docs: rewrite README for investor narrative + refresh to May 2026 models - #22
Merged
Merged
Conversation
This rewrites the README around an investor-readable narrative:
- New top-of-page pitch: "take a 3 B-active open-weight model, close most
of the SWE-bench Verified gap to GPT-5.5, overnight, hands-off, fully
audited". Replaces the GSM8K math-tutor scenario, which was easy to
read as "a small model doing small problems".
- New "30-second pitch" + "Why this matters" section before the
technical content, so non-technical readers reach the value
proposition (economics / engineering / strategy) within 30 seconds.
- "Our v1.1 target" replaces the prior "See it in action" section. The
example is now framed explicitly as a roadmap target, not a logged
run, to remove any read-as-claimed-result risk.
- Refreshed all model references to May 2026 reality:
Baseline: Qwen3.6-35B-A3B (3 B active, Apache 2.0, released
2026-04-16, 73.4 % SWE-bench Verified — publicly
reported)
Frontier: GPT-5.5 (88.7 %, OpenAI 2026-04-23),
Claude Opus 4.7 (87.6 %)
Also: Gemini 3.1 Pro, Kimi K2.6 references
All prior 2024-era model references (DeepSeek-Coder-V2-Lite, GPT-4o,
GPT-4.1, Claude 3.5 Sonnet, Qwen2.5-Coder-32B) removed.
- New "Known limitations" section: honestly states what v1.0 cannot do
(LLM evaluator can be fooled, sandbox is filesystem-only not network,
history is summary not replay, cost can compound, real-provider
integration not in CI).
- Fixed two stale numbers carried over from v0.2:
"~1,200-line runtime" → "~1,900-line runtime" (actual wc -l)
"39 tests" → "99 tests" (actual current count)
- Both English (README.md) and Chinese (README.zh.md) updated in lock-step.
No code changes. Test suite still 99 / 99 green.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
Documentation-only PR that rewrites both READMEs (English and Simplified Chinese) around an investor-facing narrative and refreshes model references to a May 2026 baseline. The repo's runtime, CI, and tests are unchanged; the README updates align stated numbers (line counts, test counts) with the current state of the code.
Changes:
- Replaces the GSM8K math-tutor "see it in action" example with a v1.1 SWE-bench Verified roadmap target (explicitly labeled "not a logged run"), and updates all model references (Qwen3.6-35B-A3B baseline, GPT-5.5 / Claude Opus 4.7 / Gemini 3.1 Pro / Kimi K2.6 frontier comparison).
- Adds a "30-second pitch" + value-prop table at the top, a "What works today (v1.0)" status table moved above Quick Start, and a new "Known limitations" section.
- Refreshes stale numbers ("~1,200-line runtime" → "~1,900-line", "39 tests" → "99 tests"), removes the obsolete
v0.3 behavior preservedsandbox note, and bumps the OpenAI example model fromgpt-4otogpt-5.5.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.
| File | Description |
|---|---|
| README.md | Full narrative rewrite + May 2026 model refresh + line/test count fixes + new Known limitations section. |
| README.zh.md | Mirror of README.md changes in Simplified Chinese. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| examples/ demo target + sandbox demo + working evolution.yml | ||
| docs/ protocol spec + first-task spec | ||
| tests/ 99 unit + acceptance tests · 14 fixture role scripts | ||
| evidence/ checked-in artifacts of runs anyone can reproduce |
| examples/ demo 目标 + sandbox demo + 可直接运行的 evolution.yml | ||
| docs/ 协议规范 + 第一个进化任务规范 | ||
| tests/ 99 个单元 + 验收测试 · 14 个 fixture 角色脚本 | ||
| evidence/ checked-in 的可复现运行 artifact |
| executor: ["bash", "roles/executor.sh"] | ||
| evaluator: ["python3", "roles/evaluator.py"] | ||
| EOF | ||
|
|
Comment on lines
+243
to
+254
| sandbox: | ||
| enabled: true | ||
| backend: firejail | ||
|
|
||
| roles: | ||
| planner: ["python3", "roles/planner.py"] | ||
| executor: ["bash", "roles/executor.sh"] | ||
| evaluator: ["python3", "roles/evaluator.py"] | ||
| EOF | ||
|
|
||
| # 3. 跑一晚上，放着不管 | ||
| evolution-kernel --config evolution.yml --repo /path/to/project --ledger /tmp/ledger --loop |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Rewrites the README around an investor-readable narrative and refreshes every model reference to May 2026 reality.
Narrative changes
Model refresh (all May 2026)
Stale number fixes
wc -l evolution_kernel/)Both
README.mdandREADME.zh.mdupdated in lock-step.Test plan
grep -n "DeepSeek\|Qwen2.5\|GPT-4\|Claude 3\|GSM8K\|Qwen3-8B" README.md README.zh.md— no residue.🤖 Generated with Claude Code