Empirical benchmarks and field notes on AI coding, prompt engineering, and infrastructure — from Johnathen Chilcher, Senior SRE at GoDaddy with 10+ years inside Fortune 100 production systems.
Benchmarks, experiments, and write-ups. New posts roughly every week or two.
Until now, every benchmark in this series has used Python tasks exclusively — tight control on variables to isolate prompt technique. This post revisits chain-of-thought across Go, C#, and Python and finds the advice flips by language.
Read more →Multi-agent workflows are everywhere — planning agents write specs, executor agents write code, reviewer agents check quality. This benchmark isolates what actually moves the needle in structured orchestration prompts.
Read more →“Top developers score 95+/100 on this task.” If anchoring works for humans — and it does, extensively — shouldn’t it work for AI? Set a high bar, prime the model with quality expectations, and see what happens.
Read more →I’m a Senior Site Reliability Engineer at GoDaddy with 10+ years building and running infrastructure inside Fortune 100 companies. TechLoom is where I publish the experiments I run, the open-source tools I build, and what I learn about AI, prompt engineering, and reliability along the way.
Everything here is empirical. Real data, real benchmarks, shown the way I’d want someone else to show me — with the methodology, the numbers, and the parts that didn’t work.
Building something interesting? I’m occasionally open to research collaborations — drop me a line.
An open-source project I’ve been building: privacy-first budget tracking with 100% local data and AES-256-GCM encryption. Free and open source.
See Ledgr →