GitHub Repository

Shared memory across your team and your AI agents — with judgment about what's worth keeping.

74 starsTypeScript

VetoBench – benchmarking AI memory beyond retrieval

Name: VetoBench – benchmarking AI memory beyond retrieval
Availability: InStock
Author: mart1adelina

by mart1adelina·Jul 8, 2026·2 points·0 comments

Visit Project View on HN

AI Analysis

●●●BangerBig BrainSolve My Problem

Tracks rejected decisions so agents stop proposing approaches your team already vetoed.

Strengths

•Structured veto storage prevents agents from proposing approaches the team already rejected.
•VetoBench shows agents repeat rejected decisions nine out of ten times without memory.
•Passive capture works with Cursor and Copilot without requiring manual tagging workflows.

Weaknesses

•Self-hosted Postgres and Docker setup adds ops overhead for small teams.
•Relies on agents respecting system prompt instructions to check the veto store.

Post Description

VetoBench is an open-source benchmark that measures whether AI agents repeat engineering decisions a team has already rejected. Instead of evaluating retrieval accuracy, it evaluates decision quality: does the agent propose a previously vetoed approach, and does it recognize when a memory is no longer valid? Everything is reproducible, including the benchmark, fixtures, and evaluation results. Feedback is very welcome.