A 5 BR, 3.5 BA renovated home on a rare 9,000 sq ft flat triple lot in Aurora Highlands — one block to Pentagon City. Listed at $1,785,000. The $45k a 2.5% agent would charge stays on the table.
Every decision — pricing, staging, vendor selection, marketing, this very page — made by a language model. I'm just the hands. A public market test: you're watching a $1.7M+ home price-discover itself under an AI-run sale.
The triple lot is three original 25' × 120' platted parcels (Lots 5, 6, 7 Block 20, Addison Heights) — one block to Pentagon City Metro. Refinished light-stain hardwoods, covered front porch, rear deck, real front and back yards, oversized 1.5-car garage plus driveway for 7 cars. 2022 Trane dual-zone HVAC, Tesla Wall Connector, active radon mitigation, $70K in 2022–2026 invoice-backed improvements.
Rare 9,000 sq ft flat triple lot in Aurora Highlands (22202). Refreshed 5 BR / 3.5 BA, ~3,500 sq ft home (3,400+ measured) with a large covered front porch, rear deck, and spacious yard. ~$70K in invoice-backed 2022–2026 improvements. Suburban-scale yard with urban walkability.
Walkable to Crystal City Metro, Pentagon City Metro, Amazon HQ2, Virginia Tech Innovation Campus, Costco, Whole Foods, Pentagon City mall, and the Pentagon itself. Long Bridge Park and the Mt. Vernon Trail within walking distance.
Large covered front porch and a rear deck. Three original 25' × 120' platted lots (Lots 5, 6, 7 Block 20, Addison Heights; 9,000 sq ft total) give you a proper front yard and backyard — scale that simply doesn't exist in most of the neighborhood. East-facing front with morning light. Professionally landscaped front and back with new sod, new trees, and seasonal plantings. Oversized 1.5-car garage plus a driveway that fits 8 vehicles — unusual inside the neighborhood's permit-parking zones.
Refinished light-stain hardwood floors. Updated lighting, plumbing, and vanity fixtures throughout (2026). Gas-log fireplace. Vinyl double-pane windows. Separate exterior basement stairwell gives the lower level its own walk-out access — potential in-law or guest suite.
New KitchenAid 30" gas cooktop (2026), new LG 27 cu ft InstaView side-by-side refrigerator with Craft Ice (2025), and an LG 5.0 cu ft Mega Capacity front-load washer (2020). Washer and dryer included.
The 5th bedroom conversion is complete and is awaiting Arlington County inspection and permit approval, which will be finalized prior to closing. Included a code-compliant basement egress window professionally installed (April 2026) and hardwired smoke detectors.
Two independently zoned systems.
Upper zone (2nd floor + attic) — 2022 Trane install, under Trane manufacturer warranty:
Lower zone (first floor + basement):
Also: active radon mitigation. Tesla Wall Connector Gen 4 (48A / 11.5 kW, NACS + switchable J1772 adapter) outside the garage.
Abingdon Elementary · Gunston Middle · Wakefield High.
All improvements and refresh decisions on this home were made by the seller in consultation heavily with AI tooling.
A short cinematic walkthrough — the way the light moves through the rooms, the flow from kitchen to porch to yard. Watch this first, then dive into the full 3D tour below.
Walk through every room, or flip to floor-plan view for dimensions.
Pick a day if you know which one. Just want the post-mortem when it sells? Same form — leave the day blank.
There is no direct comp for 909 in Aurora Highlands today. AI built the analysis the way an appraiser would: hold what matters constant, adjust everything else, and triangulate from independent angles.
| Address | Status | Price | 2026 adj* | Finished sf | Lot (sf) | Built | $/sf land |
|---|---|---|---|---|---|---|---|
| 909 19th St S This home |
Active | $1,785,000 | $1,785,000 | 3,438 | 9,000 | 2005 | $198 |
| 921 22nd St S Same parcel size, parallel street |
Coming 5/16 | $2,575,000 | $2,575,000 | 5,563 | 9,000 | 2026 | $286 |
| 637 21st St S | Sold 12/25 | $2,340,000 | $2,385,000 | 4,554 | ~4,500 | 2024 | $530 |
| 1015 20th St S | Sold 8/25 | $2,225,000 | $2,300,000 | 4,634 | 5,663 | 2025 | $406 |
| 2155 S Ives St Age peer · 2009 build |
Sold 2/24 | $1,650,000 | $1,840,000 | 3,968 | 6,238 | 2009 | $295 |
*Time-adjusted to April 2026 at 5%/yr Arlington single-family appreciation (Bright MLS 12-mo median, 2024–26).
| Comparable | 2026 price | Vintage adj1 | Size adj2 | Lot adj3 | 909-equivalent | vs $1.785M list |
|---|---|---|---|---|---|---|
| 921 22nd St S 5,563 sf · 9,000 lot · 2026 |
$2,575,000 | −$273,000 | −$319,000 | $0 | $1.983M | +$198K · +11% |
| 637 21st St S 4,554 sf · 4,500 lot · 2024 |
$2,385,000 | −$320,000 | −$167,000 | +$652,000 | $2.550M | +$765K · +43% |
| 1015 20th St S 4,634 sf · 5,663 lot · 2025 |
$2,300,000 | −$296,000 | −$179,000 | +$484,000 | $2.309M | +$524K · +29% |
| 2155 S Ives St 3,968 sf · 6,238 lot · 2009 age peer |
$1,840,000 | −$19,000 | −$80,000 | +$401,000 | $2.142M | +$357K · +20% |
| Mean / median adjusted value | — | — | — | — | $2.246M / $2.226M | +$461K / +$441K |
1 Vintage: structure value (price minus land) depreciated from comp's age to 909's age 21 on standard appraisal curve. 2 Size: (comp sf − 909's 3,438 sf) × $150/sf marginal cost. 3 Lot: (909's 9,000 sf − comp lot sf) × $145/sf land basis.
| Method | Date | Price | 2026 adj | Lot (sf) | $/sf land | Note |
|---|---|---|---|---|---|---|
| 921 22nd St S — raw land | Jul 2023 | $1,100,000 | $1,260,000 | 9,000 | $140 | Direct teardown sale, identical parcel size |
| 637 21st St S — extracted basis | 2024 teardown | ~$820,000 | ~$900,000 | ~6,000 | $150 | Implied land basis from $2.34M new-build sale |
| 909 — triangulated land value | Today | — | $1.26M – $1.35M | 9,000 | $140 – $150 | Two methods converge |
Take the midpoint of the land range — $1.30M for the dirt. That leaves roughly $485,000 attributable to a 3,438 sq ft 2005-built structure with 5 BR / 3.5 BA and a finished walk-out lower level. That's $141/sf for the structure alone — well below current Northern Virginia replacement cost of $325–$400/sf for comparable build quality. A buyer at $1,785,000 acquires the lot at fair market and the structure at a meaningful discount to building new.
All comps from Bright MLS, Arlington County tax records, and Redfin/Zillow public listings as of April 2026. Full CMA, disclosures, and inspection report available on request.
Every line item is receipted. Roughly 55% major renovation, 21% a 2022 Trane HVAC system still under warranty, 12% the 2026 code work that turns the basement into a legal 5th bedroom.
General contractor & handyman ($13,000) · floor refinishing — light-stain hardwoods ($14,000) · front + back landscaping with new sod, trees, seasonal plantings ($11,250)
2022 Trane HVAC — 16 SEER A/C, variable-speed gas furnace, Aprilaire 800 steam humidifier, 2× UV air scrubbers, maintenance plan ($13,700) · Google Nest install ($440.33) · Nest hardware + 2 temp sensors ($242.74)
Clifton Contracting basement package — code-compliant egress window + 2 hardwired smoke detectors, April 2026. The work that legally enables the 5th bedroom.
KitchenAid 30" gas downdraft cooktop ($2,249.99, Best Buy) · LG 27 cu ft InstaView side-by-side refrigerator with Craft Ice ($1,801.99, Costco 6/2025)
Gorilla Playsets "Outing w/ Monkey Bars" cedar swing set, Aug 2021. Professional install figure pending.
Jensen medicine cabinets, replacement bulbs and hardware throughout.
LG 5.0 cu ft Mega Capacity front-load washer in black (Costco 11/2020).
Full vendor log available on request — AI will email it to you.
Transparency is the whole product. Here's the actual log of how AI ran the sale — pricing, prep, vendors, marketing, copy. The detail is the credibility.
After scanning recent Aurora Highlands sales, AI concluded only 9 lots in the neighborhood share this triple-lot footprint. It proposed a price inside a narrow band; I chose to list slightly under to test demand velocity. The decision to undershoot is mine; the range isn't.
> propose_list_price(comps=n:42, constraints=triple_lot=true) → $1.78M–$1.82M
> choice(strategy="test demand velocity") → $1,785,000
Frame by frame: refinish floors lighter, swap dated ceiling fans, add landscape spotlights, power wash the deck, fix every small thing a buyer notices in the first 30 seconds. I did the labor — AI ran the prioritization.
AI returned a replace/keep list with source links. Replacement bulbs from Home Depot, swap-in vanity lights from Amazon, a specific brushed-nickel faucet for the primary bath. Zero aesthetic disputes because there was no one to dispute with.
AI identified that an egress window would reclassify the finished basement room as a conforming bedroom under Virginia code. It scoped the job, specified the window size, and drafted the contractor email. Two days of excavation, one inspection, +1 BR on the listing.
It also drafted the outreach emails, monitored replies, and flagged when a vendor went cold. I signed every contract. AI wrote every first draft. No cold calls, no awkward back-and-forth.
Including the tagline, the color palette, the fence post type, and the print shop. Every element carries a small made by AI tag where you'd expect a credit line. It's self-referential on purpose — if AI can run the sale, it can build the artifacts that advertise the sale.
Most of what a traditional agent charges 3% for is decisions: price this at X. Stage like this. Swap this fixture. Reply with this tone. Push back on that clause. That's the judgment layer. AI does it now — quickly, cheaply, and in my experience, well.
What no software can do is climb a ladder, crimp a wire, drive to Home Depot at 9pm, or reposition a hose when the sprinkler timer misfires. I did those things. Hundreds of them. Over two weeks, with a newborn in the house. It was a lot of work, even with the decision burden removed.
So here's the honest take: I don't think most homeowners should do this. Not yet. The interesting finding isn't AI replaces realtors. It's that the judgment layer is now solved, and the execution layer is still entirely human. A robot in your house changes that equation. We're not there. But we're closer than I expected when I started.
Yelp, Angi, Handy, Thumbtack — they don't rank by quality. They rank by who pays to advertise and who has time to chase reviews. The contractors at the top of those lists charge more and aren't usually the ones doing the best work. The genuinely outstanding ones are heads-down, fully booked through word-of-mouth, and basically invisible online.
So I built a different way to find them.
I'll write this method up in the post-mortem. It's the most generalizable thing I learned.
TL;DR — The experiment is: when an AI runs the sale instead of an agent, can the market tell the difference? We're about to find out.
I'll share the full prompt log, the vendor outreach archive, and the things I got wrong. No NDA, no embargo.
Writing about AI agents, real estate disintermediation, or any of the above? I'll send the full materials — prompt log, vendor outreach archive, mistakes included.
This is one experiment, not a startup. There is nothing to sell you.