jevscript
Early language design where typed Jev judgments are a runtime operator, with an engine that batches judgments over shared state.
Overview
Jevscript is a planned programming language where judgment is a runtime primitive: programs ask Jev narrow typed questions (noul, choice, score) as they run, and policy such as thresholds and gates is ordinary code. Its engine, Hocket, is meant to parse and interpret the language and coalesce lazy judgments into one request per shared state, splitting only where a question depends on an earlier answer, with answers memoised on question identity and state hash.
The example asks for a ticket's kind and urgency and pages on-call when urgency exceeds 0.6 for an enterprise tier, with both judgments sent as one request. Only milestone M0 is done: a spike that confirmed the answer contract and measured a batching baseline, about 2.7 times fewer input tokens at a 230-character state. Later milestones (interpreter, coalescing and cache, policy library, examples, determinism suite) are planned in PLAN.md. It is a design experiment for language enthusiasts rather than a usable tool.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What is the main request in ticket.body? | bug · billing · feature · other |
| Yes / no | The message conveys urgency or time-sensitivity | — |
How it works
- Write judgments with ask blocks in Jevscript
- Hocket collects lazy judgments over the same state
- Coalesced judgments are sent as one Jev request
- Policy code branches on returned probabilities
Strengths
- Batching reduces input tokens over shared state
- Policy stays in ordinary code
- Memoised answers avoid repeat calls
Caveats
- Only a spike exists (M0); no interpreter yet
- Measured savings come from a small baseline
- Requires a TypeSafe key
At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- planned interpreter (spike in Node)
- Has code
- yes
- Has tests
- no
Use cases and integrations
How curated lists describe it
An early language experiment whose current implementation is a Jev request-batching spike.
logicrw/awesome-jev-projects
Frequently asked questions
Can I use Jevscript now?
Not yet; only the M0 spike is done and the interpreter is planned.
What is Hocket?
The planned engine that runs Jevscript and batches judgments into as few requests as possible.
What saving did the spike measure?
About 2.7 times fewer input tokens at a roughly 230-character state.
Related projects
DeepSeek Harness Web GUI plugin giving the agent a jev_ask tool, usage tracking and an optional command guard with explicit caveats.
PluginbetaTypeScriptBrowser tool that finds bounded decisions in LLM prompts and proposes Jev Choice, Score and Noul questions with SDK code.
ProjectbetaJavaScriptTiny JavaScript semantic matcher that classifies text locally in microseconds and escalates ambiguous cases to Jev.
ProjectbetaJavaScriptHack Club's lightweight LLM proxy, listed for adding a Jev forwarding endpoint behind its auth, limits and usage logging.
MCP serverproductionTypeScriptZilliz semantic cache for LLM queries, listed for a Jev evaluator that uses Noul checks to decide whether a cached answer can be served.
LibraryproductionPythonSimon Willison's LLM plugin that adds Jev as a model for yes/no, choice and score questions from the command line.
PluginproductionPython