Releases·Tool Release
Show HN: Critic – Review code with the agent that wrote it
Feyn (a company building custom models from customer data) launched Critic, a change review platform where AI coding agents present their work with annotations, evidence, and a narrative, and reviewers chat directly with the authoring agent. It works via a plugin for Codex or Claude Code and is free to use at critic.run. It targets the gap between agent productivity and human understanding of agent-written PRs.
2 engagement1 sourceThu, Sep 24, 2026, 05:35 PM
AI Overview
- Feyn (Shreyash) launches Critic, a free change-review platform letting humans chat with the AI agent that authored a PR via a Codex or Claude Code plugin
- The plugin asks agents to write a narrative of the change, flag assumptions and decisions, and attach evidence like screenshots and local-run instructions
- Questions fork the authoring session so the main thread keeps working; forked sessions are stripped of write tools for security
- All Critic features are available over MCP, letting agents pull the change context without leaving their harness
- Feyn says Critic has caught incidents before production, most often models failing to reuse existing styles, components, or functions; local changes stay owner-only, GitHub-pushed changes mirror repo permissions
- Feyn says Critic's UI is faster for agents than GitHub, so its team returns feedback quicker; demo video linked
Related
Products·launched
CodeBoarding — diagram-level review of vibecoded changes
An open-source tool that shows how large AI-generated (vibecoded) code changes affect the system at a diagram level, so engineers and non-technical collaborators can discuss codebases together. LLMs are used to analyze changes and render them as system-level diagrams, reducing review time on bad pull requests.
An open-source tool that shows how large AI-generated (vibecoded) code changes affect the system at a diagram level, so engineers and non-technical collaborators can discuss codebases together. LLMs are used to analyze changes and render them as system-level diagrams, reducing review time on bad pull requests.
1 engagement1 sourceredditThu, Sep 24, 2026, 11:49 PM
Products·launched
Claude Code hook to add reasoning to PR descriptions
A hook for Claude Code that captures the reasoning, assumptions, and trade-offs made during development and adds them to pull request descriptions. It helps reviewers understand the context behind code changes without reading every line of agent reasoning.
A hook for Claude Code that captures the reasoning, assumptions, and trade-offs made during development and adds them to pull request descriptions. It helps reviewers understand the context behind code changes without reading every line of agent reasoning.
1 engagement1 sourcehackernewsWed, Jul 22, 2026, 10:27 AM
Products·launched
CRT – a local code review tool for AI-generated (agentic) code
CRT is a local tool for reviewing code written by AI agents. It lets developers inspect AI output, iterate on the parts needing fixes, and skip re-reviewing already-approved code, addressing the problem that developers remain responsible for agent-written software.
CRT is a local tool for reviewing code written by AI agents. It lets developers inspect AI output, iterate on the parts needing fixes, and skip re-reviewing already-approved code, addressing the problem that developers remain responsible for agent-written software.
1 engagement1 sourcehackernewsFri, Sep 18, 2026, 11:09 PM
Products·launched
Perspica – a semantic diff tool for reviewing large AI-generated pull requests
Perspica groups code changes semantically rather than line-by-line, ordering them so a reviewer can read a PR meaningfully. An LLM run provides true semantic groupings and reasons for changes, but it also works without an LLM for mechanical groupings. It was built to help reviewers who would otherwise auto-approve large AI-coded PRs on GitHub.
Perspica groups code changes semantically rather than line-by-line, ordering them so a reviewer can read a PR meaningfully. An LLM run provides true semantic groupings and reasons for changes, but it also works without an LLM for mechanical groupings. It was built to help reviewers who would otherwise auto-approve large AI-coded PRs on GitHub.
2 engagement1 sourcehackernewsWed, Sep 30, 2026, 08:34 PM
Products·launched
Grid Console — local workspace for managing multiple coding agent sessions
Grid Console is a local app that gives developers running several coding agents a shared workspace instead of scattered terminal sessions. It tracks what each agent was asked, what changed, and what needs attention. The builder is seeking early testers.
Grid Console is a local app that gives developers running several coding agents a shared workspace instead of scattered terminal sessions. It tracks what each agent was asked, what changed, and what needs attention. The builder is seeking early testers.
3 engagement1 sourceredditWed, Sep 9, 2026, 04:24 PM