PulseAugur
EN
LIVE 22:49:08

I scanned 6 popular repos the way a coding agent reads them. LangChain's lockfiles outweigh its code.

By PulseAugur Editorial · [1 sources] ·

A developer analyzed six popular code repositories to understand how much data coding agents process that is not useful for learning. The analysis revealed that large lockfiles in monorepos like LangChain, extensive test data such as Shakespeare's complete works in Ollama, and large diagram files in FastAPI contribute significantly to the token count processed by agents. The developer created a CLI tool called ctxray to quantify these AI

RANK_REASON [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

I scanned 6 popular repos the way a coding agent reads them. LangChain's lockfiles outweigh its code.

How we ranked this

Signal score
5 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Samyak Vaghela ·

    I scanned 6 popular repos the way a coding agent reads them. LangChain's lockfiles outweigh its code.

    <p>Coding agents don't read a repo the way you do. You skim past <code>uv.lock</code> without thinking; a <code>grep</code>, a glob or a careless <code>cat</code> from an agent doesn't. I wanted to know how much of a typical repo is text a model pays for but can't learn anything …