Jev in 25 Lines of Python
Blog post building the smallest working Jev-style decision loop in 25 lines of Python, with a busy HN debate.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
Jev in 25 Lines of Python is a blog post on nobodywho.ai that builds the smallest working Jev-style decision loop in Python. According to the list summary, it drew a Hacker News thread with 464 points and 139 comments, where commenters argued over whether the decision step needs a dedicated model at all.
Only the summary was collected, so the code and specific arguments are not reproduced here.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Does a decision step need a dedicated model?(inferred) | — |
How it works
- Build a minimal state-plus-question loop in Python
- Read typed answers and act
- Discuss trade-offs versus general LLMs
Strengths
- Compact illustration of the decision-model pattern
- Lively public discussion
Caveats
- Summary only; code not collected
At a glance
- Type
- Article
- Status
- discussion
- Language
- Python
Use cases and integrations
How curated lists describe it
Blog + HN thread (464 points, 139 comments): builds the smallest working Jev loop in Python, and the thread argues over whether the decision step needs a dedicated model at all.
yibie/awesome-jev
Frequently asked questions
What does the post show?
A minimal decision loop in about 25 lines of Python.
What did HN debate?
Whether the decision step needs a dedicated model at all.
Where is it?
On nobodywho.ai.
Related projects
Independent article walking through Jev and separating TypeSafe's published claims from public evidence.
ArticlediscussionWavect's independent review of Jev, whose four caveats a directory maintainer adopted into its IO contract section S8.
ArticlediscussionHarnessRouter blog post explaining what a System One model is, noted by a curator for consistent terminology.
ArticlediscussionChinese X thread rounding up sixty Jev use cases and twelve standouts, framed as: the LLM writes, Jev classifies, scores and chooses.
DiscussiondiscussionX thread by Moritz Kremb curating the strongest Jev projects posted within 72 hours of launch.
DiscussiondiscussionJapanese X post from LayerX about an urgent internal Jev study session for 50+ engineers, with published materials.
Discussiondiscussion