OSS program and maintainer-needs research
Last verified: 2026-07-11 (JST). This document separates statements made by program owners from ReproProof’s own conclusions. Program rules can change; re-check every linked source before applying.
Executive finding
Both named programs exist and currently expose application forms. OpenAI uses a qualitative, rolling review for maintainers of active and important projects. Anthropic publishes several objective tracks, but a new project with no adoption does not satisfy them. ReproProof should be built for genuine maintainer value and measured adoption; neither program should be treated as a launch-time entitlement.
OpenAI: Codex for Open Source
Source: application form and program terms.
- Status: the form displayed a Submit control and stated that applications are reviewed on a rolling basis on the verification date.
- Who can apply: maintainers of active open-source projects. OpenAI looks for meaningful usage, broad adoption, or clear ecosystem importance, plus evidence of active maintenance.
- Maintainer role: the form asks whether the applicant is a primary or core maintainer.
- Signals named by OpenAI: repository usage, ecosystem importance, pull-request review, issue triage, release management, and other ongoing maintenance.
- Stated benefits: six months of ChatGPT Pro including Codex; possible API credits for eligible maintainer workflows; conditional Codex Security access for qualified repositories.
- Form fields observed: first and last name, ChatGPT-account email, public GitHub username, public repository URL, maintainer role, a 500-character qualification explanation, interest in Codex Security and/or API credits, OpenAI Organization ID, a 500-character API-credit use explanation, and an optional 500-character note.
- Terms: benefits are discretionary, non-transferable, and not guaranteed. Accurate information and verification of repository affiliation/control may be required. Unauthorized scanning is forbidden. Fraud, multiple identities, and misleading metrics may cause rejection or revocation.
Not specified by the official sources
- No numeric minimum for stars, downloads, dependents, contributors, or Criticality Score is published.
- No fixed application response time or guaranteed credit amount is published.
- No explicit minimum age appears in the program-specific form or terms reviewed. General account/service terms still apply.
ReproProof judgment
ReproProof is not currently a credible applicant: the repository has an initial public source release, but it has no external users or established maintenance history. Apply only after public evidence shows active maintenance plus meaningful use or ecosystem importance.
Anthropic: Claude for Open Source
Sources: program page and terms.
- Status: the page displayed “Apply now”; terms say applications are rolling until Anthropic closes the period or reaches its cap.
- Benefit: six consecutive months of Claude Max 20x. The terms state the base subscription is complimentary, but overages may be charged. Existing paid subscriptions resume after the benefit period unless canceled.
- Cap: the terms state up to 10,000 approved recipients unless Anthropic increases it.
- The applicant must satisfy at least one maintainer or ecosystem-impact track and every general requirement.
- Published maintainer-track alternatives:
- 500 dependent repositories in aggregate; or
- 100 dependent packages in aggregate; or
- 200,000 combined monthly registry downloads; or
- listed committer/maintainer of a recognized foundation or language project; or
- 100 merged PRs to public repositories the applicant does not own in the prior 12 months; or
- 20 unique external contributors with merged PRs to a maintained repository in the prior 12 months; or
- a maintained repository with OpenSSF Criticality Score at least 0.4.
- Discretionary ecosystem-impact track: maintainers or significant contributors may explain why the ecosystem meaningfully depends on their project even without a numeric threshold.
- General requirements include: natural person; at least 18 or local age of majority, whichever is greater; eligible country and export/sanctions status; GitHub account in good standing and at least two years old; public OSS activity in the preceding 90 days; contribution to an OSI-approved project; not Anthropic staff/contractor or immediate household/family.
- Application information: GitHub OAuth sign-in, GitHub-associated email, planned use, and a qualification explanation of at most 500 words.
- Anthropic explicitly discounts trivial, duplicative, automated, or manipulated activity and may revoke benefits for artificial inflation.
ReproProof judgment
The owner’s stated age of 18 satisfies the age floor only if 18 is also the age of majority in their jurisdiction. GitHub-account age, eligible residence, and recent public activity are unverified. ReproProof starts at zero on every project-adoption threshold. Do not apply yet.
OpenSSF Criticality Score
Confirmed official source: OpenSSF Criticality Score.
- It defines project influence/importance on a 0-to-1 scale.
- Its goals are to score open-source projects, identify projects the community depends on, and guide proactive security improvement.
- Anthropic explicitly uses 0.4 as one eligibility route. OpenAI does not publish a Criticality Score threshold.
- Judgment: it is an influence signal, not a target to game. A new repository should expect a low or absent score.
Similar support programs
These are adjacent rather than interchangeable:
- GitHub Secure Open Source Fund: rolling security-focused support for maintainers; the page states $10,000 per selected project, security programming, an 18+ rule, community traction/adoption, governance, and GitHub Sponsors regional availability.
- GitHub open-source benefits: GitHub states that eligible maintainers can receive benefits such as Copilot Pro, Actions capacity, maintainer community access, Sponsors, and security/funding programs.
- OpenSSF Alpha-Omega: partners with critical projects and ecosystems on sustainable security improvement and vulnerability finding/fixing; not a general new-project credit program.
- Sovereign Tech Agency programs: funds and supports critical open digital infrastructure through fund, resilience, standards, fellowship, and challenge mechanisms; selection and timelines differ by program.
Maintainer needs and evidence for the product
Confirmed evidence
- OpenAI’s own program page names review, issue triage, releases, security, and code quality as significant maintainer workload.
- GitHub’s issue guidance asks bug reporters for reproduction steps and expected versus actual behavior.
- scikit-learn’s minimal reproducer guidance calls a minimal reproducible example key to efficient community communication.
- The peer-reviewed/preprint systems LIBRO, Issue2Test, and Echo demonstrate both demand and difficulty. Their reported success rates remain far below 100%, so an honest tool must expose uncertainty and execution evidence.
- GitHub Copilot cloud agent and SWE-agent focus primarily on resolving issues and opening code-change PRs.
Inference
There is room for a narrower, maintainer-controlled product that stops before the fix: parse a report, identify missing facts, generate a candidate reproduction, execute it in isolation, and provide a reviewable patch plus evidence. This is not proof of market adoption; it is the product hypothesis to validate with maintainer interviews and opt-in pilots.
Open research questions
- Will maintainers trust a reproduction-only tool more than a full repair agent?
- Is local setup cost lower than the time saved on triage?
- Which repositories permit safe, deterministic dependency setup without network access?
- What validation best detects a test that fails for the wrong reason?
- Which report fields are actionable across different project cultures?
No user count, recommendation, benchmark beyond the controlled fixtures, or program acceptance is implied by this research.