KYLNOR.COM

Ventures

I build things I needed and could not buy. More than twenty years selling and running real estate operations taught me exactly where the workflow breaks, and lately the tooling caught up to the point where one person can close the gap. So I do. Some of these are real estate products, some are developer tools, some are the plumbing that runs the rest. That range is not a scatter. It is the same operator solving whatever problem is in front of him, in whatever domain the problem happens to live. What follows shipped. Each one started as a specific frustration and ended as software with a URL.

IN DEPTH03

CorrectCMA

LIVE · IN PILOT
WHAT IT IS
A pricing and CMA tool for real estate agents. It takes a listing, pulls the comparable sales, and produces a defensible value with an honest confidence band, a full chart library, and a client-ready presentation. It is a real blended automated valuation model, not a spreadsheet with a logo.
WHY I BUILT IT
Pricing a listing is the daily high-stakes decision in this business, and every tool on the market hands you a number without telling you why. Agents get "Zillow says X" and no way to defend it or correct it. I watched my brokerage evaluate buying a competitor's platform, watched leadership pass, and decided the honest version was buildable. So I built it.
WHAT IT PROVED
The engine measures things the category asserts. On a corpus of 44,871 closed sales it quantified the real cost of mispricing: homes priced right give up about 2.8% and sell in about 34 days, overpriced homes give up about 10.9% and sit for over 100. That is the wedge, and it is measured, not claimed. The harder proof is the moat: comps are selected by parcel geometry, not by matching subdivision text, so the tool separates a golf-course estate from the condo two streets over that shares a neighborhood name. A live pilot exposed exactly the failure this prevents, a listing that text-matching valued at $6.6M and the geometry gate brought back to reality. The whole thing runs the valuation in the agent's own browser, which is also the compliance firewall: the servers never touch raw MLS rows. Applied valuation, geospatial analysis, and a real compliance posture, in one product.
WHERE IT LIVES
Live and auth-gated at correctcma.com. In pilot on my own Las Vegas listings, with Boise lined up as market number two. This is the one I am building now.

RoofDorks

BUILT · GATED PREVIEW
WHAT IT IS
The operating system for a roofing contractor's entire business, marketing, sales, and operations in one platform. Lead generation and a sales pipeline bring the job in. A measurement engine reads the roof off free public LiDAR and turns it into an accurate quote. A supplement engine recovers the money insurance carriers leave on the table. The trade runs on spreadsheets and a stack of $20-a-report point tools. RoofDorks replaces the stack with one system.
WHY I BUILT IT
Same instinct as everything else, aimed at a whole industry instead of one task. A roofer's day is a chain of manual, expensive, repeatable work: chasing leads, measuring roofs, writing quotes, and fighting carriers for the money an adjuster left out. Every link in that chain was a product waiting to be written, and nobody had written the chain. The measurement had a free data source nobody was using. The supplement was hand-work that is secretly a diff engine. The CRM was the thread that ties the job from lead to paid. So I built all three and wired them together.
WHAT IT PROVED
Three genuinely hard problems, solved and unified. Measurement: pitch comes back bind-quality at 4cm plane accuracy, the precision that justifies steep-roof line items on a real claim, and it pulls 16 of 18 facets on a live house off free USGS LiDAR, the job roofers pay EagleView twenty dollars a report for. Area lands within about 10% on a tree-covered lot, and I state exactly why: canopy hides roof the scanner never sees, and I would rather name the ceiling than fake precision in a fraud-sensitive space. Supplements: the scope-compare engine diffs a carrier's estimate against ground truth across the five levers where the money hides, measurement gap, code-driven adds, waste, overhead-and-profit, and improper depreciation, and it has run clean on live claims. On one real Allstate claim it extracted 15 of 15 line items with zero fabrications and built a priced, carrier-facing rebuttal letter off the carrier's own numbers. That is not a template, it is encoded domain expertise. Operations: a real CRM with lead capture and a sales pipeline, the front of the business, in the same platform as the measurement and the money. Geospatial signal processing, insurance-claim expertise, and full-stack SaaS, in one product for one trade.
WHERE IT LIVES
Built and consolidating under the RoofDorks brand. The supplement engine is validated on live insurance claims. Built with a roofing subject-matter-expert partner whose real claims are the validation set. Password-gated preview, pre-public-launch. This is genuine domain depth, not a demo, and it is honest about being early: proven on the supplement side, not yet at scale.

Unquill

LIVE IN PRODUCTION
WHAT IT IS
A multi-tenant publishing platform built on Cloudflare's edge. Each publication lives at its own subdomain, articles serve from 300-plus points of presence, and the whole thing is engineered to be found and cited by AI systems, not just read by people.
WHY I BUILT IT
I needed somewhere to publish that I actually owned, and I did not want to rent a CMS to do it. When you are the landlord, the platform bends to your products instead of the other way around. My content engine needed a home, my own writing needed a home, and building the host was less work than integrating three of somebody else's.
WHAT IT PROVED
This is systems engineering, not a blog. Metadata lives in a SQLite database at the edge, article bodies and images live in object storage, newsletters go out through a mail layer, and the serving tier will not be outgrown because that is what edge workers are for. The distinctive bet is answer-engine optimization treated as the actual product: structured data on every article, a crawler policy that explicitly welcomes the 16 AI bots most sites block, and a machine-readable map of every publication for models to discover. It hosts several live publications today and has a clear, deliberate path to scale that I chose to defer rather than over-build.
WHERE IT LIVES
Live in production, serving multiple publications on Cloudflare Workers, D1, and R2.
ALSO SHIPPED05

Lararium

PUBLISHED · LIVE
WHAT IT IS

A free, clone-and-run scaffold for a personal AI system. One command, npx lararium, and you get the whole architecture: a file-based knowledge store, a persona that survives across sessions, an agent roster, the hooks that make it remember yesterday. It ships completely empty on purpose.

WHY I BUILT IT

I kept rebuilding the same scaffolding every time I set up an AI assistant to actually know my life. So I extracted it, stripped every trace of my own data, and gave it away. The bet is that the architecture was never the moat. Your life is, and a second brain cannot be copied off a shelf because it has to be lived.

WHAT IT PROVED

That I can package a hard-won personal system into something a stranger can run in one line, and that I ship developer tooling to a developer audience, not just products to my own industry. It even installs itself: open the folder in an AI coding assistant, say "run the install interview," and an agent writes your files while you answer.

WHERE IT LIVES

Published and live. npx lararium, MIT licensed, at elorati.com/lararium.

ContentFurnace

LIVE IN PRODUCTION
WHAT IT IS

A content service that runs itself. An AI editor named Margot texts an agent, learns their voice, writes them a piece, and sells it one article at a time. The client approves by replying to a text, gets charged, and the article publishes and fans out into social assets on a private dashboard.

WHY I BUILT IT

Agents know they should be publishing and never do, because the friction is real: briefs, drafts, revisions, formatting, distribution. I collapsed the entire pipeline into a text-message conversation. Approve by SMS, everything downstream is automatic.

WHAT IT PROVED

A full revenue loop with a real AI persona on the phone and real money moving. The charge fires on approval, the payment clears, the piece goes live, and the assets land, end to end, proven with real dollars. The genuinely hard parts were the ones nobody sees: teaching Margot to tell a status question from a content idea, and guarding every path where free work could be farmed.

WHERE IT LIVES

Live in production on a cloud host with live payments and a working publishing pipeline.

GetProval

LIVE · INTERNAL
WHAT IT IS

A proposal engine for service businesses. A sales-call transcript becomes an editable, line-item proposal. The client accepts it on a web page, and that single act generates a signed agreement and pushes the onboarding tasks into a project manager. The name is the pitch: get approval.

WHY I BUILT IT

Proposals are the number-one bottleneck at a service agency, and the owner is usually the constraint, weeks behind, closing well when they finally send. The transcript-to-proposal step is the unlock. The rest of the market drafts from a template. Drafting from the actual conversation is the difference.

WHAT IT PROVED

A closed loop nobody else ships: transcript to proposal to acceptance to signed contract to project tasks, where the contract and the tasks can only ever read the line items the client actually accepted, so a proposal-versus-contract mismatch is impossible by construction. The money path is live and verified end to end, with the entitlement gate enforced before generation, not after.

WHERE IT LIVES

Live at getproval.com. Subscription billing wired and proven. Currently in use internally, seeking its first outside agencies.

FurnitureSpotter

LIVE · IN TESTING
WHAT IT IS

A simple tool for real estate agents that runs an AI furniture audit on a listing and produces a client-ready addendum PDF, the document that spells out what conveys with the house and what does not.

WHY I BUILT IT

It is a small, annoying, recurring piece of listing paperwork that people wanted done for them. Two agents asked for it directly. It was the right size to ship fast and prove the umbrella could turn a request into a live product in a week.

WHAT IT PROVED

Speed to ship, and distribution built in rather than hoped for. Every signup gets one free audit, and a referrer earns a credit only when their referral's first paid checkout clears, capped and non-farmable. The referral loop is the growth mechanism, wired into the product from day one, not bolted on later.

WHERE IT LIVES

Live at furniturespotter.com, out with testers.

Atrium

LIVE · INTERNAL
WHAT IT IS

The internal operating app for Elorati, the software company I run with two partners. One place to code, track, sell, and show the work, including a dashboard that measures the hours our automations save clients.

WHY I BUILT IT

Three people building for clients need one surface to run the business on. I stood it up in a weekend by forking a CRM spine I already had and reskinning it.

WHAT IT PROVED

That the company has its own operating system rather than a folder of spreadsheets. It is honestly early: it is live, but it has no outside tenants on it yet. I am keeping it as internal infrastructure until a real client and a real automation have run the whole loop.

WHERE IT LIVES

Live and internal at atrium.elorati.com.