We just open sourced Chelis, a programming language we built focused on agent authorship and human supervision. You can find a longer intro on our site: https://chelis.ch/blog/launch/ But here are the key things a HN reader might like to know:
* Statically typed, compiled, functional language, with heavy influences from ML family languages
* Built for agent authorship of programs across the compiler, MCP, skills, and remdiation paths
* Formal methods baked in via SMT and related solvers being connected with the type system in the compiler
* Rich support for numerical programs and mathematically rich domains like quant finance and quant econ
* GPU support is in an experimental state, for folks who want to build from source and run on AMD or Apple acceleration
* All built in Rust
From our experience, agents can already write Chelis as well as they can write Python. And they are a lot less overconfident in incorrect approaches with all the help Chelis provides.
Several of us on the core team are hanging out on HN. Drop us a comment if you check out the work.
Comments URL: [https://news.ycombinator.com/item?id=49993063](https://news.ycombinator.com/item?id=49993063)
Points: 1
source & further reading
github.com — original article
Argos –- A 1MB Rust Security Shim for Model Context Protocol Servers
Show HN: CloudGrip – Open-source AI proxy to cap LLM API budgets
We tried to reproduce LLM judge position bias