[hsliuustc0106]

Required human contributor answers

1. What outcome are you trying to achieve?

   Answer: Put the post-v0.2.0 CHANGELOG.md Unreleased bullets under ## 0.3.0 - 2026-10-30, and keep leftover 0.2.0-era notes out of that section. Source: this change request, which matches the cutting note on the published v0.3.0 release.

2. What is the highest-leverage variable for that outcome?

   Answer: Which Unreleased bullets were added after the v0.2.0 tag (0a912ba, 2026-10-14) versus which were already in Unreleased at that tag. Source: this change request.

3. How will this work move that variable rather than create busywork?

   Answer: Relocate those bullets verbatim. Post-tag bullets go to 0.3.0. Tag-time Unreleased bullets, which the published v0.2.0 notes point at as the full list, are backfilled under ## 0.2.0 - 2026-10-14. No version bump, tag, or release edit. Source: this change request.

4. Who is the customer or user, and what do they actually need?

   Answer: Readers of CHANGELOG.md and the already-published v0.3.0 release, who need the notes under the version that shipped. Source: this change request.

After answering the four questions: is a PR needed at all to achieve this outcome?

Human decision and rationale (or explicit source reference): Yes. This change request asks for a PR. The v0.3.0 release was already cut at 113ef26 without promoting Unreleased, and that tag must stay where it is. A docs-only PR is how the headings land on main.

For a bug fix: link actual observed failure evidence.

Evidence: Not applicable. This is changelog hygiene, not a bug fix.

Why

v0.3.0 was published on 2026-10-30 at 113ef26 while ## Unreleased still held both the post-0.2.0 notes and the 0.2.0-era notes that were never promoted at the 2026-10-14 tag (0a912ba). The published release says to cut only the post-0.2.0 bullets.

How

CHANGELOG.md only. A bullet whose text is absent from v0.2.0:CHANGELOG.md moved under ## 0.3.0 - 2026-10-30 (Added / Changed / Fixed). A bullet present at the v0.2.0 tag moved under ## 0.2.0 - 2026-10-14. ## Unreleased is empty. ## 0.1.0 is unchanged. pyproject.toml is not bumped. No git tag or GitHub Release was created, moved, or deleted.

What

• ## 0.3.0 - 2026-10-30: the 17 bullets added after v0.2.0. Served OneJev and its shared HTTP deadline, jev-vl-served and trace scoping, the TypeSafe jev-1.13.0 / TYPESAFE_MODEL / requested_model pin, unified Choice/Noul confidence, injection-guard withhold/coverage/labelled-set behavior, injection-public.jsonl, the BrowseSafe sample, MCP titles/hints/schemas and long-run progress/cancel, mcp>=1.26,<2, s1a probe --order, the offline public231 quality consumer, duplicate --option rejection, and ticket JSONL errors that name the dataset and line.
• ## 0.2.0 - 2026-10-14: the 41 Unreleased bullets already in the file at the v0.2.0 tag (CLM, rethink/recovery, omnijev, laya-served, Windows CI, and the rest of that cut).
• #72 (changed-image decision replay) merged after v0.2.0 and is a v0.3.0 release highlight, but that PR did not add a CHANGELOG bullet. This change does not invent one.

Verification

• Compared bullet text with git show v0.2.0:CHANGELOG.md and with CHANGELOG.md at 113ef26: 17 post-tag bullets in 0.3.0, 41 tag-time bullets in 0.2.0, no overlap, no dropped or rewritten bullets, and ## 0.1.0 byte-identical.

• git diff origin/main --stat is CHANGELOG.md only.

• git ls-remote --tags origin v0.2.0 v0.3.0 still points at 0a912ba and 113ef26.

• uv run ruff format --check . && uv run ruff check . && uv run ty check — not run; this diff does not change Python

• uv run pytest -q and scripts/smoke.sh — not run; same reason

• CHANGELOG.md and the docs say what the code does now — changelog headings now match the published cuts; no other docs changed

Demo / evidence

• Video link, or exemption/blocker with follow-up: N/A. Minor documentation change; it does not alter behavior or introduce a quality or performance claim (CONTRIBUTING.md, Agent video demos).
• Application/task → agent actions → independently checked result: N/A, same exemption.
• System1-Agents / System1-Omni commits, checkpoint and worker/frontend provenance: N/A, same exemption.

  

[Birfy]

Checked the relocation mechanically against v0.2.0:CHANGELOG.md (tag 0a912ba), v0.3.0:CHANGELOG.md (113ef26, identical to current main) and this head (358998d):

• main's ## Unreleased has 58 bullets; this head has 17 under 0.3.0 - 2026-10-30 and 41 under 0.2.0 - 2026-10-14. Their union equals main's Unreleased exactly, with no duplicates and no dropped or reworded text.
• None of the 0.3.0 bullets were in Unreleased at the v0.2.0 tag, and every 0.2.0 bullet was.
• Every bullet keeps its Added / Changed / Fixed subsection, ## 0.1.0 is byte-identical, and the diff touches CHANGELOG.md only. Both tags still point where they did.

Not adding a bullet for #72 is right.

Non-blocking nit: the PR adds a blank line between every bullet in the 0.3.0 and 0.2.0 sections (26 to 72 blank lines in the file). ## 0.1.0 and the rest of the file use tight lists, and on GitHub the blank lines turn these into loose lists (each item in its own paragraph). They also make this diff look bigger than the move it really is. I'd drop them for consistency, but it's fine to merge as is.

Approving. CI was still running when I reviewed, but the change is docs-only.