https://github.com/greynewell/greynewell

My personal README!
https://github.com/greynewell/greynewell

about-me architect aws aws-lambda blog cpp17 developer electronjs juce llamacpp llm nodejs reactjs typescript

Last synced: about 2 months ago
JSON representation

My personal README!

• Host: GitHub
• URL: https://github.com/greynewell/greynewell
• Owner: greynewell
• Created: 2023-10-25T18:17:36.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2026-02-25T20:56:07.000Z (3 months ago)
• Last Synced: 2026-02-25T22:44:54.431Z (3 months ago)
• Topics: about-me, architect, aws, aws-lambda, blog, cpp17, developer, electronjs, juce, llamacpp, llm, nodejs, reactjs, typescript
• Language: HTML
• Homepage: https://greynewell.com
• Size: 85.9 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Grey Newell

ML infrastructure engineer. Founding Engineer [@supermodeltools](https://github.com/supermodeltools), building code analysis tooling for AI agents. Creator of the [MIST stack](https://github.com/greynewell/mist-go). MS CS (ML) @ Georgia Tech. Ex-AWS.

## MIST stack

**M**atchSpec · **I**nferMux · **S**chemaFlux · **T**okenTrace

Go. Zero external deps. All six repos are pinned below.

| Repo | Purpose |

|------|---------|

| [`matchspec`](https://github.com/greynewell/matchspec) | Benchmark suites. Runs evals against any backend, produces structured reports. |

| [`infermux`](https://github.com/greynewell/infermux) | Inference routing. Abstracts providers, tracks tokens and cost per request. |

| [`schemaflux`](https://github.com/greynewell/schemaflux) | Data compiler. Pass pipeline, pluggable backends, no runtime allocs in hot path. |

| [`tokentrace`](https://github.com/greynewell/tokentrace) | Observability. Collects spans, computes latency percentiles, fires threshold alerts. |

| [`mist-go`](https://github.com/greynewell/mist-go) | Shared core. Protocol, transport, metrics, circuit breakers, checkpointing. |

Methodology: [eval-driven development](https://evaldriven.org).

## Research

[![ORCID](https://img.shields.io/badge/ORCID-0009--0001--0714--3800-A6CE39?style=flat-square&logo=orcid&logoColor=white)](https://orcid.org/0009-0001-0714-3800)

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.18627369.svg)](https://doi.org/10.5281/zenodo.18627369)

**[mcpbr: Benchmarking Model Context Protocol Servers on Software Engineering Tasks](https://doi.org/10.5281/zenodo.18627369)** — 500-task SWE-bench Verified evaluation showing MCP tools trade resolution rate for efficiency (42% fewer tool calls, 15% lower cost). Georgia Tech, 2026.

## Blog

- [SWE-bench Verified Is Broken: 5 Things I Found in the Source Code](https://greynewell.com/blog/swe-bench-verified-broken-5-things-source-code/)

- [SWE-bench Tests Run 6x Faster on ARM64 with Native Containers](https://greynewell.com/blog/swe-bench-arm64-native-containers-6x-faster/)

- [Why Code Graphs Matter for AI Agents](https://greynewell.com/blog/why-code-graphs-matter/)

- [Building Uncompact: Lessons from Production](https://greynewell.com/blog/building-uncompact-lessons-from-production/)

- [The Architecture of Supermodel's Code Graph API](https://greynewell.com/blog/supermodel-code-graph-api-architecture/)