Every January, California opens a window. For roughly ten weeks, owners of older houses can register for a state grant worth up to $3,000, earmarked for one of the cheapest life-safety upgrades in residential construction: bolting a wood-frame house to its foundation. Then the window closes, the money gets divided, and the queue resets for next year. About 2,400 houses a year get fixed, in a state where as many as 1.2 million are vulnerable.
Divide those numbers. I did, and at the historical pace the backlog clears in five centuries.
The program is called Earthquake Brace + Bolt, run by the California Residential Mitigation Program, a joint powers authority of the California Earthquake Authority and the Governor's Office of Emergency Services (CRMP), and it addresses a specific, well-understood failure. Pre-1980 houses with raised foundations often sit on their concrete perimeters with too few anchor bolts, and the short wood-framed cripple walls in the crawl space can fold in a quake, letting the house slide or topple off its foundation. Anchor bolts go in, and plywood goes on the cripple walls. Two to three days of work, no need to move out, and the foundation connection performs to current code, according to the CEA's chief mitigation officer (Santa Monica Daily Press).
Eligibility reads like a building-code haiku: wood frame, built before 1980, raised foundation, level ground or a gentle slope, and a ZIP code among the 1,118 covered as of the 2025 expansion (CRMP FAQ). Registration opens for a limited window each year; in 2025 it ran January 15 through March 26. Miss the window and you wait a year, which is the first of several ways this program rations by calendar instead of by risk, a system that would be defensible if the calendar knew anything about soil conditions, fault distance, or framing age, but it does not.
The Arithmetic Nobody Prints
Methodology first, since the numbers deserve it. CRMP reported "more than 28,500" assisted homeowners in February 2025, twelve years after the 2013 launch, which averages roughly 2,375 homes a year (CRMP), a figure with the virtue of being official and the defect of sitting two orders of magnitude below the need. CRMP separately estimates "as many as 1.2 million" vulnerable homes (CRMP, 2021), which divides to about 505 years. Fund it at the generous 2025 level instead, more than $20 million a year, which buys roughly 6,600 grants at $3,000 apiece, and the backlog clears in a mere 180 years. Nobody in the program's literature performs this division. It is a good deal for the winners and a rounding error against the risk pool, and both things are true at once.
The household math, by contrast, is kinder. A typical retrofit costs $3,000 to $7,000, per CRMP's FAQ. CEA's Janiele Maffei puts the statewide average at $5,300, with bolt-only jobs near $3,000 and Bay Area jobs running $6,000 to $8,000 (Santa Monica Daily Press). So the $3,000 grant covers roughly 57 percent of an average job, essentially all of a bolt-only job, and barely half of a Bay Area one, which means the headline number is calibrated to the cheapest version of the work in the cheapest labor market where the program operates. Households earning $94,480 or less can get a supplemental grant covering up to 100 percent of the cost (CRMP FAQ). For a bolt-only house and a qualifying income, out-of-pocket cost rounds to zero, a sentence worth reading twice.
The Camera That Reads Your House
Here is where AI enters, and it enters sideways. Researchers working with the GEM Foundation have shown that deep-learning models can classify a building's seismic vulnerability from ordinary street-view photographs: number of storeys, structural system, whether it predates modern codes. The work was named an Editor's Choice in the August 2025 issue of Earthquake Spectra, the journal of the Earthquake Engineering Research Institute (GEM Foundation). A related effort calibrated such models on 3,000 photographed buildings in Lisbon (GEM webinar). Earlier work demonstrated neural-network soft-story detection directly from street imagery (MDPI review, 2026).
What once took surveyor teams months now takes software minutes. A city could screen its entire pre-1980 housing stock before lunch and know, to a first approximation, which blocks hold the riskiest houses, a capability that once required an army of clipboards and now requires an intern, a GPU, and a street-view API key.
California does not do this. EBB awards are first-come, first-served within the registration window, which means the grant goes to whoever heard about the program and filed fastest, and risk never enters the queue. A 1923 bungalow on liquefiable soil waits behind a 1978 ranch on bedrock because its owner saw the flyer first, and no part of the system records the fact that one of these houses is far more likely to harm its occupants than the other. Nobody defends this as principled; it is simply how the program was built, back when screening a million homes meant clipboards.
The uncomfortable part is that the technology to aim the money exists, is published and peer-reviewed, and sits largely unused on the triage side, even though building-level vulnerability estimates are exactly the kind of inference commercial risk modelers already sell. Vulnerability screening at city scale is a public good currently serving private underwriting, which raises an awkward question about who gets to buy the inference that could be aiming public retrofit dollars instead.
The Honest Objections
Now the case against everything above, stated at full strength. Start with the heaviest objection. Money is the binding constraint, not information. Perfect triage does not change the funding math; it reorders a queue that still spans centuries. A smarter waiting list is still a waiting list, and nobody should mistake queue management for a solution.
Worse, street-view AI has a blind spot aimed precisely at this program. Cripple walls live inside the crawl space, behind skirting boards and vent screens. A facade photo can estimate vintage and structural system, but the thing EBB pays for, bolts and cripple-wall plywood, is invisible from the curb, which puts a hard ceiling on what photo-based screening can promise no matter how good the models get. Models can flag the houses most likely to need the work, though they cannot say which ones actually do. Only a crawl-space visit can, which is why the program sensibly skips the technology altogether and lets FEMA-trained contractors give free estimates (CRMP FAQ).
Lastly, the retrofit is narrower than "earthquake-proof" suggests. It secures the house-to-foundation connection, but it does nothing for the chimney, the masonry missile that does disproportionate harm in garden-variety quakes, and nothing for houses on steep hillsides, which the program excludes. A bolted house can still become a very expensive pile of broken dishes. Anyone selling this as total protection is selling something.
What This Analysis Could Not Prove
Boundaries, honestly kept. The 1.2 million figure is CRMP's upper-bound estimate, not a surveyed count, so read the 505-year figure as an illustration of scale rather than a forecast. Funding varies year to year and the historical pace is not a promise. Cost figures mix vintages: Maffei's $5,300 average comes from a 2024 interview while CRMP's $3,000 to $7,000 range is current guidance, and your contractor's quote is the only number that matters, which is the polite way of saying statewide averages predict any specific crawl space poorly. AI screening results were validated on urban building stocks, Lisbon's among them, not on California's crawl-space bungalows specifically. I found no consumer-facing tool that scores an individual home's seismic vulnerability from a photo, and no evidence any California agency uses street-view screening to aim retrofit money. The 2026 registration window had not been announced at publication time; the January-to-March window cited is 2025's.
What to Do If Your House Predates 1980
Six moves, in this order, none of them requiring any new technology.
First, check whether your ZIP is among the eligible ones at EarthquakeBraceBolt.com, and put the registration window on your calendar, because it opens once a year and does not wait for you.
Second, get a free estimate from a contractor in the EBB directory. They carry FEMA training, and the estimate doubles as your qualification check, since the program currently requires no inspection to apply.
Third, do not pull a permit or start work before formal acceptance, since either act disqualifies you. This is the easiest $3,000 anyone ever lost.
Fourth, if your household income is $94,480 or less, apply for the supplemental grant, which can cover the entire cost, making this the rare program where the lowest-income applicants get the best deal instead of the worst paperwork.
Fifth, after the work is done, file Form BOE-64 with your county assessor. California excludes seismic retrofits from property-tax reassessment under Revenue and Taxation Code section 74.5, but the exclusion needs the form to attach.
Sixth, have the earthquake insurance conversation separately from the retrofit decision. Your homeowner's policy does not cover quake damage. A retrofit does not change that. Whether a standalone CEA policy pencils out after the retrofit discount is arithmetic worth doing with your agent.
Five centuries is not a plan. It is a confession that the current approach, however well-intentioned, cannot reach the houses it was built for, and confessions, however honest, do not keep houses on their foundations. The state funds thousands of retrofits a year and the technology exists to find the riskiest thousands first, which means the missing piece is neither money nor science but the administrative decision to connect them. Using one without the other is a choice, and it is worth asking, every January, who that choice serves.