hatif03
storybound
Commits
storybound
Log in
Branch
is
main
Hide
October 12, 2026
36 commits
Merge: untrack video intermediates
919d5e0
-115
28 files
Untrack the video build intermediates (scene files)
b827279
1 checkpoint
-115
28 files
Merge PR 8: Vercel front door, Entire links, end cards
4c3055d
+102
/
-60
11 files
Stop tracking the video build intermediates (scene files)
46b8db2
1 checkpoint
+1
1 file
Vercel https front door, Entire links on the judges page, the unique-story answer, end cards with the https address
d0405cd
1 checkpoint
+101
/
-60
10 files
Merge PR 7: the reworked 3-minute video
5a682a3
+292
/
-28
53 files
The reworked 3-minute video (about two minutes of the app, then the tools), script doc, release links, lint allowance for the approved length
44463d4
1 checkpoint
+292
/
-28
53 files
Merge PR 6: the 3-minute demo video and status
dde58a9
+4
/
-1
2 files
The 3-minute demo video (175 s) and the status rows for the deployment, judges page and videos
c42f6b2
1 checkpoint
+4
/
-1
2 files
Merge PR 5: judges page, plain-language README, deployment notes, video tooling
fceee53
+2097
/
-731
96 files
Final links in the README and judges page, the tour builder, links file, evening build log, privacy wording in the deployment notes
2c15351
1 checkpoint
+50
/
-12
9 files
Judges page, plain-language README, WHY, submission answers for the three tools, dashboard screenshots, deployment notes, video tooling and voice-over, a note not to upload a real child's photo
9c70887
1 checkpoint
+2059
/
-731
93 files
Merge PR 4: architecture and failure-drill experiments, fixes, pages shown only when complete, daily caps, recordings
7f8e97d
+40133
/
-23
66 files
State the narration rule precisely: a page reads itself after a pause, nothing else starts on its own
c6462fb
1 checkpoint
+2
/
-2
2 files
Daily caps (all visitors and per visitor) for stories and movies on a public deployment, with a calm message
2b95001
1 checkpoint
+44
/
-2
3 files
Merge main into the experiments branch
c692d89
Chaos injection does nothing unless the run also names a drill (found by the Entire trail runner: a stray STORYBOUND_CHAOS must not reach production)
5f9253b
1 checkpoint
+14
/
-3
2 files
The final app capture (pages read themselves, Listen again), regenerated voice-over, submission checklist, docs matched to the recording's real receipt
ff651f6
1 checkpoint
+671
/
-841
25 files
Each page reads itself aloud 1.8 s after it opens (sound is on by default); the Listen button stays and becomes Listen again for replays
a601424
1 checkpoint
+23
/
-12
5 files
Evidence and docs for 12 Oct: the picture-words detection fired, real before and after traces, cfo.ai Version-history screenshots, the app capture's cues and receipt and voice-over clips, the recording guide
0274411
1 checkpoint
+7449
/
-15
35 files
Tools for the demo: the picture-words drill, trace span saver through the MCP, a recording of the real app with sound (every sound logged with its time, timestamped frames), voice-over clips and the composer; a repainted picture is painting, not fixing a page
f590ea1
1 checkpoint
+738
/
-2
9 files
A page is shown only when its words, picture and voice all exist: narration starts as each page is accepted, story.verified marks the end of the whole-story check, the reader and the making screen wait for it
1242660
1 checkpoint
+99
/
-27
12 files
Record the review fixes (HR-015), the review output, two post-review runs
771e800
1 checkpoint
+76
/
-7
9 files
Fixes from Entire's review of this branch: a block is still a refusal, safety promises stay unverified in the ensemble, a failed judge is not waited for twice, a verifier crash is not an outage, guarded experiment settings
ace0764
1 checkpoint
+142
/
-15
10 files
Docs for 12 Oct: STATUS, TOOLS round four, STORY, API error code, HANDOFF, build log, the demo video script
883e076
1 checkpoint
+74
/
-16
9 files
HR-016 and the neatlogs round: the fragment-trace measure, the three incident cards resolved, the rebuilt evaluation
dbf0a1f
1 checkpoint
+581
20 files
Experiment and drill results: HR-014 (writers, judge, repair budget, audit) and HR-015 (eight failure drills, four fixes), the comparison table, ledger rows
8e2f415
1 checkpoint
+828
5 files
A model service that fails after its retries ends in the try-again screen, not a generic error (found by drill C3b: 1 of 5 stories)
26c81a1
1 checkpoint
+18
/
-1
2 files
The workflow span now carries the request and the result, so a reviewer opening a story trace in neatlogs sees input and output instead of {} and null
0bfeecc
1 checkpoint
+4
/
-2
1 file
Start judge calls and pictures from the workflow's context so their spans stay in the trace (2 of 4 pictures and every picture check were missing; found in the diagnostic run)
74349c9
1 checkpoint
+24
/
-3
3 files
Merge PR 3: cfo.ai workbooks, ages and languages (HR-013)
539040c
+21228
/
-18
30 files
Drill and experiment results so far in the ledger; diagnostic run for the missing picture spans
9356cdd
1 checkpoint
+5660
2 files
Fixes from drills C3 and C4: a judge failover to another family, and an outage shows the try-again screen instead of a refusal
dfcad8b
1 checkpoint
+98
/
-1
6 files
Tolerant page parsing for other writer models (found by the Sonnet writer experiment); an independent audit pass for sweeps; the experiment comparison script
e918923
1 checkpoint
+25579
/
-4
7 files
Experiment driver; the reserve Gemini model priced as an assumption
2ec0edb
1 checkpoint
+71
3 files
Experiment tooling: failure injection (chaos), a second judge from another family, a repair-budget setting, ledger fields for the configuration, sweep options
fe643e3
1 checkpoint
+158
/
-13
12 files
October 11, 2026
14 commits
Trail 4 gates after the second push: the review runner passed again with no findings
edf510c
1 checkpoint
+181
1 file
Docs: HR-011 notes on the orphan traces, runner result on trail 4, launch readiness and status for ages and languages
10f6986
1 checkpoint
+205
/
-14
9 files
Ages 4 and 8 and four non-English caption fixtures (HR-013): a sweep --age option, fixtures, runs
d0519f0
1 checkpoint
+20838
/
-3
13 files
cfo.ai workbooks (five pages) saved as the offline copy; public link checked anonymously
8ba1266
1 checkpoint
+4
/
-1
8 files
Triage cards handled; the Entire runners did not run because Auto build on push is off
e752c4e
1 checkpoint
+12
/
-11
6 files
neatlogs alerts: thresholds retuned, a rule on the corrected quiz detection, email delivery on for every rule, a test alert; docs updated
bd982e3
1 checkpoint
+52
/
-7
11 files
Merge PR 2: credits routing, stall hedge, review fixes, cfo.ai plans, docs
6661b7d
+38714
/
-39
71 files
Entire review and gate snapshots for trail 3
381988d
1 checkpoint
+1
1 file
Docs: the Entire review of PR 2, runner status, and the story
f79f4ae
1 checkpoint
+214
/
-2
4 files
EVIDENCE.md rebuilt
674a7dd
1 checkpoint
+2
/
-2
1 file
Docs: what Entire's review found in this branch and how it was fixed
bf1b819
1 checkpoint
+9
/
-6
5 files
judge_agreement: error cells are not passes and the judge setting is restored; numbers recomputed
bd6fa2e
1 checkpoint
+14
/
-9
6 files
Review fixes: a failing gateway falls back to Gemini and the receipt lists it; hedged calls use daemon threads, a 45 s delay and count only the winner; quiz text gets the soft/hard steering split; unknown gateway models are priced at the dearest entry
430d017
1 checkpoint
+95
/
-33
9 files
Docs: the story of how the three tools helped, tool usage round three, status, handoff, launch readiness; lint warns when docs lag the code; CLAUDE.md asks for current docs
bd0343e
1 checkpoint
+107
/
-11
11 files