This proposal is shared privately with Ashish Dole ahead of the Mon May 4 review. Enter the access code to continue.
Butter watches the Encompass e-folder, applies your guidelines literally, and writes a hyperlinked pre-approval back into the loan file before your UW opens it. Built so your senior UWs do more with less, your CAC drops, and your securitization files arrive cap-markets-ready at underwriting time. If v0.1 doesn't ship in your sandbox in two weeks, Phase 0 is free.
Three swimlanes. Each one starts with a doc landing in your e-folder and ends with a number you can hand to a board.
Every PDF classified, every gap flagged with §-cite.
§4-2-6 income, §4-3-7 CLTV, §4-5 AVM applied verbatim.
Pre-approval lands in Encompass before the UW opens the loan.
Senior UWs stop sorting PDFs. They open files decision-ready.
Fat-finger reconnect on SSN, DOB, income typos before deny.
CMO sees which lead segments actually clear UW.
Rates drop, dormant pipeline rescored and surfaced.
Marketing spend lands on loans that close. CAC compresses.
Every value hyperlinks to its source PDF + §-cite.
One rule, every loan. Documentation variance closes at UW time.
Audit packet ships with the loan, not after.
AAA expected-loss line compresses. Bps your investors keep.
One agent. Three lanes.
Each compounds into the next.
1 → 2. Better UW drives better CAC.
2 → 3. Better CAC feeds tighter securitization.
Click below to test Butter, see how it responds to 4 scenarios.
1. Underwriter view for the person opening the loan today
2. Audit view for the cap-markets analyst opening the same loan three months later before securitization
401K_Linda_Q1_2026.pdf page 1 line 14.
W2_2025_*.pdf + W2_2024_*.pdf + Paystub_*_YTD.pdf.
Credit_TriMerge_Joint.pdf trended-data section.
0x8a3f...c2e1.
→ See Section 03·b for where this PDF lands in your e-folder + the integration architecture.
Butter is not a new tab, not a new tool, not a new training program for your UWs. Butter is a thin agentic layer over the Encompass tenant you already run. It watches the e-folder, runs a pre-approval pass, and writes the answer back. Your underwriter opens the loan and the work is already there.
Two new attachments per loan, written by Butter, named consistently, hyperlinked to source.
A thin agentic layer between your UW and Encompass. Reads the e-folder. Runs §4-2-6 / §4-3-7 / §4-5. Writes back. Encompass untouched.
Tom validates the path in Free Week 0 before any $10K commitment. If none work for your IT setup, we re-scope and you owe nothing.
Service account in your tenant. Document API v3 for reads, Attachment API for write-back. Standard ICE pattern. Production-ready Day 1.
Same data surface, different pipe. Battle-tested. Slower to build than REST. Falls back here if your tenant's REST endpoints are scoped down.
Functional v1 prove-out only. Used if A and B are both blocked. We re-scope before Phase 0 starts if this is the path.
Every rule in your guidelines compiles to a node in this tree. 8 phases. 50+ rule families. 298 §-cited decision points. Click any node to see the doc text, the formula applied, and the edge cases. The three places where the doc and verbal disagree are flagged inside the tree and surfaced again in Section 05 below.
If you want to inspect the engine before signing, this is the source of truth. Run your team's read against the tree. Every disagreement becomes a Phase 1 Week 1 alignment with Ryan / Kim / Firas.
We read your guidelines literally before drafting any code. Three places came up where the doc and earlier conversations gave different answers. None are blockers. We flagged them because the prior three vendors didn't, and that's the trust gate. Butter follows the doc until you tell us otherwise. Phase 1 Week 1 is where Sid + Tom sit with Ryan, Kim, and Firas to align the overlay logic that lives in their judgment but not in §X.Y, and codify it into the production engine.
"20% of 401K value, ÷ 12 → monthly income."
Asset depletion: (eligible liquid assets − closing costs − reserves − 10% penalty if applicable) ÷ loan term in months.
For Linda's $300K 401K, the verbal framing yields $5,000/mo; the PDF formula yields $736/mo. Demo math above runs the PDF formula as written ($736/mo, traceable line by line). Phase 1 Week 1 codifies any in-practice overlay with Ryan / Kim / Firas as a written §4-2-6-13b addendum that becomes citable in TPR review and Hildene/Goldman securitization memos.
"Single AVM. No double-checking. No second AVM pull."
Two AVMs required for $250K–$400K loans (one with FSD ≤ 0.10, or two with FSD ≤ 0.20). Single AVM path for ≤$250K.
"Average credit utilization weighted by payment history, not snapshot bureau pull."
Section requires "documented utilization analysis" but does not name the calculation method. Implementation is operationally inferred from underwriter practice.
Phase 1 Week 1 closes all three before production code touches your e-folder.
Four credible ways to ship HELOC underwriting AI at Button. SEI or a 4th vendor. Jason's $10K LinkedIn coding-contest hire. Butter with 100x. Or Tinman with Better. Five MECE dimensions below, instant-read icons. Phase 0 lets the data pick.
If you'd rather redeploy Tinman, Nik can introduce you to Vishal Garg @ Better. Note: Better could be a potential competitor, so tread carefully.
Phase 0 + Phase 1 = $110K total, ~3.5 months to production. You keep the code. You run it in your tenant. You hand the audit packet to the analyst Hildene puts on your next deal.
Early bird only.
Phase 0: $10K or $0 if we fail in 10 days.
Phase 1: $100K (60% discount on $250K, 100x.inc/pricing)
Phase 2: Pick your support model (2A DIY or 2B $100/close)
Phase 3: Optional future
Prove Butter integrates with your Encompass tenant. Three pass criteria: auth working, classifier + W-2 calc shipped, demo accepted. Miss any → 100% free.
Take Butter from Phase-0 50% accuracy to 95% production accuracy across all six income paths. Output: a Butter command + control center your team owns. 1 onsite FDE + Nik PT + 1-2 AI Engineers.
You run Butter. Your team operates the command center. We license + provide 2 FTE support.
We run Butter. One price covers everything: support, maintenance, command center, audit reports, AI tokens, infra, drift management.
Five expansion paths. Pick zero, one, or all five. Each is a separate Phase 3+ engagement.
P&L view per loan for the CFO. Real-time margin, attribution, channel-partner economics.
Borrower feedback agent (chat / voice / email / social AI). Closes file gaps before deny.
CMO feedback engine. Which lead segments clear UW, where to reallocate spend.
Head-of-Cap-Markets feedback engine. Pre-deal pool review, Hildene/Goldman analyst handoff.
Scale Butter to DSCR, Bank Statement, HELOAN, Closed-End Second.
Phase 0 starts the day mNDA + Encompass auth land. The $10K clock doesn't start until both. Each phase is an option, not a commitment. You decide whether to step up.
100x builds named AI agents in real ops roles. Sophei runs servicing for Home.LLC. Angei supports treasury at Angel Oak. Ernie powers product ops at Opendoor. Three verified partners. Zero pilots.
Nik is the C-level whisperer. He knows exactly how to cut through the noise and deliver clarity that moves mountains.
Thank you! Couldn't have done this without your kick-ass team. They absolutely crushed it!
SOC 2, GDPR, CCPA, TCPA, FINRA all cleared on our end (same as SEI).
All Button PII processed inside a Button-dedicated workspace. Inference runs against an enterprise LLM contract that does not retain customer data for training. Standard SOC 2 Type II controls (logging, encryption at rest + in transit, access reviews). Pen test report and SOC 2 letter available on request before contract.
All artifacts produced during Phase 0 + Phase 1 (pre-approval files, classification training data, decision-logic specs, integration code) become Button property at end of engagement. If 100x can't continue, Button keeps Butter's logic and routes the operational layer to a successor. This is in the SOW, not just a promise.
Integration architecture moved to Section 04·b — three integration options (REST API · SDK · TPO Connect) live there with the stack diagram.
In 100 days, get what others will take 10-months.
Founded 100x. Built Sophei to save Home.LLC. Executive interface with Ashish, Jason, and C-suite.
Owns Phase 0 to Phase 2. UW "discovery" sessions with Ryan / Kim / Firas onsite during Phase 1 to rapidly iterate Butter from 0% to 95%.
Owns Butter's engine, ask him anything (e.g. doc classifier, §4-2-6 income paths, etc.) He already built 298-rule engine to align with your 65-page guide, ask him about your errors.
Owns the pipe to Encompass integration. REST + Document API, service-account auth, write-back, TPO Connect fallback and will keep up with Encompass changes.
$10K. Ten days. Working Button inside your Encompass sandbox. Anything we get wrong by Day 10, you walk and pay $0.
Every quarter the documentation tax stays in your files, your CAC compounds, your UWs grind, and your securitization pool ships with avoidable variance. Phase 0 is ten days to find out if that gap closes.