Automating Software Feedback Loops with Claude and Slack

Every2026-06-11go watch the original →

the gist

Kieran Klaassen automates a software feedback loop by using a scheduled Claude routine to ingest Slack feedback, generate a consolidated pull request, and execute fixes in batch using the LFG workflow.

The Automated Feedback Factory

The breakthrough involves transforming Slack into a structured input stream for software development by using a scheduled Claude Cowork routine to aggregate, classify, and resolve user feedback in batch. Instead of managing individual pull requests for every bug report, the system consolidates feedback into a single, comprehensive pull request that is processed and verified while the developer is offline.

Implementation Workflow

Feedback Capture: The team uses RiffRec, an open-source React wrapper, to record user interactions, network requests, and console errors directly from the application. These rich data packets are posted to a dedicated Slack channel.
Structured Ingestion: A scheduled Claude Cowork routine monitors the Slack channel via the Slack MCP, downloads attachments, and classifies feedback. It maintains a YAML-based state file to track which items are resolved and which require human intervention.
Batch Execution: The developer uses the LFG (Let’s Fix Group) workflow within Cursor to process the consolidated feedback list. The model iterates through the items, applies fixes, and generates video walkthroughs of the changes for review.
Continuous Improvement: The system utilizes the Compound Engineering framework to learn from previous mistakes. If an automated fix fails or is rejected, the system updates its internal strategy to avoid repeating the error in future cycles.

Context

The author developed this workflow to manage the high volume of feedback generated during the development of Cora, an AI-native email application. By batching 17 distinct feedback items into a single pull request, the developer significantly reduces the cognitive load of code review and allows the AI to perform complex refactoring tasks overnight. This approach shifts the developer's role from manual coding to managing a high-level feedback loop where the AI handles the implementation and verification of UI and functional improvements.