docs: sync launch outreach copy for v0.1.20

Author: aaronlab

docs/launch/day-3-targeted-communities-packet.md

@@ -344,8 +344,9 @@ uv run --python 3.11 python scripts/launch_metrics.py --append --note "after com
 
 ## Directories And Newsletters
 
-Use these after `v0.1.19` because the release has the packaged demo command and
-safer redacted export path.
+Use these after `v0.1.20` because the release has the packaged demo command,
+failed-vs-good compare output, local compare API, and safer redacted export
+path.
 
 Use `docs/launch/directory-submission-sheet.md` for exact entry points, link
 order, owner-only notes, and tracking instructions.
@@ -394,9 +395,11 @@ records each Browser Use step as a timeline: screenshot, URL, action, model
 input/output, status, and error. Stagehand, Skyvern, and Playwright + LLM
 workflows are secondary integrations.
 
-The v0.1.19 release includes `browsertrace demo` for first-run onboarding,
-`browsertrace export --public` for public-safe sharing, and individual
-redaction flags when users want to keep some fields visible.
+The v0.1.20 release includes `browsertrace compare` for failed-vs-good run
+pairs, a local `/api/compare/<failed>/<good>` JSON endpoint,
+`browsertrace demo` for first-run onboarding, `browsertrace export --public`
+for public-safe sharing, and individual redaction flags when users want to keep
+some fields visible.
 
 Try locally from PyPI:
 

docs/launch/directory-submission-sheet.md

@@ -1,6 +1,6 @@
 # BrowserTrace Directory And Newsletter Submission Sheet
 
-Use this sheet after `v0.1.19`. Submit only where the target accepts tool
+Use this sheet after `v0.1.20`. Submit only where the target accepts tool
 suggestions, launches, or editorial pitches. Do not ask for stars, votes,
 upvotes, swaps, or artificial engagement.
 
@@ -49,10 +49,11 @@ as secondary integrations.
 BrowserTrace is an MIT-licensed local debugger for Browser Use failures. It
 records each Browser Use step as a timeline with screenshot, URL, action, model
 input/output, status, and error, then exports a standalone HTML trace with
-optional redaction. The `v0.1.19` release includes the public-safe
-`browsertrace export --public` mode, a packaged `browsertrace demo` command,
-Browser Use/Stagehand/Skyvern/Playwright guides, and a no-dependency Browser
-Use-shaped callback demo.
+optional redaction. The `v0.1.20` release includes `browsertrace compare` for
+failed-vs-good run pairs, a local `/api/compare/<failed>/<good>` JSON endpoint,
+the public-safe `browsertrace export --public` mode, a packaged
+`browsertrace demo` command, Browser Use/Stagehand/Skyvern/Playwright guides,
+and a no-dependency Browser Use-shaped callback demo.
 
 ## Submission Queue
 

docs/launch/outreach-targets.md

@@ -44,7 +44,7 @@ Post only where the community rules allow it. The ask is feedback from people bu
 2. Hacker News once the first-run path is clearly understandable.
 3. Browser Use and Stagehand communities because they match the product.
 4. Reddit only with a useful write-up and after checking local rules.
-5. Directory/newsletter submissions after `v0.1.19` links are stable. Use
+5. Directory/newsletter submissions after `v0.1.20` links are stable. Use
    `docs/launch/directory-submission-sheet.md` as the queue.
 6. Product Hunt after the visual assets are strong enough for a broader audience.
 
@@ -71,9 +71,11 @@ failures.
 
 It records each Browser Use step as a timeline with screenshots, URL, action,
 model input/output, status, and errors, then exports a standalone HTML trace.
-The v0.1.19 release includes `browsertrace demo` for first-run onboarding,
-`browsertrace export --public` for public-safe sharing, and individual
-redaction flags when users want to keep some fields visible.
+The v0.1.20 release includes `browsertrace compare` for failed-vs-good run
+pairs, a local `/api/compare/<failed>/<good>` JSON endpoint, `browsertrace demo`
+for first-run onboarding, `browsertrace export --public` for public-safe
+sharing, and individual redaction flags when users want to keep some fields
+visible.
 
 Try locally from PyPI:
 

docs/launch/response-templates.md

@@ -94,10 +94,11 @@ error: upload preview did not appear
 ```
 
 The comparison reports the first divergent step across the existing `action`,
-`url`, `status`, and `error` fields. It is intentionally explicit in v0.1.19:
+`url`, `status`, and `error` fields. It is intentionally explicit in v0.1.20:
 you choose the failed run and the good run instead of relying on automatic
-baseline selection. Use `--json` when you need structured output for automation
-or CI checks without pasting a large JSON blob into a reply.
+baseline selection. Use `--json` or the local `/api/compare/<failed>/<good>`
+endpoint when you need structured output for automation or UI checks without
+pasting a large JSON blob into a reply.
 
 Guide: https://aaronlab.github.io/browsertrace/browser-use-debugging.html#compare-run-metadata-checklist