Your Energy Code Demands an Airtight House. Your Fire Code Assumes a Drafty One. A Waterloo Lab Burned Fifteen Homes to Measure What Falls Through the Gap.
Joshua Pulsipher set fifteen fires in a building on the University of Waterloo campus and watched each one through 175 sensors sampling four times per second: temperature, airflow, humidity, burn rate, gas composition, all of it streaming into a dataset so large that traditional analysis could not extract the pattern buried inside millions of data points per experiment. Pulsipher, a chemical engineer working alongside fire researchers Beth Weckman and Vinny Gupta in Waterloo's Fire Research Group, was not looking for a way to fight fires faster. He was looking for the moment a fire changes its mind.
That moment has a name. It is the transition from well-ventilated combustion to under-ventilated combustion, the point where a fire has consumed enough oxygen in a sealed space that its chemistry shifts, incomplete burning begins to dominate, and the gases it produces become profoundly more toxic. In a leaky house, that transition may never arrive because air infiltrates through gaps around windows, soffits, rim joists, and recessed lights, feeding the fire a continuous oxygen supply. In a house built to current energy code, sealed to 3 ACH50 and wrapped in a continuous air barrier exactly the way your energy auditor told you to do it, the transition arrives sooner, and nobody watching the smoke from outside can tell.
Pulsipher's team published their framework in Fire Safety Journal in August 2026, combining statistical methods, data science, and machine learning to identify this ventilation transition in real time from sensor data. It works. Detect the shift, and you can predict what the fire will do next, route evacuations, adjust suppression tactics, save lives. But the paper's implications extend beyond fire response into a question that two separate code-writing bodies have not addressed together: what happens when the envelope your energy code requires creates fire conditions your fire code was never written to handle?
Two Codes, One House, Zero Coordination
IECC energy provisions and NFPA/IRC fire provisions are developed by separate technical committees on separate timelines with separate stakeholder constituencies. Energy code committees focus on envelope performance, mechanical ventilation sizing, and thermal bridging, and their metric of success is reduced heating and cooling load. Fire code committees focus on egress width, detection placement, suppression requirements, and fire separation, and their metric is occupant survival during an event that lasts minutes.
Neither committee references the other's work in a way that accounts for the interaction between airtightness and fire behavior. A home can pass both the IECC 2021 airtightness requirement of 3 ACH50 in climate zones 3 through 8 and every applicable fire provision in the IRC, while exhibiting fire dynamics that neither code was designed to predict. Passive House construction, at 0.6 ACH50, pushes the envelope tighter still, and every code cycle since 2009 has demanded lower air leakage rates than the last, a steady ratchet with no corresponding adjustment on the fire side of the ledger.
What Burns Differently Now
Airtightness is only half the problem. UL's Firefighter Safety Research Institute ran controlled burns comparing legacy rooms furnished with solid wood furniture, cotton upholstery, and cotton batting against modern rooms furnished with synthetic fabrics, polyurethane foam, particle board, and laminates. Legacy rooms reached flashover in 29 minutes and 30 seconds. Modern rooms? Three minutes and 40 seconds. That ratio, roughly eight to one, means a fire in a new house filled with new furniture reaches unsurvivable conditions before any fire department in the country can arrive, given that urban response time objectives run 9 to 14 minutes.
Put those two trends together and the arithmetic is grim: tighter envelopes accelerate the ventilation transition that produces more toxic gases while synthetic furnishings accelerate the heat release that produces flashover. Forty years ago a smoke alarm gave you an average of 17 minutes to escape, according to UL FSRI's 2017 analysis. Now you get three minutes or less, and the gases you are breathing during those three minutes are chemically different from what fire codes assumed when they set smoke alarm placement rules.
NIST full-scale armchair tests quantify one available countermeasure. Without barrier fabrics, an upholstered chair reaches a peak heat release rate near 3 megawatts and drives flashover in 2 to 3 minutes; with barrier fabrics, peak heat release drops by a factor of three and flashover delays to 10 to 20 minutes. NIST Technical Note 1920 found that barrier fabrics reduce the probability of lethal conditions in adjacent rooms by at least 50 percent. CPSC's 2007 open-flame mattress standard, which effectively mandated barrier fabrics in mattresses, reduced bed-fire deaths by approximately 70 percent over the following decade.
No equivalent standard exists for upholstered furniture sold in the United States. None is pending.
808,000 New Reasons Per Year
U.S. Census Bureau data for July 2026 show 808,000 single-family housing starts at a seasonally adjusted annual rate. Every one of those homes will be built to an airtightness standard tighter than what fire codes were calibrated against, furnished with synthetic materials that burn eight times faster than their predecessors, and protected by a fire detection and response framework that assumes conditions from a generation ago.
IRC Section R313, adopted in 2009, requires residential fire sprinklers in new construction. Most jurisdictions have repealed or exempted the requirement through local amendments, leaving ionization or photoelectric smoke alarms, devices calibrated for well-ventilated fires, as the primary warning system in homes where under-ventilated fires are becoming the default condition. A closed bedroom door, per UL FSRI research, can hold temperatures to 100°F while the hallway outside exceeds 1,000°F. Close your door is a public education campaign. It is not a code requirement.
Where the Counterargument Stands
Residential fire deaths have declined from over 6,000 annually in 1977 to approximately 2,700 in recent years, a trend driven by smoke alarms, improved fire separation, better electrical wiring standards, and reduced smoking rates. Most fires still occur in older, leakier homes with outdated systems, not in Passive House builds. No epidemiological study has isolated excess fire risk attributable specifically to post-2012 airtightness levels, and the Waterloo study used a controlled campus burn house, not occupied residences. Every piece of evidence connecting tighter envelopes to worse fire outcomes is mechanistic and experimental, not observational. Nobody has counted the bodies that this code gap is supposed to produce, because they may not exist yet, or because the data infrastructure to detect them across 3,143 counties does not exist either.
Both of those explanations deserve investigation, and neither has received funding.
What Pulsipher Built and What It Cannot Do
Waterloo's AI system can detect the ventilation transition in sensor data from a burning structure. It cannot retrofit a building code process that takes a decade per cycle to acknowledge a problem that compounds with every new home permitted. It cannot force the IECC energy committee and the NFPA fire committee into the same room to reconcile requirements that interact physically but not institutionally. And it cannot tell the homeowner who just achieved a blower-door result of 1.8 ACH50 and celebrated with their energy rater that the same sealed envelope changes the survival math if a candle tips over at 2 a.m.
What Pulsipher built is a measurement, a precise, sensor-verified, machine-learning-validated measurement of a phenomenon that building science has discussed theoretically for years: airtight homes burn differently. Now there is data. Whether anyone in a code-writing position reads the paper before the next cycle closes is a question the AI cannot answer.
Limitations: Waterloo's fifteen controlled burns were conducted in a campus burn house, not in occupied residences with diverse layouts and contents. No published data quantify the specific fire-risk delta between homes at 3 ACH50 versus 7 ACH50 versus pre-code legacy construction. UL FSRI's legacy-versus-modern comparison measured furnishing material differences, not envelope airtightness effects specifically; isolating the contribution of each variable to fire outcomes has not been attempted in published research. NIST barrier-fabric data address furniture, not the interaction between furniture fire behavior and building envelope airtightness. No data exist on whether residential mechanical ventilation systems, required in airtight homes, affect fire behavior when running or when the system shuts off during a fire event. Cost data for residential sprinkler systems in new construction ($1 to $2 per square foot is commonly cited) vary by project and jurisdiction.
Sources: Mazzadi, S., et al., "A framework for high-dimensional fire sensor data analysis," Fire Safety Journal, 2026 (DOI: 10.1016/j.firesaf.2026.104738); UL Firefighter Safety Research Institute, controlled burn comparison studies and "Close Your Door" campaign data, 2017; NIST Technical Note 1920, barrier fabric full-scale fire tests; CPSC 2007 open-flame mattress flammability standard (16 CFR Part 1633) impact assessment; IECC 2021, Sections R402.4.1.2 and Table R402.4.1.1; IRC 2021, Section R313 (residential fire sprinklers); NFPA fire death statistics, 1977–2024; U.S. Census Bureau, New Residential Construction, July 2026 (SAAR); University of Waterloo Fire Research Group, news release, August 2026.