XMarts room

Fall 2026

Tire Swing

AI Compliance for Affordable Housing

Visit website

What they do

AI runs resident eligibility certification for affordable housing

How close to us

4, very close

one compliance layer for one firm type, end to end; size unclear, no checking described.

Wedges

  • W5 One layer for one type of small firm adjacent
L1 Company brain (knows the organisation) 1, light
L2 Agent team (runs agents, tools, approvals) 1, light
L3 Does and learns (finishes, checks, measures, updates the brain) 2, clear

Who buys

Firm size
not known
Buyer type
business
Role
property compliance staff at owners and managers of subsidised housing
UK
no
Their own description (from YC)

Tire Swing uses AI to power resident and applicant eligibility evaluations for affordable housing.

Our software manages the resident certification and recertification process end-to-end for public and private properties supported by Project Based Rental Assistance, Low Income Housing Tax Credit, and Housing Vouchers. It serves as a drop-in replacement for third-party consultants and auditors, and helps compliance staff complete their work faster and more accurately.

Full teardown

Tire Swing (YC Fall 2026): teardown vs XMarts

In one line: Tire Swing uses AI to read tenant paperwork, work out household income against housing-programme rules, and chase missing documents, for US affordable housing properties [site] [yc]. Threat: 1, because it is the same shape as our W5 (one vertical layer that replaces outside consultants) but for a US housing buyer we do not target [yc] [site].

Batch warning: the YC page lists Fall 2026, but a founder bio on the same page says "W25", and the company's own site says "Backed by Y Combinator (W25)" [yc] [site]. A third-party directory describes a health-insurance compliance product under the same name [press]. A search also showed a second YC launch post titled "Healthcare Compliance with AI" that I could not open [yc]. Possible reading: a W25 company that moved from healthcare compliance to housing and is listed again (inference, UNVERIFIED). The founders should check this before using Tire Swing as a "what YC funds now" example.

1. Fact file

Item Finding Source
What they say Site headline: "Faster, cleaner income certifications". YC line: "AI Compliance for Affordable Housing". [site] [yc]
Founded, team, funding Founded 2025, 2 people, New York City. No funding amount found. YC primary partner: Grey Baker. [yc]
Founders Paul Witten (CEO): Princeton CS 2019, then senior consultant at Stroz Friedberg on cyber investigations. Lucas Irvine (CTO): Princeton CS 2019, Instacart front-page engineer, earlier Capital One (open-source rules engine Cloud Custodian). They are college friends. What YC likely saw: two strong engineers who know rules engines and security, with SOC 2 already claimed (inference). [yc]
Customers None named on the site, the YC page or the launch post. No revenue or usage numbers. [site] [yc]
Price Not public. Demo only (30-minute session). No job ads to hint at a price band. [site] [yc]
Path Founded 2025. Site and founder bio say W25. YC page says Fall 2026. Launch post for housing is about 3 months old (about late July 2026, inference from "3 months ago"). A health-insurance compliance version of the product is listed by a third-party directory. [yc] [site] [press]

2. Their proposition, reverse engineered

Part What they do
Who buys Owners, managers and compliance teams of subsidised properties. Their ask is for property managers, developers, public housing authorities and state housing agencies [yc]. Size of buyer unknown: no customers named.
The pain Slow move-ins cost weeks of rent. Mistakes can cost "credit recaptures and fines" in the millions (launch post paraphrase). Outside consultants and auditors are paid to check the files [yc].
The promise "Drop-in replacement" for third-party consultants and auditors. Reviews that took weeks take minutes (they claim) [yc]. Live in "two weeks" with "no IT lift" on the platform page, "under a week" on the home page (the two pages differ) [site].
How it works Reads paystubs, W-2s, benefit letters, bank statements and scans, including handwriting [site]. Applies programme rules and each property's own forms and policies [site]. Gives a certification with each value linked to its source, plus a confidence score [site]. Sends requests for missing documents, forms and e-signatures by phone, email and text [site]. A human reviewer can override, and each override is logged [site]. Everything sits in one queue of files [site].
The output A finished certification packet marked "Ready to send", findings with citations and next steps, and an audit log of every field, flag and override [site].
Proof style Sample screens only. The site says the demo figures are illustrative, not verified results [site]. Claims SOC 2 Type II and a trust centre [site]. No customer numbers.
Who sells it Founders, by demo and warm introductions. The launch post asks for introductions [yc]. No sales hires visible [yc].
Price model Not public [site].
The story The enemy is slow, manual file review done by costly consultants. The belief: affordable housing is the first of many "needs-based eligibility" jobs. They cite over $1.5 trillion a year in programmes such as Medicaid and SNAP, and about $20B a year to administer them (their figures) [yc].

3. Positioning

  • Category they claim: "AI Compliance for Affordable Housing"; "Agentic income certification" [yc] [site].
  • Category a YC partner would put them in: vertical AI that replaces an outsourced service (GovTech and real estate compliance), pitched with a large "next programmes" story (inference).
  • What they want to be known for: the audit-ready file that is correct before the applicant leaves the office [yc].
  • Where they sit on the map: layers L1/L2/L3 = 1/1/2, buyer = unknown size, business, US only.
  • Their rivals (public): ExactEstate sells affordable-housing property software with an income-certification flow, public price of $3 to $5 per unit per month [blog]. Pronto Housing (founded 2020, about $7.6M raised per CB Insights) sells compliance software plus managed compliance services [press]. EliseAI has launched affordable features (recertification notices, voucher pre-qualification) and hires affordable-housing compliance staff [blog] [jobs]. AppFolio markets its own affordable housing module [press]. Tire Swing names none of them (UNVERIFIED that its own site omits them everywhere).

4. Three layers vs ours

Layer Them (score + evidence) Us (from the system inventory, "Marton's logs show")
L1 Company brain 1. Uses the property's forms, policies and programme rules for one task [site]. No sign of a lasting memory that other work reads. Marton's logs show a vault of about 1,400 pages, 112 call transcripts and about 8,700 customer quotes, searchable by meaning. It holds Marton's business only.
L2 Agent orchestration 1. One document pipeline with outreach and a reviewer approval step [site]. Called "agentic" by them; no multi-agent routing shown (inference). Marton's logs show about 5,600 jobs handed to agents from three makers, each checked by a different AI before merge. This is code work for his own system.
L3 Execution and learning 2. Does the work (the certification) and checks it with rule checks, source links, confidence scores and a human reviewer [site] [yc]. No public claim that overrides or outcomes change the next run (UNVERIFIED), so not 3. Marton's logs show about 160 failure records that each added an automatic check. Offer feedback lines fell from 44 to 18 as 22 rules were added. Only about 15% of jobs reached the main code, and merges were unreliable in October.

5. Same and different

Topic Same or different Them Us
Core belief Same Do the work and show proof for each value, with a human in the loop [site]. Same: work is checked by a different engine and logged (inventory).
Replace an outside service Same Replace consultants and auditors [yc]. Same shape for W5 (brief), no vertical chosen yet.
Audit trail Same Every field, flag and override logged with source and reviewer [site]. Marton's logs show every failure gets a record and an automatic check.
Buyer Different US affordable housing operators; size unknown [yc]. SMEs and scale-ups, UK first (brief). No customer yet.
Kind of work Different One narrow, rule-heavy job: income certification [site]. A whole business loop: memory, agents, checks (brief).
Owner of the result Similar Reviewer signs; the software prepares the file [site]. Not set. Our claim to test is owning one number.
How it learns Unknown Overrides are logged [site]; no public claim they improve the model (UNVERIFIED). Rules added from failures and feedback (inventory).
Proof Both thin Illustrative demo figures, SOC 2 claim, no customers [site]. Own-business logs only; nobody else has run it.
Size Same 2 people [yc]. 2 founders (brief).
Price Different Not public; the nearest rival publishes $3 to $5 per unit per month [blog]. Not set.

6. Wedge overlap

Wedge Fit Why
W5 One layer for one type of small firm adjacent Same shape: one vertical job done end to end, replacing an outside service [yc]. Different buyer: US housing operators, not small UK firms (brief). The screen pass said direct; at deep pass the buyer and problem differ, so adjacent by the lens.

7. Whitespace (where they stop)

  • Layer gap: one job, one pipeline. No company memory beyond the property's rules, and no stated learning from outcomes (UNVERIFIED) [site]. To own it we would need a rule-change log tied to checked runs, with rework falling.
  • Buyer gap: US programmes only (HUD, LIHTC and others) [site]. They do not serve UK or European firms. To own it we would need UK SME buyers who say they have a similar rule-heavy back-office pain.
  • Number gap: no customer numbers, and the site says the demo figures are illustrative [site]. To own it we would need one workflow run for 4+ weeks with a before/after number the buyer signs.
  • Price gap and speed gap: not true here. They claim live in one to two weeks [site]. Price is not public, so no gap can be shown.

8. For the YC application

  • Proven today: Marton's logs show his system already runs agents whose work is checked by a different AI before it counts, and every failure gets a record and an automatic check (about 160 so far). Tire Swing's public proof is sample screens and a security claim; it names no customer [site].
  • Draft answer to "how are you different from Tire Swing?": DRAFT. "Tire Swing does one US housing job. We are not in that market. Our system runs and checks work across a business and learns from the checks, and we have run it on ourselves." Founders say it their own way.
  • Future claim (needs proof): "For one type of UK firm we do the whole back-office loop and own one number." Needs one customer workflow run for 4+ weeks with a before/after number.
  • What we understand that they don't: a result that is checked and logged should change the next run, not just fill an audit file. needs proof on a customer; Marton's logs show it on his own work (offer feedback lines fell from 44 to 18).
  • Copyability (M3): medium. They already log reviewer overrides [site], so learning from them is a natural next feature, months of work (inference). Chasing the whole business loop for other buyers would pull them off a narrow regulated niche. Not hard, so M3 cannot score above 3 against Tire Swing.
  • Why now (M4): None found. Do not use "agents can now read documents": that is general AI progress.
  • Do not claim: "no one does this for housing". That is not our market and we have not checked their rivals deeply. "SOC 2" or "audit-ready" for us: nothing in the inventory shows it. "Company brain" as a headline.

9. Ideas worth taking

  1. Every value links to its source, and every human override is logged [site]. It builds trust with a cautious buyer and is the raw data for a learning loop.
  2. Name the paperwork buyer's real fear: fines and lost rent, in the buyer's terms [yc].
  3. Show a security page and a trust centre early, even with two people [site]. Our open question is "how is client data kept safe?" (brief).

10. Unknowns

  • Which batch they are in (W25 or Fall 2026), and whether the housing product is a pivot from healthcare compliance. Look at the second Launch YC post and the Wayback copy of the old site.
  • Any customer, revenue or pilot. Ask in the YC Bookface or check their launch comments.
  • Whether the SOC 2 Type II report is real (it is a very fast claim for a company founded in 2025). Open trust.gettireswing.com in a browser; my fetch returned only a title.
  • Which AI model providers and how tenant data is kept (the security page names none) [site].
  • Whether they integrate with Yardi, RealPage or AppFolio. The platform page names none [site].

Sources

Technical detail
Slug
tire-swing
Lens hash
179096c479
Pass
deep
Stale
no