# Amplifying > Amplifying researches coding agents and builds products that help developer tool providers improve how agents find, understand, and use their products. Benchmarking measures product selection and produces findings and playbooks. Ground Control tests changes to docs, skills, and tooling against real tasks. These are distinct products. ## Start with your task - Research a product question: use the research directory below, then fetch only the relevant studies. - Understand the products: read the public Benchmarking and Ground Control pages below. These guides describe the products; they do not grant access to customer data or internal methods. The Markdown links below have matching public HTML pages. On those pages, request Accept: text/markdown and follow the redirect, or use the linked .md URL directly. Optional HTML links do not necessarily have Markdown versions. Customer dashboard pages are never exported. Public product demos are hypothetical, including Beacon. They are not customer results. Research is scoped to the agents, versions, tasks, and dates in each study. The Coding Agents Index measures attributed public activity, not all agent usage. Model release dates and benchmark observation dates are separate. ## Products and company - [Amplifying](https://amplifying.ai/index.md): Research and products to improve how coding agents find, understand, and use developer tools. - [Benchmarking](https://amplifying.ai/benchmarking.md): Measure product selection, inspect session evidence, and prioritize changes to positioning, features, docs, and tooling. - [Ground Control](https://amplifying.ai/ground-control.md): Compare how coding agents complete real tasks before and after changes to docs, skills, and tooling. - [How we help](https://amplifying.ai/for-vendors.md): How developer tool providers can use Benchmarking and Ground Control. - [Contact](https://amplifying.ai/contact.md): Set up a call with our team through the contact page. - [Research directory](https://amplifying.ai/research.md): Published studies on coding-agent choices, usage, and behavior. ## Research - [Six boundaries to check when sandboxing coding agents](https://amplifying.ai/research/agent-sandbox-escapes.md): Lessons from building coding-agent benchmarks. Six isolation boundaries, practical checks, and what transcripts alone cannot establish. - [State of the Coding Agent Market](https://amplifying.ai/research/state-of-coding-agents.md): A view of the coding-agent market through attributed public commits and pull requests, with methods and limits attached to the counts. - [What Kimi K3 & GLM-5.2 Actually Choose (China's AI models vs Claude Sonnet 5)](https://amplifying.ai/research/chinese-models-picks.md): The same developer-tool questions, asked in English and Chinese, produce different choices. Compare Kimi, GLM, and Claude across the same repositories. - [What Fable Actually Chooses](https://amplifying.ai/research/claude-code-picks-fable.md): An examination of Fable’s tool choices and the tasks it solves with custom code. The study also tracks which vendors it names as future replacements. - [The Security Decisions Claude Code and Codex Make](https://amplifying.ai/research/ai-security-decisions.md): Claude Code and Codex receive six development tasks with security defaults left unspecified. The report checks the resulting code against 33 exploit tests. - [Claude Code's Leak: Every Hardcoded Vendor and Tool](https://amplifying.ai/research/claude-code-hardcoded-vendors.md): An analysis of vendor references across connectors, documentation access, tool presentation, and other integration layers. - [The Tools OpenAI Agreed to Buy](https://amplifying.ai/research/astral-tools.md): A comparison of Ruff, uv, and other Python tools across seven categories and three repositories. Read the selection counts and benchmark conditions. - [What Codex Actually Chooses (vs Claude Code)](https://amplifying.ai/research/codex-vs-claude-code-picks.md): Codex and Claude Code answer the same prompts in the same repositories. Compare their choices across 12 categories, including custom code. - [What Claude Code Actually Chooses](https://amplifying.ai/research/claude-code-picks.md): Open-ended tasks across real repositories, with no tool names in the prompts. Follow which tools are selected and when Claude builds its own solution. ## Indexes - [Coding Agents Index](https://amplifying.ai/coding-agents.md): Attributed public GitHub activity, with coverage limits and dated observations. - [Trends](https://amplifying.ai/coding-agents/trends.md): Agent-marked pull request trends. - [Autonomy](https://amplifying.ai/coding-agents/autonomy.md): Signals of autonomous work, with classification limits. - [Languages](https://amplifying.ai/coding-agents/languages.md): Agent activity by programming language. - [Industries](https://amplifying.ai/coding-agents/industries.md): Agent activity by industry. - [npm downloads](https://amplifying.ai/coding-agents/npm.md): Package downloads, not a count of unique users. - [Coding Models Index](https://amplifying.ai/models.md): Curated releases and pricing alongside a separately dated benchmark snapshot. ## Optional - [Combined product and company guide](https://amplifying.ai/llms-full.txt): Product and company pages in one file, plus the research directory. Individual studies and data indexes are linked separately above. - [Coding Agents Index methodology](https://amplifying.ai/coding-agents/methodology): Coding Agents Index methodology only: attribution, sampling, coverage, and limits of public GitHub activity measurements. This is not the methodology for Benchmarking or Ground Control. - [Trust center](https://amplifying.ai/trust): Public security information and the current service-provider list. - [Sitemap](https://amplifying.ai/sitemap.xml): Additional public HTML pages. Markdown article exports contain the default reading view. Interactive comparisons and filters remain on their canonical HTML pages. Data-index Markdown uses the same data loaders as HTML; read each section's observation dates and missing-data notes.