v1.0.0 — What actually matters across N files

WhatTheDiff v1.0.0 🔍

What actually matters across N files. The full intent.md vision is shipped — plus two capabilities that weren't even in the original spec.

Point it at a folder. It decomposes every file into stable semantic primitives, hashes them, and tells you what they agree on (consensus), how far each one deviates (drift), which is the outlier, and which form families (factions) — with the exact evidence behind every claim. Deterministic engine; an AI only ever explains what the engine proved.

Reads everything

• Text formats: JSON, YAML, XML, INI/env/TOML, Markdown, plain text
• Documents: PDF (text extraction, FlateDecode-compressed streams included)
• Binaries/executables: SSDeep-class fuzzy analysis via content-defined chunking — clusters executables into variant families by shared code (ELF/PE/Mach-O/Wasm detected)

Highlights

• Cross-format unification — the same fact in JSON, YAML, INI, or XML hashes to the same identity, so mixed-format corpora find real consensus
• Faction detection — groups that deviate together, with their defining signature
• wtd ask — grounded AI explanation (Anthropic API, OpenRouter, or a local model via Ollama); it can never invent a finding
• --keys-only — secret-safe schema comparison for credential/.env profiles; no value ever reaches the report
• Million-file scale — ~22 µs/file, 3.8 GB RSS at 1,000,000 files, exact detection (20,000/20,000 planted outliers, zero false positives)

Quality

Deterministic (byte-identical reports), zero dependencies, 67 tests + 1,065 property iterations, six static platforms (Linux/macOS/Windows × x86_64/aarch64).

Install

curl -fsSL https://raw.githubusercontent.com/copyleftdev/whatthediff/main/install.sh | sh

Windows: irm https://raw.githubusercontent.com/copyleftdev/whatthediff/main/install.ps1 | iex

All binaries below are checksum-verified against SHA256SUMS.