-you-sue/lynk-labs-v-samsung): A published application is prior art from the day it was filed, under the pre-AIA statute. A search that filters by publication date can miss it.
  - [Socket Solutions: no presumed irreparable harm](https://obviouslynot.ai/if-you-are-sued-or-you-sue/socket-solutions-v-import-global): Proving your patent is valid and infringed does not by itself get you an injunction. Decided August 4, 2026, so check it is still current.
- [When Must I File?](https://obviouslynot.ai/when-must-i-file): Hub: patent timing via Celanese v. ITC (on-sale bar for secret processes) and Minerva v. Hologic (assignor estoppel).
  - [Celanese v. ITC: secret process, public sales](https://obviouslynot.ai/when-must-i-file/celanese-v-itc): Celanese sold a sweetener made by a secret process for years, then filed on the process. The patents were invalid. Secrecy plus selling the output is the combination that cost them.
- [How Broadly Can I Claim?](https://obviouslynot.ai/how-broadly-can-i-claim): Hub: claim breadth via Amgen v. Sanofi (enablement) and Juno v. Kite (written description). The two requirements fail broad claims differently.
  - [Amgen v. Sanofi: the more you claim, the more you enable](https://obviouslynot.ai/how-broadly-can-i-claim/amgen-v-sanofi): Two methods for finding the rest of a claimed class were "little more than two research assignments." Breadth is allowed; it has to be paid for in disclosure.
  - [Juno v. Kite: two examples, one billion lost](https://obviouslynot.ai/how-broadly-can-i-claim/juno-v-kite): Two examples of an enormous functional class did not show the inventor possessed the class. A section 112 case where the number is the consequence, not the subject.
  - [Wyeth v. AstraZeneca: dosage claims and enablement](https://obviouslynot.ai/how-broadly-can-i-claim/wyeth-v-astrazeneca): A dosage range the specification cannot support is an enablement problem, not an FDA question.
  - [Showing possession of what you claim](https://obviouslynot.ai/how-broadly-can-i-claim/showing-possession-of-what-you-claim): Flow charts are not required, says one passage. Function-only description may lack support, says another. They are in the same section for a reason.
  - [Enablement and the Wands factors](https://obviouslynot.ai/how-broadly-can-i-claim/enablement-and-the-wands-factors): The word doing the work is undue. Almost every invention requires some experimentation.
  - [Three requirements in one sentence](https://obviouslynot.ai/how-broadly-can-i-claim/three-requirements-in-one-sentence): Describing a function is not, by itself, describing an invention.
- [AI Patent Legal Precedents](https://obviouslynot.ai/ai-patent-legal-precedents): A comprehensive guide to how courts and patent offices treat AI-assisted inventions, with authoritative links to USPTO guidance and case law. Understand the legal landscape with plain English explanations of key rulings.
- [LKQ v. GM: design patent obviousness reset](https://obviouslynot.ai/lkq-v-gm): LKQ v. GM (Fed. Cir. en banc, 2024): the Rosen-Durling test for design patent obviousness is overruled in favor of the Graham factors.
- [AI Prompts in Federal Discovery (2026)](https://obviouslynot.ai/ai-prompts-discovery-federal-court): Magistrate Judge Farrish ordered AI prompts in CLF v. Shell, then was stayed pending Rule 72(a) review. The broader emerging framework distinguishes attorney litigation prompts (often work product), expert prompts (Rule 26(b) methodology), party-side prompts (fact-specific), and consumer-AI use under protective orders.
- [AGI SureTrack v. Farmers Edge §101](https://obviouslynot.ai/agi-suretrack-section-101): Federal Circuit, June 2, 2026, precedential. Five farm-equipment data patents fall §101 at Alice step two. Sensor-plus-GPS pattern treated as abstract on generic hardware. Section 285 fee-shifting vacated and remanded.

## Case Studies

- [27 Strategic Concepts Hiding in a Flight-Control Codebase](https://obviouslynot.ai/px4-autopilot-concept-scan): We pointed the scanner at a mature open-source flight stack and it returned 27 strategic concepts in 39 minutes. A worked example of what a scan produces, with three concepts walked through end to end.
  - [Hover-Thrust Integrator Compensation](https://obviouslynot.ai/px4-autopilot-concept-scan/hover-thrust-compensation): A reproduced scanner result from the PX4 Autopilot scan: updating a hover-thrust estimate without a step change in controller output, by pre-compensating the velocity integrator.
  - [Sequential Desaturation with Airmode Priority](https://obviouslynot.ai/px4-autopilot-concept-scan/sequential-desaturation): A reproduced scanner result from the PX4 Autopilot scan: eliminating actuator saturation iteratively while preserving roll and pitch authority ahead of yaw and thrust.
  - [External Mode-Executor Registration](https://obviouslynot.ai/px4-autopilot-concept-scan/external-mode-registration): A reproduced scanner result from the PX4 Autopilot scan: registering external navigation modes with stable identifiers across reboots, coupled to liveness-aware arming checks.

## Product

- [Concept Scanner App](https://app.obviouslynot.ai/): The scanner application (separate domain)

## Optional

- [Privacy Policy](https://obviouslynot.ai/privacy): How code and IP data is protected
- [Terms of Service](https://obviouslynot.ai/terms): Terms and conditions