Skip to main content

shojin

Shōjin (精進): sustained, diligent effort toward improvement — the Japanese rendering of vīrya-pāramitā, and the everyday phrase (精進します) for "I will keep getting better." Investigations into whether and how AI agents improve themselves, run on hgym. The framework lives there; the science lives here.

Findings so far

  • Claude Code self-improved on business workflows through conversation, not memories (post, replication package) — told only to get better, Claude Code raised held-out success from 34% to 48% and built itself a directed knowledge base along the way. Held-out ablations locate the improvement in the conversation: the compaction summary carries most of it at less than a hundredth of the size, the memory notes add nothing measurable across nine comparisons, and stripping thought content changes nothing but the price.

Registrations

Pre-registered designs live in registrations/. A study is registered when its file merges without the DRAFT banner; every later change is a commit, so the freeze is git history itself.

  • Generation effect — DRAFT — does the act of writing memory notes contribute, independent of the notes ever being read?

Studies

Each study directory is self-contained: data, stdlib-only analysis, the as-run pipeline, provenance, and a re-run path pinning the hgym commit it targets.

Metadata

Release files for shojin 0.0.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for shojin 0.0.1
File Size Uploaded
shojin-0.0.1.tar.gz 4.4 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for shojin 0.0.1
File Interpreter ABI Platform
shojin-0.0.1-py3-none-any.whl Python 3 none any Details

Total release size: 6.6 kB

Release files / shojin-0.0.1.tar.gz

Download URL shojin-0.0.1.tar.gz
Size 4.4 kB
Tags Source
SHA-256 checksum
How to use checksums
8c5a9fef6e43b262411d0a6751cb09fa8fb32db735f52fb8204ef88d39b87d81
BLAKE2b-256 checksum
How to use checksums
4d7f42d3f15820aacc6d3410ad7aec83d0d914c19ca769ed0c815973ad87ec1a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.11.20 {"installer":{"name":"uv","version":"0.11.20","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release files / shojin-0.0.1-py3-none-any.whl

Download URL shojin-0.0.1-py3-none-any.whl
Size 2.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
1c5138daf5a7acfbb49a25e92f1cc94243a8629f654f27a224965e3399c4aff0
BLAKE2b-256 checksum
How to use checksums
045cd224408762a4ac126261aa73411d1777e8a01460e70386f9e51efa78106c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.11.20 {"installer":{"name":"uv","version":"0.11.20","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release history Release notifications | RSS feed

This release

0.0.1 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page