/dev/fast: architecture-level code review for AI-generated changes

/dev/fast builds Review, an open-source desktop app for code review in the era of coding agents. The buyer decision is whether a team needs architecture-level understanding of a large diff and its agent trace, rather than another line-by-line reviewer or a summary bot.

What it does

Review presents interactive visualizations, diffs, and agent trajectories in one review surface. The public demo describes sequence and database diagrams, chat with the authoring agent, and trace links that let a reviewer connect an implementation decision back to the change. The broader /dev/fast roadmap also names local trajectory storage, merge rules, code storage, and programmable CI (Review homepage; /dev/fast about).

Fact	What the public sources say
Buyer	Engineers and staff reviewers working with AI-generated or very large changes
Core product	Open-source macOS Review app for diffs, architecture views, agent chat, and trajectories
Review boundary	Review helps a human understand and critique code; it does not claim to replace approval
Availability	Desktop beta; hosted Review is described as coming soon
Founders	Sid Menon, Ketan Agrawal, Milan Bhandari, and Alex Iansiti

Why it fits

The product addresses a real shift in review work. When an agent can produce a thousand-line change in minutes, reading every line is both expensive and insufficient; the reviewer needs to understand the architecture the change touches, the assumptions the authoring agent made, and whether the resulting system still behaves coherently. Interactive diagrams and trace-linked context are aimed at that decision.

The tradeoff is local workflow and team adoption. The current public app runs on macOS, and the hosted collaboration layer is not yet described as generally available. A buyer should resolve repository permissions, secret handling, trace retention, generated-diagram accuracy, monorepo scale, CI integration, and whether a review artifact can be shared with the rest of the team. Pricing was not published in the reviewed sources.

The founders bring direct systems and AI-tooling context. YC describes Menon and Bhandari as former Palantir infrastructure engineers, Agrawal as a Robust Intelligence and Snowflake ML engineer, and Iansiti as a Pinterest and Flowcode engineer (YC company profile).

Short version: /dev/fast is worth trying on the hardest agent-generated change in a repository. It is most useful when a reviewer needs to see what the code means, not merely whether a linter passed.

Sources checked — 2026-09-19

Cohort context

/dev/fast is listed in Winter 2026. In our 2026-09-18 directory snapshot, 126 of 199 listed companies in that cohort have YC’s primary industry label B2B (63.3%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.

Public website snapshot

Observed 2026-09-19T16:19:59.520Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.

Signal	Homepage observation
Product description metadata	Observed
Canonical link	Not observed in this response
H1 or H2 heading	Observed
Typed structured data	Not observed in this response
Docs/developer link	Not observed in this response
Pricing link	Not observed in this response
llms.txt link	Not observed in this response
Markdown alternate	Not observed in this response

Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.