SiftStack setup in 3 to 5 days

Stand up Claude Code inside VS Code, wire it to the repo and put your credentials in an ENV file, then let it pull, clean and upload records every morning.

• What the SiftStack is: Claude Code running inside VS Code wired to the public GitHub repo, your credentials in an ENV file, and the full skills library operating as one system; an operating layer, not
• Realistic setup expectation: 3-5 dedicated days (the old "about 45 minutes" claim is retired); install VS Code and Claude Code, paste the repo, "implement every single thing inside it," fill the ENV (
• What it automates end to end : public-notice scraping with 2Captcha and ScrapFly, scanned-PDF OCR import, courthouse terminal photo import for all 8 notice types (liens, tax delinquent, tax sale, foreclosure, probate, obituary, code violations, condemned properties; the list Ty pulls on tape, D1 3:36)
• The assembly-line data pipeline SOP the stack executes : (1) pull SiftMap precision same day (AI score band, acute distressors as parallel pulls, native obituary) with auto-add on, (2) clean and skip
• The consistency threshold: a single precision strategy needs ~200 new records per month or deal flow gets lumpy; stack niches to reach it
• The stack can go the other way on entities when you choose to: buy-box configuration flags (--include-vacant, --include-commercial, --include-entities) adapt the pipeline to land, commercial, or entit
• Governance and hygiene: fork the public repo (read-only upstream), keep credentials in ENV, close sessions by updating the memory files (9.3), and staff it with ONE operator hired for an engineering m
• Setup 3-5 dedicated days
• 200 new records/month consistency threshold
• 8 notice types, 41-column upload CSV, 12 + 9 presets, 26 sequences
• The assembly line: courthouse/SiftMap in, seven stations (pull, clean, skip, tag-route, score, sequence, deep prospect), marketing-ready records out
• The 50-cent factory: a tiny server icon stamping out enriched records with a $0.50/day meter
• The tag switchboard: one tag lever routing a record to niche vs bulk vs invisible (the bug)
• The 20% iceberg: clean records above the waterline, incomplete/entity records below, with a small "entity researcher" submarine
• Incomplete records and the LLC policy: roughly 20% of pulled records come in incomplete and about 95% of those are entities; the pipeline routes them to enrichment or the entity skip, never to marketing