Blog
Engineering
Featured
Engineering 11 min read
How AI Coding Agents Find Context: 7 Codebases Read
How do AI coding agents find context? We read the code of 7 open-source agents: none indexes your repo by default. They grep, read, then trim the window.
Read
Engineering 10 min read
Does Code Health Predict Bugs? 21 Repos, 9 Languages, ROC AUC 0.74
A reproducible defect-prediction study: 21 repos, 9 languages, 2,826 labeled files, ROC AUC 0.74, and 2.3x more defects caught under a fixed review budget.
ReadEngineering
Code Review at Agent Scale: Deciding Which Changes Get a Person
When merge volume grows 12x, reading every diff stops scaling. How to review AI-generated pull requests by risk, and what to automate, with real output.
Read9 min read
Engineering
AI Code Review Cost per Pull Request When Coding Agents Write the PRs
What AI code review costs per pull request when coding agents open 1,000 PRs a month: seat, credit, per-run and token billing worked out, with measured data.
Read9 min read
Engineering
Leiden Communities vs PageRank for Finding Code Hotspots
Leiden community detection maps the modules, PageRank ranks the central files, and git history finds the code hotspots. Real output from the repowise repo.
Read9 min read
Engineering
The Healthiest Open-Source Repos by Language, October 2026
Code health rankings for 147 popular open-source repos across Python, TypeScript, Go, JavaScript, Java, Rust, C# and C++, with the method behind each score.
Read7 min read
Engineering
Agent Memory Compared: How mem0, cognee and hindsight Are Built
mem0 stores a fact list, cognee builds a knowledge graph, hindsight forms observations and reflects. We read the write and read path of each, file by file.
Read10 min read
Engineering
Agent framework overhead: one tool call through five repos
We traced one tool call through LangChain, Google ADK, Composio, mem0 and GPT Researcher in their real code: how many steps each takes and what each one buys.
Read9 min read
Engineering
AI Code Provenance: How to Tell Which Code Agents Wrote
Three ways to track AI code provenance in git: commit trailers, tools that record agent sessions like Git AI, and inference from history, compared.
Read9 min read
Engineering
Cross-Repo and Cross-Language Dependency Graphs: Tools That Actually Do It
What a cross-repo, cross-language dependency graph really resolves: imports, HTTP routes, gRPC, queues, tables and packages, and which tools link each one.
Read9 min read
Engineering
Deterministic vs LLM Code Review: What Each Catches, and the Guard Pattern
What deterministic checks catch, what an LLM reviewer catches, what each costs per pull request, and how a guard lets one keep the other honest.
Read9 min read
Engineering
The First Commits of 20 Famous Open-Source Repos
The first commit of React, requests, Hugo, FastAPI and 16 more famous repos, checked against GitHub: three kinds of start, from an empty README to 43,123 lines.
Read6 min read
Engineering
Do Code Health Flags Find Buggy Files? 199 Repos, Past Fixes Taken Out
The 20 lowest-health files in a repo had a recent bug fix 51% of the time against 17% for a random file, with the score's own bug-fix input removed.
Read8 min read
Engineering
Healthy at Scale: What Chatwoot and Grafana Do Differently
Chatwoot keeps 369,000 lines healthy and Grafana 2.2 million. Across 45 big repos, the clearest difference is how little code sits in long files.
Read7 min read
Index your repo free.
See what repowise finds in your own code in minutes. Rolling it out to a team? We will walk you through it.