BaseHalf
All posts

Keep a settled note, and leave the agent's draft as a draft

We checked RFC 9562 UUID version rules in a BaseHalf folder, grew a note from the official PDF, had Claude Code write a draft, and kept only the facts we checked in a separate settled page.

I wanted a page I could reopen tomorrow about UUID versions, and I did not want the first model draft to become that page.

The questions were ordinary ones. In a dashed UUID string, which character is the version? I had been glancing at the first hex digit. RFC 9562 obsoletes RFC 4122: does that mean UUIDv4 is gone? And when people recommend UUIDv7, is the Unix time in that value counted in seconds? I also wanted to read the appendix test vector 017F22E2-79B0-7CC3-98C4-DC0C0C07398F myself, instead of trusting a remembered blog example.

I put the official text into a folder and opened it in BaseHalf 0.6.1. The document is RFC 9562, Universally Unique IDentifiers (UUIDs), by K. Davis, B. Peabody, and P. Leach, an IETF Standards Track document from May 2024 that obsoletes RFC 4122. I retrieved the official PDF and the official text from the RFC Editor on 2026-11-13, Asia/Shanghai, and kept both in the folder to read them. I am not republishing the document. A short SOURCE.md records those facts. I also wrote question.md with the four questions and my first guesses, before I had checked the text.

In Card Detail I opened the PDF, went to section 4.2, Version Field, on page 10, selected the heading and the sentence that states where the version bits are, and used Grow branch. That created rfc9562-note.md, already listing rfc9562.pdf as its upstream, with a blockquote and a source link to page 10. The quote stops at bits 48 throug. I did not tidy that away. I also read Figure 5 on page 11, section 5.4 on page 14, section 5.7 on pages 18 and 19, and Appendix A.6 on page 43, and I read the same wording in rfc9562.txt.

I connected the official texts and the grown note to question.md by writing them into its upstream list. Lines appeared on the canvas between the same files. SOURCE.md sits next to the texts, but I did not connect it. Being nearby is for my eyes. A connection is the context that should stick.

A BaseHalf canvas of the uuid-versions folder: the question on the left, rfc9562.pdf in the center, rfc9562.txt and SOURCE.md on the right, and the grown note below the text file, with lines from the official texts and the grown note into the question, and no line from SOURCE.md.

Card Detail open on rfc9562.pdf at section 4.2 on page 10, with the heading and the version-field sentence selected and the Grow branch button in a small toolbar under the selection.

I selected the question card. The status bar showed that agents could see question. In the Agent Area I started Claude Code and asked it to read the question and only the files listed in its upstream, then write draft-answers.md in the same folder, quoting the rule from RFC 9562, without editing the existing notes. I told it that the grown note may be cut off, and to use the official text for the complete wording.

The same canvas with the question card selected, and the request typed into Claude Code in the Agent Area: write draft-answers.md from the official texts and the grown note, and do not edit the existing files.

After 1 minute 29 seconds, at 18:12 Asia/Shanghai, the draft was there. It appeared in the sidebar and as a new card on the canvas. It listed the question and the official texts as its upstream. It did not change question.md, SOURCE.md, the grown note, or the RFC files. I read the draft, and I still did not treat it as the page I would reopen.

I checked the official sentences, and I checked one piece of arithmetic.

  1. The version is not the first hex digit of the string. Section 4.2 says: "The version number is in the most significant 4 bits of octet 6 (bits 48 through 51 of the UUID)." Figure 5 on page 11 writes that position as M in xxxxxxxx-xxxx-Mxxx-Nxxx-xxxxxxxxxxxx, the first character of the third group.
  2. Table 2 in section 4.2 describes version 7 as a "Unix Epoch time-based UUID specified in this document." Section 5.7 is more precise about the unit: "the number of milliseconds since midnight 1 Jan 1970 UTC, leap seconds excluded." It is not a count of seconds.
  3. RFC 9562 obsoletes RFC 4122, and it still specifies UUIDv4. The abstract says: "This document obsoletes RFC 4122." Table 2 still lists version 4 as "The randomly or pseudorandomly generated version specified in this document." Section 5.4 opens: "UUIDv4 is meant for generating UUIDs from truly random or pseudorandom numbers."
  4. In the Appendix A.6 test vector 017F22E2-79B0-7CC3-98C4-DC0C0C07398F, the M character is 7, and the RFC sets ver to 0x7. The same appendix gives the timestamp as "Tuesday, February 22, 2022 2:22:22.00 PM GMT-05:00, represented as 0x017F22E279B0 or 1645557742000." Those first 12 hex digits are the 48-bit unix_ts_ms field from section 5.7. I also converted 1645557742000 milliseconds to 2022-02-22 19:22:22 UTC, which is 14:22:22 at GMT-05:00, as a check of the arithmetic, not as a second authority.

I wrote those checked answers into a new file, settled.md, and I left the first guesses in question.md. I connected the official texts, the grown note, and the question to the settled page. I did not connect the draft to it. The draft stays in the folder as a draft.

The same folder after draft-answers.md and settled.md appeared on the canvas, with the settled note under the question, the draft to the right of the grown note, and Claude Code in the Agent Area recapping the four answers.

Card Detail open on settled.md in the Preview, showing the heading What I checked in RFC 9562 and the four checked answers, with Claude Code's recap still visible in the Agent Area.

This is not a notebook that turns every model reply into a finished page, and it is not a built-in summary. The selection in the status bar lasts only while the window is open. The official texts, the grown note, the questions, the draft, and the settled page are ordinary files in the folder. If I close the chat, I still have the page I wrote the checked answers into, and I can still see that the draft is a draft.