What's New — May 9, 2026
Verify it yourself — free, no login
See how AI medical-record review links every fact to the exact Bates page that proves it — click any citation and jump straight to the record.
See the 60-second demo →This release fixes the single most-requested correctness issue in Case Record Search: the Bates labels shown next to each search hit weren't always the real Bates from that page. They are now. This week's reindex re-attributed Bates citations on 50,145 pages of customer records. Page-image rendering also got a 10-30× speedup on large files. Plus a new admin metric.
Headline fixes
Case Record Search: Bates labels are now correct
The previous indexer's regex for Bates extraction was too loose — strings like MRN12345678, ROOM1234, DATE2024, and LAB123456 all matched. The first one in a page's header would be returned as the page's "Bates stamp" — even if the real Bates was in the footer. On medical-record productions where every page header has an MRN, the result was that nearly every search hit showed the same wrong label.
Before
Search for "informed consent" → returns the right page text, but Bates label shows MRN_012345678 on every result (false positive — that's the Medical Record Number, not the Bates stamp).
After
Search for "informed consent" → returns Bates JOHNSON_001247, JOHNSON_001248, etc. — the actual production stamps from the file's footer. Every prefix is the file's dominant Bates prefix, detected automatically.
How it works now: when a PDF is indexed, the system scans every page and finds the prefix that appears with sequential numbers across the file. That's the file's "real" Bates prefix. Anything else (MRN, ROOM, LAB, CPT, etc.) gets ignored. If a page has no Bates stamp at all, the result card shows "no Bates detected" instead of inventing one.
The denylist of noise prefixes covers 50+ common medical-record metadata patterns: MRN, DOB, DOS, DOA, ROOM, BED, OR, ED, ICU, ER, LAB, CPT, ICD, RX, EKG, ECG, MRI, CT, TIME, DATE, BP, HR, NPI, ZIP, NDC, and others.
Page-image rendering: 10-30× faster on large records
The "View Bates page" button now shows the actual rendered PDF page in roughly 1-3 seconds instead of 30-60+ seconds for large files. Previously, the rendering route did a per-page pdftotext scan to find the physical PDF page that matched the search hit's text — on a 500-page file that meant up to 500 sequential text extractions before the right page could be rendered.
The new indexer stores the physical PDF page index directly when it indexes the file, so the renderer goes straight from "search hit → exact PDF page → PNG render → cached in S3." The first render takes 2-4 seconds; every subsequent click on the same hit is <1 second from the S3 cache.
Concretely: on the largest reindexed case (3,984-page production), a search hit's "View Bates page" used to take 60-90 seconds the first time. Now it's about 3.
Admin: Daily Active Users sparkline
Owners can now see a 30-day DAU trend on /admin alongside the existing Total Users / Active Subscribers / Free Trials / MRR cards. The new card shows today's distinct active users plus the 7-day and 30-day rolling averages, with an inline sparkline of the last 30 days. Today's value is highlighted in yellow; the dashed line is the 30-day baseline.
The footer line splits today's count into "real users" vs. "admin/test" so internal accounts (jeff@, support@, healthcheck@, demo@, etc.) don't inflate the number prospects evaluate against.
Smaller things
- Pricing page corrected. Professional ($249/mo) and Firm ($499/mo) tiers now display the prices Stripe actually charges. Previous values ($490 and $1,499) were marketing inflation that didn't match the underlying SKUs.
- Lead pipeline. The Google Places attorney-firm scraper now expands to all 263 cities across all 50 states + DC (was 50 cities, 29 states). Runs as a paced background job that auto-stops at the $190/mo Google free-tier cap. New firms flow through the email-discovery + NeverBounce pipeline nightly.
- Save-guard for case analysis. Long Claude streams that get truncated by network drops or upstream timeouts now refuse to overwrite a complete prior analysis. The frontend shows "Analysis was interrupted, please re-run" instead of silently saving partial output.
- Sequential publishing schedule. Four new SEO posts queued for May 12, 14, 19, 21 (Plaintiff Deposition Checklist, Nursing Red Flags, Statute of Limitations by State, Wrongful Death case-building).
What this means for you
Plaintiff attorneys + LNCs: If you've been using Case Record Search to pull citation-ready Bates stamps for memo-of-law footnotes, you can now copy them directly without verifying — they actually match the page they're on. The "View Bates page" speedup also makes large-file workflow (1,000+ page productions) practical instead of frustrating.
Court reporters + paralegals: The dominant-prefix detection works on any consistent Bates scheme: JOHNSON_, SCI, RECORDS_, PBC, MR, CV, AA, RDP, JEFF, etc. — all live across reindexed customer cases. If your firm uses a non-standard prefix (3+ letters, separator or 5+ zero-padded digits), it should auto-detect.
Try the rebuilt search on your own records.
14-day free trial · no credit card · all 23 tools included.
Start Free Trial →Questions? Reply to [email protected] — I read every email.