Diego Cabello

I sell codebase auditing. My individual work includes highly performant utilities measured in time in seconds and size in bytes.

Auditing

Codebase deduplication for messy codebases. The first half of the deduplicator pipeline is dozy (below), but to be really useful it needs automated candidate selection, which is the second half I've kept proprietary
Blame analysis for good or bad contributors
Dependency analysis (circular dependencies, etc.)
Insights into where aggressive CPU parallelization can streamline pipelines (uses code generators)
Recommendations as to where formally verified parsers can be used
Those are for decent codebases; for poor codebases the vast majority will be just cutting out crap
Two pricings: a written report, or a report and implementation

nab
- A unix-style single-file diffing utility
- Overall best on Splicebench
- Repo
- try it out: git clone https://gitlab.com/diegocabello/nab && cd nab && make
- Write-up coming soon :(
Splicebench
- A benchmark for measuring patch speed, rebuild speed, and file size for diffing utilities. Made for demonstrating the advantages of nab because there was no pre-existing benchmark for this category
- Bench and about bench lives at splicebench.dev
dozy
- Best-in-class (for CPU-bound, as opposed to GPU-bound) duplicate detector in accuracy and speed on BigCloneBench (the standard benchmark for these)
- Benchmark results, methodology, and write-up haven't been compiled yet, so for now you'll have to try it out yourself
- Repo

git sstat
- a better view for git diff --stat command
- Repo
Other cool stuff that I've made but don't have publishable materials ready for yet include a file format that beats JSONL on parsing speed for common tasks and a granular locking utility that enables swarms of agents to coordinate.