cubic

What do they actually do

Cubic is a GitHub‑integrated AI code review service. When a pull request opens, it runs an automated review in an isolated sandbox, posts inline comments, and generates a PR summary; it can also be triggered on existing PRs by mentioning the bot. Reviewers use Cubic’s web/desktop app to triage findings and accept fixes, with some changes applied via one‑click actions or delegated to automated background agents for more complex updates (docs, GitHub App).

The company emphasizes security and data handling: reviews run in short‑lived environments, source code isn’t stored persistently, AI providers aren’t allowed to train on customer code, and SOC 2 Type I controls are in place (docs). Cubic lists early customers and example reviews (e.g., cal.com, n8n, Linux Foundation) so teams can see real outputs before installing (site). Pricing includes a 14‑day free trial with unlimited reviews, a free tier with 40 PR reviews/month, and a Team plan at $24/month per developer billed annually ($30/month billed monthly) (pricing/docs, site).

Beyond PR comments, Cubic is rolling out background agents for broader auto‑fix coverage, custom rules and team memory to tailor reviews, and an auto‑generated repository wiki to help with onboarding and codebase context. They also signal upcoming analytics and integrations to help teams measure review velocity and identify where automation should apply next (docs, site, blog on wiki).

Who are their target customer(s)

• Startup engineering teams shipping frequently with high PR volume: Reviewers get overwhelmed and small changes linger unreviewed; they need fast triage, clear PR summaries, and low‑friction fixes to keep velocity. Cubic automates reviews and summaries with one‑click fixes to reduce reviewer load (docs).
• Mid‑market or enterprise engineering orgs with strict legal/security requirements: They are cautious about sending source code to third‑party models and need compliance, contractual controls, and auditability. Cubic’s no persistent code storage, no training on customer code, and SOC controls address these blockers (docs).
• Platform/dev‑productivity teams enforcing standards across many repos: They struggle to scale style, linting, and review consistency as teams grow. Cubic supports custom rules, learns from org comment history, and runs background agents to apply low‑risk fixes automatically (docs).
• Open‑source maintainers with limited reviewer time: Maintainers spend time triaging low‑quality or unclear PRs. Cubic runs sandboxed reviews via a GitHub App and posts inline comments so maintainers see vetted suggestions before investing time (GitHub App).
• Teams hiring/onboarding frequently: New engineers lose time searching commit history and asking for context. Cubic can generate a searchable codebase wiki from code and history to surface answers faster (blog on wiki).

How would they acquire their first 10, 50, and 100 customers

• First 10: Hands‑on pilots with 10 early adopters (YC/partner networks and fast‑shipping startups). Install the GitHub App on 1–3 busy repos during the 14‑day trial and provide weekly onboarding to demonstrate faster triage and one‑click fixes; capture time‑to‑merge and backlog metrics plus a short case study (docs, GitHub App, pricing/docs).
• First 50: Lean into product‑led growth via the GitHub App/Marketplace and free tier (40 PR reviews/month) with example reviews and smooth onboarding; convert installs to paid seats. Amplify with demo content, how‑tos, and community posts/meetups to drive low‑touch signups (GitHub App, pricing/docs, site).
• First 100: Begin a focused mid‑market/enterprise motion: named‑account outreach to platform/productivity teams, a pilot plus security/compliance packet (ephemeral sandboxes, no persistent storage, no training on customer code, SOC artifacts) and procurement‑friendly contracts; partner with dev‑tool consultancies and offer bot‑seat/chargeback bundles for broad rollouts (docs, site).

What is the rough total addressable market

Top-down context:

GitHub reports 180M+ developers on the platform and ~43M pull requests merged per month, signaling large review workload. GitHub also cites 20M+ users across ~77K enterprises, providing an enterprise seat anchor (Octoverse, GitHub enterprise materials).

Bottom-up calculation:

Cubic’s Team plan lists $24/month per developer billed annually → $288 per developer per year. At 1–10% penetration of the 20M enterprise seats, ARR ranges from ~$57.6M to ~$576M (200k–2M paying developers × $288) (pricing/docs, GitHub enterprise materials).

Assumptions:

• Seat-based pricing remains primary (not PR-volume pricing).
• Penetration scenarios of 1%, 5%, and 10% of GitHub enterprise developer seats are used for illustration, not forecasts.
• A meaningful share of GitHub enterprise users overlaps with teams willing to adopt third‑party AI review tools.

Who are some of their notable competitors

• CodeRabbit: AI code review bot for GitHub that comments on pull requests and suggests fixes; a direct substitute for automated PR feedback.
• CodiumAI PR‑Agent: Open‑source/GitHub App that reviews PRs, generates summaries, and suggests improvements with configurable policies.
• Amazon CodeGuru Reviewer: ML‑powered automated code reviews focused on issues like code quality and security; integrates with popular repositories and CI.
• SonarCloud: Cloud service for static code analysis that annotates PRs with code quality and security issues across multiple languages.
• Snyk Code: Developer‑first SAST that surfaces security issues in PRs and IDEs, often used alongside code review to enforce security gates.