From fe41d15c70c5a1a61eedc6d42afc666c85e711e8 Mon Sep 17 00:00:00 2001 From: Protocol Zero <257158451+Protocol-zero-0@users.noreply.github.com> Date: Thu, 14 May 2026 05:52:28 +0000 Subject: [PATCH] docs: rewrite hero around general-purpose positioning MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The previous hero opened with a specific worked example (Qwen3.6 + SWE-bench Verified + GPT-5.5). That framing reads as "this tool exists to optimize Qwen3.6 against SWE-bench", which buries the actual product: a *general-purpose evolution runtime* that works against any git repository and any measurable goal. This pass reframes the hero around what the runtime is, with the SWE-bench example demoted into a clearly-labeled worked example further down the page. Specific changes: - Hero headline: "Take a 3 B-active open-weight model..." → "A soul plugin for any code." - New second-paragraph framing: the codebase, once given a goal, begins to perceive its environment, judge its output, and rewrite itself toward the goal — sandboxed, ledgered, reversible. - New short evocative paragraph below describing the conceptual shift: code-as-text becomes code-as-process when it can read its own outputs and shed failed experiments without scarring. - Section heading "## Our v1.1 target" → "## A worked example: SWE-bench Verified (our v1.1 target)". Same content; the heading now signals that the example is *one* demonstration of the runtime, not the runtime's purpose. English (README.md) and Chinese (README.zh.md) hero blocks rewritten in lock-step. No code changes; the rest of the README (motivation, what-it-does, architecture, configuration, etc.) is untouched. Co-Authored-By: Claude Opus 4.7 --- README.md | 16 +++++++++++++--- README.zh.md | 14 +++++++++++--- 2 files changed, 24 insertions(+), 6 deletions(-) diff --git a/README.md b/README.md index adb8caa..8236c5f 100644 --- a/README.md +++ b/README.md @@ -1,8 +1,18 @@ # Evolution Kernel

- Take a 3 B-active open-weight model. Close most of the SWE-bench gap to GPT-5.5 — overnight, hands-off, fully audited.
- No retraining. No fine-tuning. No weight changes. Only the harness around the model evolves. + A soul plugin for any code.
+ Drop it into any git repository, give it a measurable goal, and the + codebase begins to perceive its own environment, judge its own + output, and rewrite itself toward the goal — sandboxed, ledgered, + every change a real git commit you can roll back. +

+ +

+ Code on disk does not evolve. But once a codebase can read its own + outputs, hold a conversation with its own past, and shed failed + experiments without leaving scars — it stops being a body of text and + becomes a process.

@@ -76,7 +86,7 @@ Every attempt is written to a **ledger**: goal, observation, plan, diff, evaluat --- -## Our v1.1 target +## A worked example: SWE-bench Verified (our v1.1 target) > 📋 **ROADMAP TARGET · NOT A LOGGED RUN.** The example below describes the next milestone we are engineering toward, not a checked-in artifact. When the run lands, the full ledger will be committed under [`evidence/`](evidence/) and this README will link to it. For runs you can reproduce **today**, see [`examples/sandbox_demo/`](examples/sandbox_demo/) and the 99-test suite in [`tests/`](tests/). diff --git a/README.zh.md b/README.zh.md index c889dbb..9ec59e0 100644 --- a/README.zh.md +++ b/README.zh.md @@ -1,8 +1,16 @@ # Evolution Kernel

- 给一个 3 B-active 的开源小模型，用一晚上无人值守，把它在 SWE-bench Verified 上推到接近 GPT-5.5 的水平。
- 不重训，不微调，不动一个权重——只进化模型背后的 solver harness。 + 任意代码的灵魂插件。
+ 把它放进任意 git 仓库，给它一个可衡量的目标，这份代码便开始感知自己的 + 环境、评判自己的产出、朝着目标重写自己——沙箱里隔离、ledger 留档、 + 每次接受的变更都是一条可回滚的 git commit。 +

+ +

+ 静止的代码本不会进化。但当一段代码能读取自己的产出、能与自己 + 的过去对话、能丢掉失败的实验而不留疤——它便不再是被写下的文本， + 而成为一个正在发生的过程。

@@ -76,7 +84,7 @@ Evolution Kernel 把 harness 调优变成一个可复现的 runtime。把它指 --- -## v1.1 路线图目标 +## 一个具体示范：SWE-bench Verified（也是我们的 v1.1 目标） > 📋 **路线图目标 · 不是已落盘的运行。** 下面这个例子描述的是我们下一步要工程化达到的里程碑，不是已经 checked-in 的 artifact。当这个 run 落地后，完整 ledger 会提交到 [`evidence/`](evidence/) 目录，本 README 会链接过去。**今天就能复现**的真实运行，请参考 [`examples/sandbox_demo/`](examples/sandbox_demo/) 和 [`tests/`](tests/) 下的 99 个测试。