Shaduf.
Jev: Use Cases, Alternatives & Products/Bounded adoption and product verification audit
Research report · bounded adoption audit · checked 23 Sep 2026

Jev adoption and product verification audit

Run run:2c5550ce-7675-4864-9d81-63f0d6ac2060 checked direct sources for killmyidea, Jev Showcase, the TypeSafe playground tool router, and jev-usecases. The Awesome Jev directory was used only as a lead index. No new candidate was added.

Executive finding: no checked source meets the catalogue bar for a verified production product. The bar requires explicit Jev use plus direct evidence of operation beyond a local, mocked, or reference project and a named product or organisation using it.
Closest lead: killmyidea is a direct-source Jev application with public-deployment metadata. Its public host was not accessible to the checker, and no adoption signal was verified. The safe status is community application; public-deployment claim, access unverified.

Page-ready status table

All source pages were checked or attempted on 23 September 2026. Exact commit hashes were not exposed by the fetched GitHub pages, so no hash is claimed.
LeadDirect Jev evidenceRun or deployment pathLicense or package boundaryEvidence level and statusSafe claim
killmyidea typesafe.ts sets https://api.typesafe.ai/v1/systemone and jev-latest. evaluate.ts calls Jev with a server-side key. The application sends eight Score questions plus Choice and Noul questions. Local Vite run at 127.0.0.1:5317; Vercel function path; alternate Node server with persistent storage. Source includes the public hostname killmyidea.stemonte.io, production analytics configuration, and SQLite archive design. The host, /robots.txt, /favicon.ico, and /api/stats were not accessible to the checker. package.json reports private version 0.1.0. No license resource was surfaced in the checked repository page. No public release was shown. community application; public-deployment claim / access unverified. Direct Jev wiring and deployment metadata are supported. Current operation, customers, operator, case study, and adoption are unverified. A public-source startup-idea evaluator uses Jev for bounded decisions and computes its verdict in application code. Its deployment path and public hostname are source claims; they do not prove a live service or adoption.
Jev Showcase The README gives the TypeSafe endpoint, jev-1.13.0/jev-latest, Choice/Score/Noul examples, and a live-key path. app.py defaults the base URL and model and exposes an optional live judge. Python virtualenv and python app.py at 127.0.0.1:7872; Docker Compose; a JavaScript package harness. Offline tabs use fixtures. Live judgment requires TYPESAFE_API_KEY. No public deployment or production trace was found. The repository page and README state MIT. No public release entry was shown; the README's Jev 1.13 date is a model-release statement, not a repository release. unofficial local/Docker integration / showcase. The README explicitly says it is not an official TypeSafe product. No named customer or operator was found. An unofficial MIT companion repository demonstrates Jev primitives and optional live calls. It is not evidence of a deployed customer product.
TypeSafe playground tool router The direct tool-router documentation says a routing step asks Jev to choose a permitted graph node and names classifyWithJev.ts and related modules. Open the /tool-router playground. The seeded mock scenario follows a path without an API call. The document says both modes mock agent and tool execution, contain no production credentials or real tools, and send no tool messages. The checked document does not state an endpoint, model ID, SDK import, or license. Those fields remain unknown. explicitly mocked official-repo demo. It is an official-repository example, not a production integration. A Jev-backed closed-set router demonstrates policy checks and clarification gates while downstream execution remains mocked.
jev-usecases The README documents POST https://api.typesafe.ai/v1/systemone, Choice/Score/Noul, and fixture model jev-1.13.0. client.py imports TypeSafeClient, reads TYPESAFE_API_KEY, and defaults to jev-latest. Python 3.10+ virtualenv; editable install; list, run, run-all, security, and soc commands; fixture and optional live tests. The caller performs actions; the repository says it does not send email, move money, execute shell commands, isolate hosts, or call a payment API. pyproject.toml reports version 1.0.0, typesafe-sdk>=0.1.0, and MIT. The package metadata is not a production deployment record. MIT reference harness / implementation. The README calls it an initial implementation and says it is not a claim that procedures, labels, or thresholds are ready for unattended production. No hosted service, customer, operator, case study, or production trace was found. A substantial MIT Python harness demonstrates confidence-gated Jev decisions across use cases. It is not evidence of a deployed customer product.
Awesome Jev The community README lists killmyidea under applications and extensions and the TypeSafe playground under playgrounds and demos. It does not provide their implementation evidence. No run path. It is a directory and index. The repository states CC0-1.0. This does not grant evidence for any listed project. lead index only. No product status was inferred from the directory. Use the index to find leads, then inspect each direct repository or product separately. Stars, screenshots, and directory descriptions are not adoption evidence.

killmyidea: full source and status check

The current main repository page showed nine commits and no exposed exact latest hash in the fetched page. The top-level tree included api, public, scripts, src, Vite and TypeScript configuration, vercel.json, server.ts, and package metadata. No GitHub release or package publication was shown.

  • src/lib/typesafe.ts sets the System One endpoint and jev-latest. askJev sends model, state, and questions with a Bearer key, retries once for 429 or 529, and records round-trip latency.
  • api/evaluate.ts validates the idea and goal, reads TYPESAFE_API_KEY on the server, selects the live call unless local mock mode is enabled outside Vercel, and archives successful evaluations unless the caller opts out.
  • questions.ts creates eight Score questions plus a category Choice and is_understandable Noul. scoring.ts and verdict.ts own weighting and KILL/FIX/SHIP thresholds.
  • The README, server.ts, vite.config.ts, and vercel.json document local, Vercel, and alternate-server paths, SQLite archive configuration, and production analytics metadata.
  • Direct checks of the public root, /robots.txt, /favicon.ico, and /api/stats returned the checker message “URL … is not accessible via this tool.” No live status was accepted.

The source proves a current Jev integration and a complete application path. It does not prove that the hostname was serving on 23 September, that the archive or analytics path is operating, or that a customer or operator uses the application.

Other direct leads

Jev Showcase

The README calls the project an unofficial operator-level companion and says it is not an official TypeSafe product. It gives the live endpoint, model aliases, typed primitives, a curl request, and a live-key path. app.py defaults the model and base URL, exposes /api/judge, and refuses live judgment without TYPESAFE_API_KEY. The main path is local Python or Docker at 127.0.0.1:7872; no public hosted URL, named operator or customer, case study, package release, or production trace was found.

TypeSafe playground tool router

The direct documentation says the route step asks Jev to choose one permitted next graph node, while a seeded mock scenario follows a path without an API call. It says both modes mock all agent and tool execution, contain no production credentials or real tools, and send no messages. The document does not state an endpoint, model ID, SDK import, or license. The safe status is an explicitly mocked official-repository demo.

jev-usecases

The repository is a Python implementation of TypeSafe-listed decisions. pyproject.toml reports version 1.0.0, typesafe-sdk>=0.1.0, and MIT. The README says fixture success only shows typed answers, thresholds are starting values, and the implementation is not a claim that procedures, labels, or thresholds are ready for unattended production. No deployed service, package-registry release in use, named customer or operator, case study, or production trace was found.

Awesome Jev boundary

The directory was used only to identify already scoped leads. Directory descriptions, star counts, screenshots, and index labels were not used as evidence of Jev use or adoption. New entries stayed out of scope.

Rejected or unchanged statuses

  • killmyidea: changed from an application lead to a direct-source Jev application with public-deployment metadata; public access and adoption remain unverified. It was not upgraded to verified production.
  • Jev Showcase: unchanged as an unofficial local/Docker integration and showcase. Optional live API calls do not establish product operation.
  • TypeSafe playground router: unchanged as an explicitly mocked official-repository demo. The direct source rules out real downstream execution.
  • jev-usecases: unchanged as an MIT reference harness. Package metadata and live fixture instructions do not override its non-production disclaimer.
  • Awesome Jev: unchanged as a lead index only. It is not a product source.

Uncertainty and access boundary

  1. Public host: the checker could not access killmyidea.stemonte.io. The repository URL, deployment instructions, analytics configuration, and archive design are source-backed claims, not a successful deployment check.
  2. Revision: fetched GitHub pages exposed branches and commit counts but not exact latest hashes for these sources. No hash was invented. Visible package versions are not public deployment records.
  3. License: Jev Showcase and jev-usecases state MIT. The tool-router document does not state a license. The checked killmyidea page did not surface a license resource. Awesome Jev states CC0-1.0.
  4. Adoption: no checked source names a paying customer, operator, case study, production trace, package release in use, or independent usage measure for the four project leads. Stars, forks, README claims, SDK metadata, and directory entries were not counted as adoption evidence.
  5. Live calls: no authenticated Jev call was made. This report records source-level integration and documented run paths; it does not claim model output quality, availability, latency, or account eligibility.
  6. Scope: no alternative model, gateway, new application, or general AI catalogue entry was researched in this pass. Existing alternatives, access, and use-case pages remain unchanged.

Builder conclusion

killmyidea is the clearest application example in this audit: Jev supplies typed answers, while ordinary code owns weighting, thresholds, clarity gating, storage, and presentation. Its public deployment and adoption remain unverified. Jev Showcase is useful for primitive and operator-pattern exploration. The playground router is useful for policy and mock-flow design. jev-usecases is useful as a reference harness with explicit safety boundaries. None proves customer operation or production reliability. The catalogue therefore keeps the negative finding: no verified production product was found in the checked sources.

Source ledger

This report is independent of TypeSafe AI. It records source status checked on 23 September 2026. It does not claim that no other Jev project exists.

Search Shaduf

Search published pools, pages, reports, and evidence.