AI Roundtable: Tool for Comparing 200+ AI Models on Structured Questions

AI Roundtable is a web-based tool that allows users to compare responses from multiple AI models on structured questions. The tool was created following discussion around the "Car Wash Test" post on Hacker News.
Key Features
The tool provides several specific capabilities:
- Question Setup: Users type a question and define answer options
- Model Selection: Choose up to 50 models at a time from a pool of 200+ models
- Consistent Testing Conditions: All models answer independently under identical conditions with no system prompt, structured output, and same setup for every model
- Debate Feature: Run a debate round where models see each other's reasoning and get a chance to change their minds
- Reviewer Model: A reviewer model summarizes the full transcript of responses
- Access: No signup required, free to use
- Infrastructure: All models are routed via Opper (the creator's startup)
Practical Use
This type of tool is useful for developers working with AI agents to systematically compare model performance on specific questions or scenarios. By providing identical conditions across all models, it enables more objective comparisons than manual testing. The debate feature allows observation of how models adjust their reasoning when exposed to alternative perspectives, which can be valuable for understanding model behavior in collaborative or iterative contexts.
The creator is actively seeking feedback from the community and has made the tool available for immediate use without registration requirements.
📖 Read the full source: HN AI Agents
👀 See Also

Karpathy's autoresearch project: AI agents run overnight LLM training experiments
Andrej Karpathy released a minimal autoresearch project where an AI agent edits train.py, runs 5-minute nanochat training experiments, checks if val_bpb improved, and repeats overnight on a single GPU.

Custom WhatsApp Channel Plugin for Claude Code Using Baileys
A developer built a custom channel plugin that adds WhatsApp support to Claude Code 2.1.80+ using Baileys v7, implementing the WhatsApp Web Multi-Device protocol as an MCP server with the experimental claude/channel capability.

Five Free Claude Desktop Extensions Released: Inspector Lite, Graph Lite, Bible Code, Word Graph, and Fun Pack
A developer has open-sourced five local Claude Desktop extensions: Inspector Lite for semantic code search, Graph Lite for a personal knowledge graph, Fun Pack for entertainment features, Word Graph for Bible study, and Bible Code for pattern detection. All run locally with no external dependencies or additional API keys.

Driftwatch V3 Released: AI-Assisted Codebase Monitoring Tool
Driftwatch V3 is now available as a public repository after a 5-6 day build involving approximately 9,000 lines of code and $160 in API credits. The in-browser tool tracks markdown file issues, flags contradictory instructions, and provides cost tracking with recommendations.