Haystack Software

What do they actually do

Haystack Software makes Haystack Editor, a web-based pull request (PR) review tool. It replaces GitHub’s default diff view with a canvas-style page that groups related hunks into titled sections, highlights the core logic first, lets reviewers follow variables or functions across files, and includes an integrated AI chat that has already read the PR so you can ask questions about intent, blast radius, or missing tests (site, repo README, HN thread). Under the hood, it combines static analysis and language servers with LLMs to ground answers in concrete code references rather than loose text (site, HN).

Teams add it to their workflow by installing the Haystack GitHub App or logging in with GitHub; the service analyzes a PR and produces a Haystack review link so reviewers can open that view instead of the raw diff. There are public demo PRs you can try without connecting a repo, and the open-source codebase is public and actively developed (site, repo, HN). The founders have discussed a simple per-seat price around $20/person, noted that diffs are sent to external LLMs (OpenAI/Anthropic) for analysis, and stated they don’t train on or retain customer code; they’ve also said an on‑prem/enterprise option is planned for teams that need it (HN founder comments).

Who are their target customer(s)

• Front-line code reviewers at software teams: They spend time scrolling long, line-by-line diffs and struggle to see the key logic and intent of a change; they need a faster, narrative view of a PR.
• Engineers reviewing outside their domain or new hires: They lack context on where changed functions/variables are used and waste time chasing references across files; they need cross-file tracing and grouped sections to get oriented quickly.
• Engineering managers and team leads: Reviews delay shipping and quality varies; they want faster cycles and quick signals on blast radius, risky logic, and missing tests without hand-auditing every line.
• Open-source maintainers and small teams triaging many PRs: High PR volume makes thorough manual triage infeasible; they need a reordered, prioritized view to decide which PRs need deep review versus are safe to merge.
• Security/compliance and enterprise buyers: They are cautious about sending proprietary code to third-party LLMs and need on‑prem or stronger privacy guarantees before broad adoption.

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

• First 10: Personally reach out to users who tried the public demo and early HN commenters, invite them to install the GitHub App on a real repo, and onboard them 1:1 while iterating the automatic link/analysis flow (site, HN).
• First 50: Target small teams and YC startups with tailored demos using real PRs and early testimonials; enable automatic reviewer links so adoption is one click and convert with a simple per‑seat plan (~$20/person) while addressing privacy upfront (HN).
• First 100: Expand to mid-sized teams and OSS maintainers via lightweight sponsorships, a few case studies with time-saved metrics, and integrations (PR webhooks/Slack) that make Haystack part of normal review flows; pilot on‑prem/enterprise options and privacy SLAs to unlock security‑sensitive buyers (ROADMAP, HN).

What is the rough total addressable market

Top-down context:

There are tens of millions of software developers globally; Evans Data reported 26.3M developers in 2023, while GitHub alone crossed 100M developers by 2023 (and reported 180M by 2025) (Evans Data, GitHub Octoverse). Code review is a near-universal part of Git-based workflows, so a PR-review assistant priced per seat has a broad potential base.

Bottom-up calculation:

Assuming a focus on GitHub-centric orgs: if 200,000 teams have an average of 15 reviewing seats and pay ~$20/user/month (~$240/year), the TAM is roughly 200,000 × 15 × $240 ≈ $720M annually (pricing ref).

Assumptions:

• 200,000 organizations use PR-based review on GitHub/Git-like workflows at team scale (10+ developers).
• Average of 15 active reviewer seats per org that would benefit from Haystack-style review.
• Per-seat pricing averages $20/month ($240/year), consistent with founder-stated plans.

Who are some of their notable competitors

• GitHub Copilot (code review / PR summaries): Built-in AI in GitHub that generates PR summaries, leaves inline review comments, and runs automated reviews inside GitHub—competing for the “surface the important parts of a PR” workflow (Copilot code review, PR summaries).
• Sourcegraph (Cody + code navigation): Provides cross-repo search, precise “go to definition”/“find references,” and an AI assistant (Cody) for code context; overlaps on cross-file tracing and grounded answers for understanding multi-file changes (Cody, code navigation in PRs).
• CodeSee: Creates visual maps of a codebase and auto-generates review maps and AI summaries/tours of a PR, offering a visual, narrative-first PR review experience similar to Haystack’s canvas approach (product, Review Maps).
• SonarSource (SonarCloud / SonarQube): Static analysis that runs on PRs to flag bugs, security risks, and quality issues, and can decorate PRs or gate merges—competing on surfacing risky parts of a change more than narrative understanding (PR analysis & decoration).
• Amazon CodeGuru Reviewer: Automated PR reviewer that posts recommendations (security, resource leaks, API misuse) as PR comments and integrates into CI—an established option for automated review assistance (features, PR reviews docs).