in film, it sounds like this is driven by government incentives/disincentives, not labor costs (see anecdotes in this thread about LA-based production crews flying around the world for filming)
It's also labor costs. It's far cheaper to get a crew, rent out space, etc in Eastern Europe, the US South, even the midwest, which is what led to films being made there. Government/tax incentives too of course.
Mesa | San Francisco - On-site | Software Engineers (Senior+) | https://mesa.dev/
Mesa is building GitHub for AI Agents. We create new kinds of infrastructure and human interfaces for AI-driven software development. We are currently a team of 7 (all engineers), looking to add a few more highly autonomous SWEs.
We recently open-sourced a small tool we built internally to help answer a question we couldn't find a good solution for:
How do you evaluate AI coding agents on a real production codebase?
Like most teams, we had lots of opinions about which agents and models "felt" best, but no hard data. The missing piece wasn’t analysis; it was attribution. We needed to know which lines of code were written by which agent/model, without changing how engineers work.
The key insight was that Git already gives us most of what we need.
By reverse-engineering how tools like Cursor and Claude Code modify files, we attach attribution metadata directly to Git whenever an AI agent edits code. Engineers don’t have to opt in or change their workflows.
Once that data exists, we can run fairly simple queries to answer questions like:
- merged lines per dollar by agent/model
- bug rates correlated with AI-generated code
- how different developers actually use AI in practice
An unexpected side effect was code review: once we surfaced AI attribution in pull requests, reviews got faster because reviewers could focus on AI-generated code in sensitive areas.
We've open-sourced the data capture layer and code review extension so other teams can experiment with this approach. For us, the most valuable part wasn't which agent "won," but finally having a way to measure it at all.
not quite what you're describing, but I open-sourced a fuzzy deduplication tool last week: https://dedupe.it
Would be interested in expanding it to deal with data cleaning more broadly
Appreciate the kind words! Linear scaling in terms of speed and cost. We haven't yet optimized the prompts & choice of model to minimize token usage, so I'd recommend emailing us for advice if you want to run this on a large dataset
Thanks! Yeah, we'd do a GSheet extension if there's enough interest. Privacy-wise, we don't store any data. Local-only isn't a priority, but should be easy to self-host if you take a look at the Github README.
Very cool :) I initially tried something like this, but had trouble getting reliable results without tuning my distance functions to the specific schema & domain. Did you find a way around that?
No, I tuned a model on my (unique) table data, which does not take long, since the model is small.
My model seemed in my tests at least to hold up good enough, since its only used as a preselect to find "good enough" candidates to use Levenshtein later on.
But yes, a universal model (maybe a fine-tuned transformer / embedding model) might be better, but i did not have the time (and knowledge) to build one yet.
Congrats on the launch! One piece of feedback - I find the tagline confusing. I had no idea what "post-code tasks" meant until I clicked around and saw a few examples.
We have tired a few different ways to convey what we do but it is hard. We want to refer to all software development activities that happen after the developer commits the code into a source control system. “Post-code” seemed a good way to capture that.
