From a recall sample of clean finishes: "the box hasn't answered yet" is
something not done, and "One thing worth flagging: …" is a heads-up.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A sample of the 308 real pages titled "Done, with a note" read as open
work in about nine of ten. The rest were history ("Before, that early
idle could end a wait…", "used to") or "still open" as a verb ("JSON
files still open the live log viewer, as intended"). Neither counts as
something not done any more, and a history sentence is no open problem
either.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
`git init -q --bare ../origin.git` (flags before --bare) counted as Claude
starting the project's repository. And a last step that printed a usage
message or an error (`error: unknown option`, `grep: x: No such file`)
read as "the last check finds nothing"; it now fails, as it did.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 9: a background job's page showed only Claude Code's
line 'Agent "Helper B: write extra pytest tests" finished', so what the
helper said about its own work vanished, here that it had never run its
tests against the real wordcount.py.
The notice balloon now carries the helper's report like Claude's last
words (its first sentence and the key sentences from the rest):
'"Helper B: write extra pytest tests" reports: I wrote the file:
tests/test_extra.py It has 18 test cases … I haven't run it against the
real wordcount.py.' A report that is only a path is its file name, and
absolute paths keep their last two parts, so a long path does not push
the sentence that matters out of a phone's balloon.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A sweep of 1,314 real transcripts (3,863 pages) titled 330 of them
"Done, with a catch"; a sample showed about half were history or
negations: "I found and fixed a model bug", "Bug fixed: …", "found no
broken links", "Not a problem", "the caveat is gone", "was never
broken", plus names ("bug-fixing", "cron-broken-network",
"#limitations", "the memory warning").
A problem sentence now counts only while it is open: not described as
fixed, gone or the past cause (unless it says "not fixed"), not a
negation, and a bare "warning" no longer counts. "Left out" and "left
open" are open work (a note), unless Claude chose it ("left out on
purpose", "rather than guessed"). 244 pages keep the catch; samples
read as real open problems in about four of five.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Since second opinions ride on the next panel, an advisor call sorted to
the front of a helper panel tripped the old rule that kept advisor
panels apart, so "advisor, helper A, helper B" drew two helper panels,
and the first counted the advisor as a helper ("Claude calls in 2
helpers" for one helper). The rule has no job left and is gone; a
riding second opinion is never counted as a helper. One recorded page
changes: "Claude sends 1 helper off to work in the background."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 9: "Claude tries a command, and it fails" made up a
problem. The command printed what Claude was after, then exited 1 only
because its last step (`git config user.name`) found nothing set.
A command whose LAST step is a lookup that answers "no" with exit code
1 (grep, `git config`, `test`, `which`), after printing something,
now reads "Claude checks a few things; the last check finds nothing.",
and does not make the turn "Nothing Claude tried worked". A real
failure (another exit code, nothing printed, or the lookup not last)
still fails.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 9: "You stop Claude here, during its step 3." said
where Claude was, not what that meant. The stop now carries the step's
own words, "during its step 3 (new tests under test/)", and when Claude
kept a to-do list, what was still open on it: "Not done yet: Write the
new tests; Update the README and 1 more."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 9 (a newcomer to the project):
- One or two files are named, not counted: "Claude writes .gitignore and
saves the work", "Claude changes rename_by_date.sh.", "Claude writes
the new files mathx.py and test_mathx.py." Three or more are counted.
- The first save in a folder Claude turned into a repository says so:
"Claude starts a git repository and saves the work: …". A `git init
--bare` (the push target) does not count, nor a folder that already
had commits.
- A push to a repository in a folder on this computer (`To ./origin.git`)
is a copy, not an upload: "then copies it to origin.git, a folder on
this computer", and the final chip reads "copied to origin.git".
- A tool lookup for plan mode reads "Claude gets ready to plan."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 9: "Claude asks a stronger AI model for a second
opinion" stood as a whole panel on almost every page, and the reader
never learns what it said (the answer is encrypted). An advisor call now
rides on the panel of the step after it (or the last one, when nothing
follows), counted in `secondOpinion`: a small line under the caption,
"after asking a stronger AI model for a second opinion" ("then asks…"
when it came last), and "· after 1 second opinion" in the Director's
cut. Pages lose a panel of noise and keep room for the work.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 9:
- "another check" was too vague to act on. A check other than the tests
is named by its tool: a layout check (a formatter), a style check (a
linter), a type check, the build. "Fix and test, 2 rounds: a test fails
and Claude fixes it, then a layout check fails and Claude fixes that;
everything passes."
- A failure the prompt asked for reads "The first run fails on purpose,
as you asked; the second passes." ("The tests don't even load, as you
asked."); one Claude planned itself keeps "Test first: … as planned".
- A test run sent to the background that reported in a later turn reads
"Claude starts the tests in the background; later they pass.", not as
if they passed in that turn.
- The test board says "1 test" and, when nothing loaded, "1 error".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Two fresh cold readers (the person who asked for the work, a newcomer to
the project) found the round-8 balloons still cut exactly where Claude
turns from describing to asking, offering or warning: "Want me to add
that?", "I can fix the number format whenever you want", the
http.server instruction, the fix that followed a reported bug.
Key sentences are now picked by kind, one of each, at most three, and
shown in the order Claude wrote them: a problem left (with the fix right
after it, when there is one, so it never reads as open), a question or
offer, how to use what was built, something not done or checked yet, a
step that failed, a word on what you asked for, a corrected warning. A
list intro is judged on its own words and brings its first items; a
pointer ("described below") is skipped; a short offer brings what
"that" was.
The final panel reads "Done, with a catch" for a problem left, "Done,
with a note" for something pending, and "Done, with a question" when
the next move is yours (`openNote` is now 'catch' | 'note'; `asks`).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 8: you asked to see a dry run, Claude's balloon said
"Here's the dry run you asked for", and the page showed nothing of it.
A run panel's terminal card now prints, dimmer under the commands, the
first lines its last command printed (the last ones when it failed:
that is where an error is).
The card is drawn on the canvas, so it goes into a saved image someone
may share: a line that looks like a secret (token, password, api key,
credential, inside identifiers too) is left out, addresses and long
token-like strings are masked, and Claude Code's own markers ("(Bash
completed with no output)", tool-error tags) are dropped.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 8: "STOPPED" with the balloon "Step 3: new tests under
test/." read like progress, and nothing said what the stop left. The
final panel now names the step Claude last announced and the commits made
before the stop: "You stop Claude here, during its step 3. Saved before
the stop: 2 commits."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 8: "Mission complete!" with confetti and cheering
helpers sat over balloons saying the page will not work as opened, a
helper's tests have not run, or a problem is left. The final panel is now
calm (no confetti, a neutral robot) whenever work was still out at the
turn's end, nothing worked, or Claude's last words name an open problem
(`page.openNote`: a key point about something wrong, not done, not
checked or not fixed); the last of these reads "Done, with a catch".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Two agents read only the comic of the eleven recorded sessions (one who
has never written code, one engineering manager skimming), wrote down
what they believed, then compared with the transcripts. Both were misled
in the same places, and the cause was the same: the final panel kept
Claude's first sentence or two, and what a reader must not lose sat
further into its last message.
- Claude's last words now keep key sentences from anywhere in its final
message (read in full, past the label's clip): an open problem or a
quirk, something not yet done, checked or confirmed, work still out,
what you asked for or expected, a correction of an earlier warning
(with the sentence it refers to). A bare "I didn't" counts only when
nothing stronger is there; table rows, list intros and questions never.
"I built the pomodoro timer. Opening the page: it won't work if you
double-click index.html…"; "…That note makes the current behaviour
clear, but it doesn't fix it."; "As you expected, git describe --tags
then failed because the repo has no tags."
- A turn that ended with helpers or a background command still out says
so: "Done for now", and "Claude's part is done in 36 seconds. Its 2
helpers finished later and reported back." (`finishedLater`).
- A failure you asked for, or that Claude said it expected, reads as the
plan: "Test first: the first run fails as planned, the second
passes.", "The tests don't even load, as planned." A loop that failed
once on a test and once on another check says both, and the commits
made inside a loop are part of its story; the final panel counts the
turn's commits ("5 commits, uploaded").
- One or two files read are named ("Claude checks README.md." in a
turn a report started); a tool lookup reads "Claude gets the tools it
needs ready." again (it was overwritten) without its query; the
advisor is "a stronger AI model", and one that answered with an error
"gets no answer"; a commit message loses its `feat:` prefix in the
plain caption.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Round 6's two sessions join the corpus (sanitized like the others, and
checked to draw the same pages as the originals): tests that do not load,
a fix-fail-fix-pass loop, a commit then a push in its own command, a push
with a failing step after it, an append; two helpers and a command in the
background whose reports come back as turns of their own.
Replaying them showed two captions to fix:
- A test command that ran in the background and reported back read "The
tests pass.", as if they ran during that turn: "The tests run in the
background and pass." (events keep `background`).
- A python script that rewrote README.md drew as "Claude works in the
terminal." beside a separate change panel. A plain command that writes
project files (a script fed by a here-document, or one Claude Code says
changed files) is an edit: "Claude changes 3 files."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 7 (PR review):
- With no hook path (normal after a server restart), the transcript was
looked up by conversation id only; it now also tries the Codeman id and
a recovered session's `restored-<fragment>` transcript, as
last-response does, and keeps what it found briefly (a lookup reads
every project folder; a miss is kept only 4 s).
- A session Story Mode cannot draw is not polled, and a page with nothing
drawn yet is polled like a finished one (every 15 s), not every 4 s.
- Helper steps: one request per helper is drawn, the newest; a failed
refetch never writes over steps already shown.
- The Director's cut toggle, a skin change and the fonts arriving redraw
within the frame budget, not every panel in one task.
- A bare name under an `ls` header finds its folder in one pass over the
lines; a search per link over the text before it was quadratic (about
1.5 s for 700 links on agent output).
- A page of many questions still keeps to 12 panels: questions fold last.
- Opening Story Mode while it is open never starts a second runtime; the
status announcer carries data-i18n-skip; the turn key in the ETag is
clipped.
- The crafted-input timing test runs at sizes where the old code took 6
to 29 s and the new one under 10 ms, with a 1.5 s bound.
- Docs: `older`, `origin` (`scheduled`), which route answers a 500;
mobile.css's stale comment; an orphaned doc comment.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 7 (PR review):
- A helper's drawer froze while the helper worked: a background helper
writes only to its own transcript, so the parent page answered 304 and
the refresh returned before anything refreshed the drawer. An unchanged
page now still refreshes an open drawer's working helpers (ETag'd).
- A helper messaged twice read as superseded ("The trail stops here…"):
its one-page run took its ending from the first turn. It takes the last.
- More than 60 turns in a small file listed the newest 60 with no hint;
`older` is now set whenever the list leaves turns out.
- One /story request that never settled wedged the view, even across a
reopen. Requests carry a token and a 30 s timeout, and opening resets
the request state, with what was announced and shown.
- The background canvases (megabytes each at 2x) are released when Story
Mode closes.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 7 (security):
- Eviction past the per-transcript caps dropped entries one shift() at
a time after the whole tail was parsed: quadratic and never yielding
(a 2 s event-loop stall on a 70 MB long turn, 51 s from 300k entries).
It is one splice, done at every 1 MB yield, so the peak stays near the
caps; and a window that saw a turn start which was then trimmed away no
longer widens to a bigger tail that would be trimmed the same.
- A FIFO or a device in place of a transcript or a helper's .meta.json
(stat said size 0, then open() waited for a writer) pinned a libuv
worker per request; four of them hang every fs, dns and crypto call on
the server. Files are now opened non-blocking, checked with the
handle's own stat (a regular file only) and read through that handle.
- The cache held 6 transcripts for everyone, so a seventh open view made
every poll a full re-read, and a 304 was decided only after reading. An
unchanged file is now a 304 from a stat alone, and the cache holds up
to 24 transcripts within 30 million characters.
- Story Mode is the first route to send a transcript's content to the
browser, so the hook-reported path is used only when it really lies
(symlinks resolved) in a Claude projects folder: the session's
relocated CLAUDE_CONFIG_DIR (now also used for the lookup by id) or the
server's. Anything else falls back to the lookup by conversation id.
Tests move their transcripts into the (temp) projects folder and cover a
path outside it, a symlink out of it, a FIFO transcript, a FIFO meta file
and a 100,000-entry tail.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
API reference: `roundsOk` carries `null` for a run still going, a
background command's `output`, script writes in `edits`, what `helpers`
counts, and the page text budget behind `truncated`.
src/web/CLAUDE.md: panels kept by identity, the balloon text span the
clamp must target, the drawer's unchanged-refresh skip and held turn, the
drawer under the bar and the inert page behind a full-width one, and text
tokens versus accents.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every string Story Mode matches is agent-written (and an agent can be
prompt-injected), and one regex call blocks the single-threaded server.
Timing each helper at its own input cap with crafted shapes found five
that were quadratic or worse:
- A lazy `<tag>[\s\S]*?</tag>` rescans to the end for every opener that
never closes: system reminders, teammate messages and task-notification
fields took 1.5 to 29 s on 0.5 to 1 MB of prompt text. They are found
with indexOf now (withoutBlocks, blocks, blockText1).
- The commit message regex backtracked over many `git commit` tokens
(290 ms at 4,000 characters, growing x8.7 per doubling) and ran on the
whole command. It is a bounded line-by-line scan with the same rules.
- The "This command modified …" note was searched with a lazy match per
repeat of the phrase (7 s on 1 MB of output). It is one anchored try
at its last occurrence, where Claude Code appends it.
- firstSentence and planFromText now read only the start they need.
The worst shape is now under 10 ms at 1 MB. A test times five of them at
sizes where the old code took a second or more.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A sweep of the Story core over every local transcript (1,311 files,
3,855 turns: no errors, never more than 12 panels, the slowest page 12
ms) found pages up to 551 KB (157 KB gzipped), refetched every few
seconds while a long turn runs. The page text budget (outputs, hunks,
helper texts; the oldest events give theirs up first) goes from 400,000
to 120,000 characters: 94% of turns never reach it, and the longest page
is now 74 KB gzipped.
An edit whose lines the budget dropped now says so ("left out to keep a
long turn fast"), instead of the wording for a file a script wrote.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The speaker words added for screen readers became the balloon's first
child, and the clamp took that element instead of the text: long balloons
were no longer cut to fit and ran over the art (the loop's covered the
test count). The clamp now finds the text by its class.
The loop has no Claude to keep clear of on the left, so its balloon may
run wide and is clamped down to one line rather than over the board it
points at; on a 400 px panel "All 19 tests pass…" now ends above "19
tests".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Accessibility (review6-c):
- A panel is named "Panel 2 of 7" and its caption is read once, as its
content (it was the name and the content both).
- Balloons say who speaks, visually hidden ("Claude says: ", "You say: ",
"Claude asks: ", "Notice: "); Save image leaves those words out.
- The overlay is named "Story Mode: <session>", the drawer by its kind and
title ("Diff 2 files"), the final panel's chips by what they open, and
the wire's panel labels are headings.
- While a file preview opened from a panel is up, the overlay is not
aria-modal (the preview sits beside it), and is again once focus returns.
- The focus ring on a click region gets a light band under its dashes; on
the loop's magenta halftone it vanished.
From review6-b:
- "Mission complete!" no longer shows while a helper or a background
command is still out ("To be continued…").
- A background command's drawer says "Started in the background…", then
"Finished in the background, exit code 0…", instead of the launch note
meant for Claude ("do not poll that file").
- `sed -n '5,25p' README.md` is a printed file, offered in the drawer.
- A report turn no longer counts the helper it answers as one it started.
- A helper's `cd /work/x && ls` keeps its folder as ".".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 6 (accessibility): in the paper skin the accents were
used as text colours, and yellow, light green and pink on white fell to
1.42 to 4.48:1 (the "running" tag, the pencils' "Claude is working…",
diff markers, ok/error tags, +N/−N counts, "You approved it", LIVE); in
the dark skin the panel's LIVE tag (white on pink, 3.09) and the note
balloon's "Notice" label (2.74) were under too.
Text now takes its own tokens (--story-go-text, -bad-text, -wait-text,
-pencil-text, the LIVE tag's pair, -note-ink): the accents themselves in
the dark skin, darker ones in the paper skin. Fills and borders keep the
accents. Every pair measured 4.87:1 or better, the tinted diff rows
included.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Two control-character regexes (the NUL span markers, terminal escape
codes) get their eslint-disable directive on the line it covers, and the
montage case gets its own block for its declaration.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 6:
- An open drawer was not tied to its turn: when the newest turn moved on,
a helper drawer went blank and a report drawer silently became the next
turn's prompt. The drawer now remembers its turn; when the newest turn
moves on, the page holds that turn (the picker lists the new one, and
the status line says so) until the drawer closes, then follows the
newest again. Picking a turn yourself wins.
- Previous on the second turn (or Next onto the last) disabled the
focused button and dropped focus to the page body. Focus moves to the
turn picker first.
- The drawer covered the right half of the bar at desktop widths; it now
starts under the bar there too. Where it covers the whole page (a
narrow window, a phone) the page behind it is inert, so Tab no longer
walks the panels it hides.
- A live loop shows the runs that finished, and the one still going as an
open circle (`roundsOk` carries `null` for it).
- The diff of a command that also printed something (`cat >> NOTES.md
<<EOF ... cat calc_stats.py`) shows what it printed, with "Open
calc_stats.py".
- "Red as expected: collection error." stays on its check, not the save
after it. Parked on an older turn, a new turn is announced. A new panel
is sized as it is inserted (no empty first frame). The turn picker is
44 px on touch screens, and the diff drawer no longer says "Live diff"
beside its LIVE tag.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 6:
- An open drawer rebuilt its body on every refresh, unchanged or not (11
rebuilds in 25 s, 7 identical), so a focused button inside it fell to
the page body and a text selection was cleared. A refresh that changes
nothing now keeps the nodes; a real change rebuilds and gives focus back
to the control that does the same thing. Helper steps are filled after
that decision, so a skipped rebuild costs no request and a helper still
working still refreshes its steps.
- When a fix folded into a loop under an open drawer, Escape left focus on
the body: the opener (`diff@fix:1`) was gone. Its replacement is found by
view and panel position (`diff@loop:1`).
- Panels are matched by identity (kind and first event), not position: a
fold into a montage near the start of a long live turn shifted every
later index and rebuilt all eleven panels with their entrances. Only the
montage is new now.
- An edit with no lines (a script, or a command Claude Code says changed
the file) says so: "What changed is not in the transcript: a script or
another program wrote this file."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 6: commit, then push as a second command (the most
common way to ship) drew as "Claude saves the work in 2 commits and
uploads it." for one commit, and the commit's message was dropped. A
command's `commits` was only recorded above one, and the save panel then
counted every save command as one. It is now recorded whenever a command
commits and is the only thing counted: "Claude saves the work and uploads
it: "feat: add mode()"." A push of earlier commits reads "Claude uploads
the work."
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 5: Claude rewrote textstats.py, the change the user
asked for, through `python3 - <<EOF ... open(p, "w").write(s)`, and
the page listed only the two files the same command wrote with `cat >`.
A script fed to python or node by a here-document is now read for the
files it opens for writing (`open(target, 'w'|'a'|...)`, `Path(...)
.write_text`, `writeFileSync`), with the target a string literal or a
name bound to one (`p = "textstats.py"`); anything computed is left
out, never guessed, and scratch targets are skipped. They are edits
without lines, like the files Claude Code says a command modified,
since what the script changed is not in the command.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- The compact panel's click region opens what was packed up, with the
compaction's token counts, not an empty output.
- Parked on an older turn, the poll asks for that turn, the picker learns
of a third turn ("Turn 1 of 3"), the reader stays on turn 1 and nothing
is announced.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- The final panel's chips were rebuilt on every redraw, so the chip that
opened a drawer was gone by the time it closed and focus fell to the
page body. Unchanged chips are now kept, and a trigger replaced anyway
hands focus to the control that opens the same thing.
- API reference: `older`, background Bash calls (`running`), `subject`,
append hunks (`atEnd`), `files` resolved against the command's folder,
test counts read before the output is cut, `roundsOk`, the `note`
speaker and `noneWorked`.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The rest of review round 5 (a phone with touch):
- Text appended with `cat >> f <<EOF` or `echo x >> f` was drawn as lines
1 to 8 of the file; where an append lands is not in the command, so the
diff now says "Added at the end of the file (8 lines)" with no numbers.
- A command that prints a file and does something else (`cat textstats.py
&& pytest`) kept no record of the file; it is now on the event,
resolved against the command's own folder, and the output drawer offers
"Open textstats.py". (A file rewritten by a python here-document still
cannot be diffed from the command alone; Claude Code's own "modified"
note, when it comes, is used.)
- A background job's report that starts a turn was a balloon credited to
"you" in the API (the page already drew it as a note); its speaker is
now "note".
- The output drawer tagged a command still running in the background
"ok"; it says "in the background".
- The Header buttons hint says that on a phone, Story Mode opens from
Session Options.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- The bar on a phone took four rows (216 px of a 664 px screen) with the
close button alone on the last; it now takes three (170 px): title and
close, the turn, the view and Save image. The drawer starts exactly
under it and follows it when the phone turns (the bar is observed).
- A panel below the fold had no buttons until it was scrolled to (only
drawn panels get click regions), so Tab skipped it. A quiet frame now
draws one such panel, a band at a time; its entrance still plays when
it comes into view.
- "Fix and test: the first run fails, the second passes." was chosen by
rounds, so a loop with three runs (a failing test, then ruff, then a
pass) said it under ✗ ✗ ✓. It now needs exactly two runs; that loop
reads "Fix and test, 2 rounds: 2 runs fail, then it passes.", and the
corpus pins every loop's marks.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 5 (a phone with touch):
- Touch screens: every control at least 44 px (prev/next, close, the
Story/Director's cut toggle, Save image, the final panel's chips, drawer
buttons and links), and the small type at 13 px.
- The drawer on a phone starts below the bar (it cut through it), closes by
dragging its head down, and keeps its head small in landscape.
- The turn picker says where you are ("Turn 2 of 5"), learns about new
turns while you look at an older one, and a turn you picked yourself is
not announced as news.
- A check whose command also wrote files offers "See the 2 files it wrote".
- Test counts are read from the whole output: a later command's lines had
pushed pytest's "5 passed" out of the kept tail.
- A save names its commit: "Claude saves the work: "feat: add wrap() for
modular wrapping into a range"." (from -m, -am, --message or the
here-document of git commit -F -; a trailer is never the subject).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 5 (a phone with touch):
- Save image on a phone exported blank art for every panel not yet scrolled
into view (the frame loop draws only what is on screen). It now draws
every panel completely first, and the file name carries the turn number,
so saves of different turns no longer overwrite each other.
- A test command sent to the background ("Command running in background
with ID …") drew as "The tests pass." with POW before any test ran. It is
now running: "Claude starts the tests in the background.", tagged "in the
background" on the wire, and the final panel says a command is still
running. Its task notification settles it from the exit code, and reading
the job's output file that reports tests is a check ("The tests pass.",
7 passed) instead of "Claude looks around".
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Review round 5 had an agent read only the comic of the nine real sessions
(20 turns), write down what it believed happened, and only then compare
with the transcripts. Three panels told it something false, one hid the
failure the user asked to see:
- A push that went through, followed by a check that failed in the same
command, drew as a failed push; the output (`[new branch]`, `->`) now
decides: "Claude saves the work in 2 commits and uploads it, but a step
after that fails."
- A formatter's "would be reformatted" preview, printed before it fixed
the file, made a clean run a failure: the last verdict wins.
- A blocked command ended in confetti and "Mission complete!": a run panel
where everything failed says "Claude tries a command, and it is
blocked.", and a turn where nothing worked ends "Turn over" / "Nothing
Claude tried worked, so the job is not done." with a neutral robot.
- The loop stamped only the final verdict (POW ×2), so a first run that
failed on purpose never showed. It now marks every test run red or green,
and says it: "Fix and test: the first run fails, the second passes." /
"Build and test, 4 rounds. A check fails once, then everything passes."
Also:
- Files written inside a check or a save reach the plain caption ("Claude
writes 4 files; the tests pass.").
- Balloons keep the next sentence when the first reports a problem ("…left a
stray file. I'll remove it…") or is very short ("Not done. The harness
blocked sleep 65…"); Claude's last words keep a caveat that follows them
("Nothing is committed because this folder isn't a git repository.").
- The advisor never shares a panel with helpers that do work; ToolSearch
reads "Claude gets the tools it needs ready."; Codeman's own case file is
not drawn as a file Claude read; looking around shows what Claude meant
(the call's description), not raw commands; a montage at the start of a
turn reads "Claude gets started".
- Words: boards count tests ("19 tests"), "uploads it" replaces "sends it
up", and the prompt balloon keeps 400 characters.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Stepping through 40 turns made 46 long tasks of 51 to 286 ms, enough for
the app's guard to drop the terminal's WebGL renderer for the rest of the
page's life. Drawing switched off made none: the cost was canvas raster at
paint time, which the per-frame JS budget never saw. Now:
- Panel backgrounds are shared between panels and turns (kind × size ×
palette), and a new one is built in four bands, one per frame (a wide
halftone at 2x was ~35 ms in one go; a band is ~8 ms).
- At most two panels are drawn per frame, the first one served rotating so
none starves; clocks still run for all.
- Measured on 20 turns stepped 40 times at 2x: 1 long task (52 ms), against
39; the settled overlay is still ~2% of a core.
Also from round 4:
- The turn picker says when older turns were not loaded (`older` in the
response: the transcript is longer than the part that was read).
- A compaction after the last turn closes that turn; a local command's
output is the final panel's balloon ("Compacted"), without its markdown
ticks, and a turn under a second reads "Done.".
- Exact durations keep their seconds past an hour, so the two captions never
disagree on the minute ("1 h 10 min." / "1 h 9 min 39 s").
- A turn that arrives already finished is announced as new too.
- Whole files written by here-documents read "Claude writes 3 files."; the
advisor's caption reads "a second opinion" instead of repeating its name.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From review round 4 (a 110 MB real session, about 23 turns, read through
all four tail steps):
- A scheduled task's run was drawn inside the /compact turn before it (96
minutes and 237 tool calls credited to "You run /compact."): its prompt is
a meta line, and work after an ended turn was appended to that turn. The
scheduled prompt now starts a turn of origin `scheduled` ("A scheduled
task starts this turn."), and work that comes more than 5 s after a turn
ended starts a turn of its own ("Claude carries on"); a line Claude Code
flushed a moment late still joins.
- A turn with seven prompts added mid-turn drew 14 panels (the cap is 12):
missions were never folded. An added prompt may now fold into a montage,
as the last resort after quiet panels and work panels, and the montage
says so ("Claude keeps going (you add 2 things): ..."); the turn's own
prompt and a question to you are never folded.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Eleven batches of findings had grown it by accretion. It now reads in
parts: the routes, where turns start and end, how background work joins
back, and what a page, an event and a panel carry.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Nine transcripts from the review rounds, each a real Claude Code worker
building a small program (a Python CLI with a fix loop, a web page with
node:test, a bash script with a background helper, a long turn with a
prompt added mid-turn, teammates then an interrupt, a question then tests
first then a push, a text plan with a scratch dry run, /compact then plan
mode, a test failed on purpose). Each is cut down to the lines and fields
the reducer reads, with every path, user name and email rewritten and the
file contents of Reads left out; replaying them before and after gives
the same pages.
The expected file beside each pins every page (panel kinds, captions,
balloons, counts), so a change that alters a story fails with a diff of
exactly what changed; STORY_CORPUS_UPDATE=1 regenerates them. The beats
the reviews were about are also stated outright.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A turn with no end marker of its own (an abandoned or crashed turn) was
closed when the next prompt started and then drawn as "Mission complete!".
It now reads "To be continued…" / "The trail stops here: the next prompt
starts a new turn.", and the page carries `superseded`. The Director's cut
names what ended every turn: turn_duration, stop_hook_summary, a local
command, an interrupt, or the next prompt.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Codeman maintainer