diff --git a/.impeccable/surfaces/site-docs-html.md b/.impeccable/surfaces/site-docs-html.md index 6a42e0e..14b236c 100644 --- a/.impeccable/surfaces/site-docs-html.md +++ b/.impeccable/surfaces/site-docs-html.md @@ -197,3 +197,7 @@ add useful information for interpreting the result. Draft/source pass: Codex agent /root/portfolio_audit. Independent source review and converged browser acceptance remain with the integration owner. + +## Evidence navigation repair — October 1, 2026 + +The evaluations page links workflow screening directly to its recorded study and sends application integration readers to the current docs integration section. Preserve historical measurements, sources, dates, and limitations. This is a navigation repair, not a new model review. Independent reviewer: Codex AI agent `/root/ghostget_editorial`. diff --git a/.impeccable/surfaces/site-introducing-sys1-html.md b/.impeccable/surfaces/site-introducing-sys1-html.md index d4cf432..006241e 100644 --- a/.impeccable/surfaces/site-introducing-sys1-html.md +++ b/.impeccable/surfaces/site-introducing-sys1-html.md @@ -54,3 +54,9 @@ only when readers need it to interpret a result. Draft/source pass: Codex agent /root/portfolio_audit. Independent source review and converged browser acceptance remain with the integration owner. + +## Evergreen editorial cleanup — October 1, 2026 + +Keep the Hraness byline and accurate article metadata. Omit visible publication dates and the old release-at-publication badge; installation links lead to the current release. Name the recorded Codex AI reviewers in the disclosure. Historical study dates and negative outcomes remain beside their measurements. + +Independent editorial review: Codex AI agent `/root/ghostget_editorial`. This is a bounded copy review, not a new model study or human review. Final aggregate and rendered verification belong to the integration owner. diff --git a/.impeccable/surfaces/site-skills-html.md b/.impeccable/surfaces/site-skills-html.md index 9acd7b1..6b84ba8 100644 --- a/.impeccable/surfaces/site-skills-html.md +++ b/.impeccable/surfaces/site-skills-html.md @@ -248,3 +248,9 @@ only when readers need it to interpret a result. Draft/source pass: Codex agent /root/portfolio_audit. Independent source review and converged browser acceptance remain with the integration owner. + +## Evergreen editorial cleanup — October 1, 2026 + +Keep one command-reference link after installation. Adjacent links to the same destination add no navigation value. Existing evidence, installation requirements, and optional model-assisted workflow limits remain intact. + +Independent editorial review: Codex AI agent `/root/ghostget_editorial`. This is a bounded copy review, not a new model study or human review. Final aggregate and rendered verification belong to the integration owner. diff --git a/docs/cli-updates.md b/docs/cli-updates.md index e362168..6259b7d 100644 --- a/docs/cli-updates.md +++ b/docs/cli-updates.md @@ -1,7 +1,7 @@ # Update Sys1 -Automatic updates require Sys1 0.19.0 or newer. After that release is -published, upgrade an older installation once through its package manager. +Automatic updates require Sys1 0.19.0 or newer. Upgrade an older installation +once through its package manager to enable them. Supported Bun and npm global installations on macOS and Linux check for a newer release at most once a day before a command starts. Automatic updates are diff --git a/docs/launch-editorial.md b/docs/launch-editorial.md index e6a8611..ea6930a 100644 --- a/docs/launch-editorial.md +++ b/docs/launch-editorial.md @@ -190,3 +190,9 @@ This revision replaces the nine-beat catalogue and repeated reference material w Drafted by Codex AI agent `/root/launch_finish`; independently reviewed by Codex AI agent `/root`. No human review is claimed. Root checked the commands and workflow against source, the compact-output figures against the published replay, the whole-task limits against the reports, and the completion baseline’s actual scope. The revised prose makes no claim of proven whole-task savings or review accuracy. Preview instructions match the command’s displayed input type and model route. Root inspected the rendered desktop launch page, mobile homepage, skills page, documentation, and share image. The final local browser sweep passed 80 route/width/theme combinations, plus interactions, no-JavaScript reading, reduced motion, and video playback. The browser was Playwright 1.58.2’s Chromium 145.0.7632.6, with verified cleanup. Aggregate validation passed 584 tests and the packed CLI check; native installation passed on macOS arm64. Publication and exact production verification remain the integration owner’s next delivery steps. + +## Evergreen editorial cleanup (October 1, 2026) + +The article header keeps its Hraness byline and current installation links, removes visible publication dates and the old release badge, and names Codex AI reviewers in its disclosure. Publication metadata remains accurate; the revision date is October 1. The skills page keeps one command-reference link. The update guide describes the shipped updater in the present tense while retaining the 0.19.0 minimum. The runtime guide accurately labels disabling Jev and links to the workflow and update commands. + +Drafted by Codex AI agent `/root`; independently reviewed by Codex AI agent `/root/ghostget_editorial` on October 1, 2026. The reviewer read both complete page bodies and the updater guide, and checked the updater minimum against the changelog, release version, entry point, and installation-lock behavior. No new study, model-quality, human, or professional review is claimed. The existing article admission remains 11/12; reassess on November 9, 2026, or a relevant behavior change. Final aggregate and browser verification are recorded with the pull request. diff --git a/docs/runtime.md b/docs/runtime.md index 01889e6..6f7f7d1 100644 --- a/docs/runtime.md +++ b/docs/runtime.md @@ -164,7 +164,7 @@ sys1 jev status `hosted.enabled: true`, and sets routing to `hosted-only`. The key remains in the environment; Sys1 never writes it to its config or prints it. Avoid putting the key in source files or shell history. Restart a gateway that was -started before the key was exported. To return to local-only operation: +started before the key was exported. To disable hosted Jev: ```sh sys1 jev disable @@ -515,6 +515,9 @@ sys1 config path|get|set|unset sys1 --version|--help ``` +See [saved checks and review workflows](workflows.md) for `sys1 workflow` +commands and the [update guide](cli-updates.md) for `sys1 update`. + Supporting commands accept `--json`. When an agent runs sys1 (Claude Code, Codex, Cursor, Gemini CLI, or `AI_AGENT` is set), JSON is the default; `HRANESS_AUDIENCE=human` or `agent` overrides the guess. Machine data goes to diff --git a/site-templates/docs/evaluations.html b/site-templates/docs/evaluations.html index 770db51..e73970d 100644 --- a/site-templates/docs/evaluations.html +++ b/site-templates/docs/evaluations.html @@ -91,7 +91,7 @@

Sys1 evaluations

These are Sys1’s own tests of the models it can use, with raw reports, failure analysis, and the steps to rerun them. The small fixtures show how each adapter behaves. For cross-model accuracy and cost, use the model comparison.

-

Adapter studies: September 19–20, 2026 · Workflow screening: September 28, 2026

+

Adapter studies: September 19–20, 2026 · Workflow screening: September 28, 2026

@@ -168,7 +168,7 @@

Sys1 evaluations

Rerun these tests.

The fixtures, grading rules, model pins, source hashes, and complete responses are public. Every case is reported, including the ones a model got wrong.

  1. Quality: 72 authored cases, 24 per answer type. Grade only the first pass. Choice requires the exact label; yes/no uses a strict 0.5 threshold; score uses a unique most-probable level. Noul and Score ties, and all failures, count as incorrect.
  2. Stability: all six option orders for 24 choice cases. Report identical semantic answers separately from correctness; permutations do not add independent cases.
  3. Timing: three measured passes, 216 calls at concurrency one. Two earlier form cases warm up the route. p50/p95 use nearest rank over all attempts; throughput includes response validation.
  4. Boundaries: pin weights and the provider model version; keep local and network timing distinct. A small project-authored test does not establish production quality or calibrated probabilities.
Related projects and publisher benchmarks

Jev is a hosted decision model; Qwen supplies open model weights; llama.cpp supplies local inference. Sys1 connects these to one typed application contract. OpenJev offers a separate compatible server you can operate and explicitly register.

Publisher speed and quality results use different workloads, runtimes, and hardware. JevBench is the external cross-model suite; its scores remain separate from Sys1's local adapter evidence. The upstream evidence archive preserves other primary sources and conditions.

-
Run the evaluationDetailed evidence notesUse Sys1 in your app
+
Run the evaluationDetailed evidence notesUse Sys1 in your app
diff --git a/site-templates/introducing-sys1.html b/site-templates/introducing-sys1.html index 76f9fe6..4b5e9b6 100644 --- a/site-templates/introducing-sys1.html +++ b/site-templates/introducing-sys1.html @@ -79,7 +79,7 @@ "headline": "Introducing Sys1", "description": "Sys1 gives coding agents tools to review changes, check completion claims, and get structured decisions. Choose a skill and add it to your project.", "datePublished": "2026-09-28", - "dateModified": "2026-09-30", + "dateModified": "2026-10-01", "inLanguage": "en", "author": { "@type": "Organization", @@ -107,7 +107,7 @@ "name": "Claude Code AI revision agent" } ], - "creditText": "Drafted and revised with AI from the source code, and reviewed by AI agents.", + "creditText": "Drafted and revised with AI from the source code, and reviewed by Codex AI agents.", "about": [ { "@type": "SoftwareApplication", @@ -198,8 +198,7 @@

Introducing Sys1

Give your coding agent a repeatable way to review a change and check what it says is done.

-

Hraness Updated

-

Release at publication: v0.17.0. Open source under MIT. Install the current release.

+

Hraness

@@ -291,7 +290,7 @@

What the results cover

diff --git a/site-templates/skills.html b/site-templates/skills.html index da4bc9e..9cc0703 100644 --- a/site-templates/skills.html +++ b/site-templates/skills.html @@ -116,7 +116,7 @@

{{PORTFOLIO_HEADING:skills-first-check-heading}}

Failed check
The command’s failure status is preserved. Read the full log when the excerpt is not enough to understand the failure.
Everyday use
Ask your agent to use system-one-verify for a known noisy check. Prefer the command’s native quiet reporter when it already gives you a useful result.
-
Command referenceBehavior and limits
+
Command reference
diff --git a/site/docs/evaluations.html b/site/docs/evaluations.html index 2f99634..7ef25d1 100644 --- a/site/docs/evaluations.html +++ b/site/docs/evaluations.html @@ -91,7 +91,7 @@

Sys1 evaluations

These are Sys1’s own tests of the models it can use, with raw reports, failure analysis, and the steps to rerun them. The small fixtures show how each adapter behaves. For cross-model accuracy and cost, use the model comparison.

-

Adapter studies: September 19–20, 2026 · Workflow screening: September 28, 2026

+

Adapter studies: September 19–20, 2026 · Workflow screening: September 28, 2026

@@ -168,7 +168,7 @@

Sys1 evaluations

Rerun these tests.

The fixtures, grading rules, model pins, source hashes, and complete responses are public. Every case is reported, including the ones a model got wrong.

  1. Quality: 72 authored cases, 24 per answer type. Grade only the first pass. Choice requires the exact label; yes/no uses a strict 0.5 threshold; score uses a unique most-probable level. Noul and Score ties, and all failures, count as incorrect.
  2. Stability: all six option orders for 24 choice cases. Report identical semantic answers separately from correctness; permutations do not add independent cases.
  3. Timing: three measured passes, 216 calls at concurrency one. Two earlier form cases warm up the route. p50/p95 use nearest rank over all attempts; throughput includes response validation.
  4. Boundaries: pin weights and the provider model version; keep local and network timing distinct. A small project-authored test does not establish production quality or calibrated probabilities.
Related projects and publisher benchmarks

Jev is a hosted decision model; Qwen supplies open model weights; llama.cpp supplies local inference. Sys1 connects these to one typed application contract. OpenJev offers a separate compatible server you can operate and explicitly register.

Publisher speed and quality results use different workloads, runtimes, and hardware. JevBench is the external cross-model suite; its scores remain separate from Sys1's local adapter evidence. The upstream evidence archive preserves other primary sources and conditions.

-
Run the evaluationDetailed evidence notesUse Sys1 in your app
+
Run the evaluationDetailed evidence notesUse Sys1 in your app
by Hraness
diff --git a/site/introducing-sys1.html b/site/introducing-sys1.html index 03b59cb..b9331a2 100644 --- a/site/introducing-sys1.html +++ b/site/introducing-sys1.html @@ -79,7 +79,7 @@ "headline": "Introducing Sys1", "description": "Sys1 gives coding agents tools to review changes, check completion claims, and get structured decisions. Choose a skill and add it to your project.", "datePublished": "2026-09-28", - "dateModified": "2026-09-30", + "dateModified": "2026-10-01", "inLanguage": "en", "author": { "@type": "Organization", @@ -107,7 +107,7 @@ "name": "Claude Code AI revision agent" } ], - "creditText": "Drafted and revised with AI from the source code, and reviewed by AI agents.", + "creditText": "Drafted and revised with AI from the source code, and reviewed by Codex AI agents.", "about": [ { "@type": "SoftwareApplication", @@ -199,8 +199,7 @@

Introducing Sys1

Give your coding agent a repeatable way to review a change and check what it says is done.

-

Hraness Updated

-

Release at publication: v0.17.0. Open source under MIT. Install the current release.

+

Hraness

@@ -292,7 +291,7 @@

What the results cover

diff --git a/site/skills.html b/site/skills.html index d1282df..3b5cd55 100644 --- a/site/skills.html +++ b/site/skills.html @@ -118,7 +118,7 @@

Run your first check

Failed check
The command’s failure status is preserved. Read the full log when the excerpt is not enough to understand the failure.
Everyday use
Ask your agent to use system-one-verify for a known noisy check. Prefer the command’s native quiet reporter when it already gives you a useful result.
-
Command referenceBehavior and limits
+
Command reference