Loading page content…
Loading page content…
Hand-picked papers, lab write-ups and engineering posts. Each entry links to the original source with our own short summary — we never republish the articles themselves.
A technical guide to the parts that make up a coding agent, including its model, context, tools and control loop.
Why read it: Build a mental model of agent architecture before focusing on any one coding product.
The Pro Git book explains how Git represents file contents, directories and commits as objects.
Why read it: Use this primary reference to follow the object model behind Git history.
Anthropic describes building Scion as its infrastructure needs changed. It reports about 2,500 code changes merged per day in May 2026 and more than 30,000 per day in September; this is a company-specific case study, not a general baseline.
Why read it: See how a team reasons about source control bottlenecks when its workload changes sharply.
Anthropic reports that its typical engineer merged 8× as much code per day in Q2 2026 as in 2024. The article cautions that lines of code measure quantity, not quality, so the figure is not a direct measure of productivity.
Why read it: A first-party account of how AI changes engineering work, with an explicit discussion of what its productivity measure can and cannot show.
Prerequisites: Basic familiarity with software engineering workflows.
A visual explanation of Git's objects and references that helps connect familiar commands to the data they manipulate.
Why read it: A friendly way to understand Git's underlying model and reason about everyday operations.
An interview-led look at the product and engineering decisions behind Claude Code. The article is subscriber content; access may require a paid subscription.
Why read it: A product case study for discussing how an AI coding tool is designed and developed.
Anthropic's guide to choosing simple workflows or agent patterns, and to designing tools and evaluation. Published in 2024, so check newer sources for current platform details.
Why read it: A practical framework for thinking about agent design without assuming that every task needs a complex autonomous system.
Resources are curated manually by CORTEX administrators. The data model records a source and external id so automated discovery can be added later without redesign.