Analysis updated 2026-05-18
Study a working example of an evidence system that rejects unbacked claims in code.
Explore recreational Lisp programs that exploit homoiconicity, like self-printing programs and metacircular interpreters.
Learn how a formal review process between two AI models shaped a language design document.
| wondermonger-daydreaming/latent-lisp | norvig/paip-lisp | atlas-engineer/nyxt | |
|---|---|---|---|
| Stars | 0 | 7,453 | 10,911 |
| Language | Common Lisp | Common Lisp | Common Lisp |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 5/5 | 2/5 | 4/5 |
| Audience | researcher | researcher | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires SBCL installed and familiarity with Common Lisp to follow the code.
This repository is a collection of experiments in the Lisp programming language, split into two related parts. Everything is written to run on SBCL, a Common Lisp implementation, and the author treats a program running without errors as the basic test of correctness. The first part, called Mneme, is a small runtime built around one idea: a system should not be able to claim it verified something unless it actually holds proof of that verification. In plain terms, the code defines rules so that a bare statement like "I checked this" cannot be treated as a confirmed fact unless a matching certificate backs it up. The README is explicit that this protects against a cooperative caller using the intended interface, not against code with direct access to the internals, and that cryptographic protection is planned for later rather than already built. An included test suite reports a set of forged claims being correctly rejected alongside legitimate ones being correctly accepted. The second part, called the atelier, is a workshop of smaller, playful Lisp projects that explore homoiconicity, the property of Lisp where code and data share the same structure. Examples described in the README include programs that print their own source code, an interpreter written in the language it interprets, poems written as valid pieces of Lisp code, a recreation of the classic ELIZA chatbot, and a set of short experimental pieces exploring philosophical ideas through code. A newer part of the repository, added in July 2026, documents an attempt to design a more complete language specification for this style of system, including a review process where a second AI model critiqued the design and the author responded with revisions. Because this project centers on Lisp's unusual properties and experimental verification ideas, it will likely be most useful to readers already comfortable with Lisp or interested in formal approaches to trustworthy AI generated claims, rather than general programming beginners.
A collection of experimental Common Lisp projects: a runtime that requires proof before accepting a claim, plus a workshop of playful, self-referential Lisp code.
Mainly Common Lisp. The stack also includes Common Lisp, SBCL.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Verify against the repo before relying on details.