Killed By Claude Report
← HomeCurrent verdict
Assessment
Claude is increasingly good at reviewing code, debugging CI issues, finding vulnerabilities, and operating across repositories and tools. That absolutely pressures Graphite's AI review layer.
But Graphite is not dead, because its core value is still the workflow substrate around PRs:
- stacked pull requests
- stack-aware merge queue
- team review inbox
- GitHub-synced review UX
- adoption across CLI, editor, PR page, and notifications
So yes, Anthropic is eating pieces of Graphite.
It is not yet a clean substitute for Graphite's end-to-end PR operations product.
Biggest historical hit
The biggest hit is Claude Opus 4.6.
Anthropic explicitly says Opus 4.6 has better code review and debugging skills, can operate more reliably in larger codebases, and sustains agentic coding tasks for longer. That directly attacks Graphite's AI PR review, PR chat, and CI debugging story.
If buyers decide they mostly want an AI reviewer inside existing GitHub workflows, Graphite's AI differentiation gets thinner fast.
What still protects them
Graphite still has real protection, but it is mostly workflow moat, not model moat.
Key defenses:
- Stacked PRs are a concrete workflow product, not just an LLM wrapper.
- Stack-aware merge/CI logic is operational software with edge cases, sequencing rules, and team process embedded in it.
- Deep GitHub integration across web, CLI, editor, inbox, and Slack creates switching cost.
- Teams adopting Graphite are standardizing how they review and land code, which is harder to rip out than a chatbot.
What does not protect them much:
- plain AI review comments
- generic PR explanations
- CI debugging assistance
Those are exactly the kinds of capabilities frontier models keep absorbing.
Signals
Why this is in the blast radius
This is the cleanest overlap.
Anthropic says Opus 4.6 improved coding skills, code review, debugging, long-running agentic tasks, and reliability in larger codebases. Graphite sells AI PR reviews, PR chat for understanding diffs, and CI failure debugging.
That means Anthropic is directly attacking Graphite's AI assistant surface area, especially for teams willing to stay inside native GitHub plus Claude-powered tooling.
Sonnet 4.6 matters because it pushes strong coding capability into a cheaper, more broadly deployed default model with 1M-token context.
That makes repository-scale code understanding and review assistance easier to distribute widely, which weakens Graphite's AI review differentiation. It does not replace Graphite's merge queue or stacked PR workflow, but it makes the AI layer more commodity-like.
Graphite lives inside live developer workflows across GitHub, CI, editors, and collaboration surfaces.
Anthropic's push into computer use means Claude can increasingly act inside those tools directly, not just chat about them. That raises the risk that Claude becomes the active workflow operator for reviewing, triaging, fixing, and routing PR work.
It still does not equal Graphite's purpose-built stack and merge infrastructure, but it narrows the gap.
Graphite promises AI reviews that catch critical bugs before merge. Anthropic is showing Claude finding real high-severity vulnerabilities in a major production codebase.
That is a strong proof point that Claude can do meaningful code inspection beyond superficial comments. For buyers evaluating AI review tools, this undermines the idea that Graphite alone has special review intelligence.
This is a practical product overlap: GitHub Actions integration for automatic reviews on every PR.
Graphite's AI review product also lives on pull requests before merge. Claude Code shipping automated PR security reviews means Anthropic is not just model-capable in theory; it is shipping review automation in the same neighborhood.
It is narrower than Graphite overall, but very relevant to the AI review wedge.
This shows continued investment in Claude Code's tooling and enterprise controls, but the specific changelog items here are mostly about hooks, model settings, PowerShell support, and admin controls.
That signals platform maturity, yet it does not specifically overlap with Graphite's distinctive assets like stacked PRs, merge queues, or team review inboxes.