About
About
I'm Alexandre Belostecinic, a software engineer who builds tools for working with AI agents. This site is where I publish research: experiments with frontier models, replications of published results, and the infrastructure needed to run them carefully.
How I work
- Check before claiming. A harness is validated against known results before it produces new ones.
- Report everything. Failures, deviations and changes of plan are written down, not left out.
- Release the data. Each study links to its code and full logs.
Contact
Find me on LinkedIn, or my code on GitHub. Corrections and replications of my own work are very welcome.