# Amplifying

> Amplifying is an AI benchmark research studio. We build evaluation frameworks for AI judgment, measuring what models recommend, not just what they get right.

Amplifying studies the subjective choices AI systems make: what tools to install, what products to recommend, what to build from scratch. We benchmark these choices systematically so the patterns become visible.

## Coding Agents Index

Independent measurement of coding-agent usage from output you can verify: commits and pull requests on public GitHub that carry an agent's signature, tracked daily since Claude Code launched in February 2025.

- [Coding Agents Index (Markdown, machine-readable)](https://amplifying.ai/coding-agents.md): the full current index as Markdown, rankings plus trends, autonomy, npm install demand, weekday-vs-weekend usage, Claude model mix, and language and industry leaders. Best starting point for an agent.
- [Coding Agents Index (rankings)](https://amplifying.ai/coding-agents): agents ranked by attributed public output, with a visibility class on every figure. Markdown: https://amplifying.ai/coding-agents.md
- [Trends](https://amplifying.ai/coding-agents/trends): weekly and daily agent-marked PR series, plus weekday-vs-weekend usage. Markdown: https://amplifying.ai/coding-agents/trends.md
- [Autonomy](https://amplifying.ai/coding-agents/autonomy): human review, time-to-merge, and test-inclusion for merged agent PRs. Markdown: https://amplifying.ai/coding-agents/autonomy.md
- [Languages](https://amplifying.ai/coding-agents/languages): where agent work concentrates by language. Markdown: https://amplifying.ai/coding-agents/languages.md
- [Industries](https://amplifying.ai/coding-agents/industries): where agent work concentrates by industry. Markdown: https://amplifying.ai/coding-agents/industries.md
- [npm install demand](https://amplifying.ai/coding-agents/npm): weekly CLI package downloads as a demand floor. Markdown: https://amplifying.ai/coding-agents/npm.md
- [Methodology](https://amplifying.ai/coding-agents/methodology): visibility classes, validation, and coverage limits.
- [State of the Coding Agent Market](https://amplifying.ai/research/state-of-coding-agents): the written analysis on top of the index. Markdown: https://amplifying.ai/research/state-of-coding-agents.md

## Research

- [What Claude Code Actually Chooses](https://amplifying.ai/research/claude-code-picks): A systematic survey of 2,430 Claude Code responses across 3 models (Opus, Sonnet, Haiku), 4 greenfield project types, and 20 tool categories. Finds that Custom/DIY is the #1 recommendation in 12 of 20 categories. Includes per-model breakdowns, phrasing sensitivity analysis, and greenfield-vs-established comparisons.
- [Full Report: Claude Code Picks](https://amplifying.ai/research/claude-code-picks/report): The complete write-up with detailed findings, methodology, and per-category analysis.
- [Why AI Product Recommendations Keep Changing](https://amplifying.ai/research/ai-product-recommendations): 792 product questions asked to Google AI Mode and ChatGPT. Reveals 47% cross-platform disagreement, Shopping Graph bias, and significant output drift over time.

## Tech Directory

- [Tech Directory](https://amplifying.ai/tech): Browse 20 tool categories and 80+ tools that Claude Code picks when building real projects. Each tool has a profile with pick rates, model-level breakdowns, and per-repo data.

Categories include: CI/CD, Payments, Deployment, Package Manager, Styling, Testing, ORM, Databases, Authentication, UI Components, State Management, Caching, Email, Observability, Forms & Validation, API Layer, File Storage, Background Jobs, Real-time, and Feature Flags.

## Methodology

All benchmarks use open-ended prompts (no tool is named), real or realistic repositories, and multiple models. Tool extraction uses an LLM-as-judge pattern. State is isolated between prompts via git reset. Full methodology: https://amplifying.ai/research/claude-code-picks/report#methodology

## For Vendors

- [For Vendors](https://amplifying.ai/for-vendors): Intelligence reports for dev tool vendors showing how AI coding agents position their tool — competitive analysis, model trends, and agent verbatim quotes.

## Contact

- Website: https://amplifying.ai
- Email: team@amplifying.ai
- GitHub: https://github.com/amplifying-ai

## Optional

- [Privacy Policy](https://amplifying.ai/privacy)
- [Contact Page](https://amplifying.ai/contact)