She Built Your Climate Dashboard. She Has Notes.

This interview takes place in 2043. Kit Tamura-Bosch, her tool, and this conversation are speculative fiction — which, for a piece about the gap between what a screen shows you and what's actually happening, feels appropriate.

Before she built Canopy, the household-translation app that now aggregates utility demand-response alerts, NWS HeatRisk scores, AQI feeds, municipal water-restriction stages, and school schedule modifications into a single daily interface, Kit Tamura-Bosch spent six years designing claims portals for a health insurance company in Columbus, Ohio. She is, by her own description, "professionally experienced at making broken systems look like they work."

She moved to Phoenix in 2031. In the summer of 2037, her neighbor's air conditioning failed during a HeatRisk 4 week. The neighbor checked Google, found a cooling center listed as "open," drove fourteen minutes, and discovered it didn't accept unaccompanied adults with service animals.¹ Kit started building Canopy that fall. By version three, she had removed the cooling center field entirely.

We spoke over video on a Tuesday in late June. Her own app was showing an AQI of 178, a HeatRisk 3, a demand-response event from 2 to 7 PM, Stage 2 water restrictions, and a school schedule modification she described as "the word modified doing a lot of heavy lifting."

What does Canopy actually do, in one sentence?

Kit: It takes five institutional feeds that were never designed to talk to each other and puts them on one screen. That's the whole product. People call it a "decision support tool" or a "household resilience platform," and I let them, because those phrases get you grant funding. But what I actually built is a thing that reduces the number of apps you check in the morning from seven to one. The achievement is consolidation. I'm a very expensive bookmark folder.

You used to build health insurance claims portals. How different is this?

Kit: Barely. The data is different, the stakes feel different, but the core design problem is the same. You have a system that operates in institutional units — CPT codes, allowed amounts, provider networks — and a person standing in front of it who needs to know one thing: can I go to this doctor on Thursday. My job was to make the portal feel like it answered that question. It didn't, really. It showed you the institution's answer in the institution's language and let you believe you'd been told something actionable.

Canopy works the same way. HeatRisk gives you a number for a 24-hour period.² The household needs to know whether 4 PM is safe for a specific body. AQI gives you a single number that masks which pollutant is driving it, and whether closing your windows helps depends entirely on whether it's ozone or PM2.5 and whether your building has functional filtration.³ I can show you the number. I can't show you your building.

You removed the cooling center feature in version three. That seems like a strange thing to take out of a heat-safety tool.

Kit: Yeah. So in version two, we pulled the county's cooling center feed. Hours, locations, open/closed status. Green checkmark, red X. Very clean. Very satisfying interface element. And almost immediately we started getting feedback from users who'd driven to a center that was "open" — green checkmark, we told them it was open — and found out it didn't have weekend transit access, or didn't accept pets, or had a check-in process that required ID they didn't have, or was running its HVAC at reduced capacity because of a demand-response event happening simultaneously.

In Maricopa County, survey data showed 39% of people said they'd want to bring a pet or service animal to a cooling center. Seventy-six percent said they'd drive themselves.⁴ We had no data field for any of that. We had "open." And "open" was true. It just wasn't the thing you needed to know.

So I pulled it. People were furious. They said, you took away information.

I still don't know if removing it was right. Someone might have gone to a center they wouldn't have found otherwise. But I couldn't keep showing "open" when I knew what "open" didn't include.

Your app sometimes shows a demand-response event — "reduce cooling" — on the same day it shows an elevated HeatRisk score. What's the user supposed to do with that?

Kit: Both things. Neither thing. I genuinely don't have an answer. The tool shows both because both are true. The utility is asking you to reduce AC load between 2 and 7 PM. The NWS is telling you heat risk is elevated for vulnerable populations. If you override the demand-response event, your smart thermostat stays on, but override too many times and you get disenrolled from the program and lose the credit that was offsetting part of your rate increase.⁵ If you don't override, your house gets warmer during the exact hours the weather service says are dangerous.

My tool can display that conflict. It cannot resolve it. Resolving it would require knowing which bodies are in the house, which medical conditions apply, whether the AC unit actually works. Among indoor heat deaths where AC was present, the unit wasn't functioning in 72% of cases.⁶ You'd also need to know what the household can afford to lose financially. I don't have that information. I shouldn't have that information. But without it, I'm showing you a problem and calling it a service.

Everything in your app updates on a different schedule.

Kit: This is the thing that makes me close my laptop and go lie on the floor. At 7 AM, when a parent is deciding what the day looks like, she's looking at: a HeatRisk score from last night's model run, an AQI forecast from yesterday afternoon, no school notice yet because the district decides at 6:30 and the API lag is forty minutes, no demand-response event because the utility declares those at 9, and a water-restriction stage that was set five weeks ago.

She's assembling a jigsaw puzzle from pieces that were cut on different days. I can put them on one screen. I cannot make them simultaneous. The screen implies coherence. The data doesn't have any.

Who actually uses Canopy?

Kit: Households with smart thermostats, stable internet, and enough bandwidth — cognitive and literal — to install another app. So, roughly 3% of the population that would benefit from it.⁷

The demand-response feed requires enrollment, which requires a smart thermostat, which requires either owning your home or having a landlord who installed one. A renter in an older building with a landlord-controlled thermostat gets nothing from my tool regardless of how elegant the interface is. I could redesign it every week for a year and it wouldn't matter.

I see it in my user demographics. My median user has newer appliances, a stable lease or a mortgage, and enough financial margin to act on what the app shows them. The household that most needs hour-level heat guidance and AQI breakdowns is the one least likely to have my app, or a smart thermostat, or a working AC unit, or a car to get to the cooling center I stopped listing. I built a tool for the people who needed it second-most.

What does the tool actually do well?

Kit: It surfaces timing conflicts you'd miss otherwise. If a demand-response window overlaps with school pickup, you now know before noon that you need to choose between running errands now or waiting until 7 PM. That's real. People shift laundry loads, check AQI before outdoor plans, rearrange their afternoons. Small stuff. But the friction reduction is genuine, even if the underlying systems are incoherent. I'll take genuine and small.

Should you have built it?

Kit: [Long pause.] I don't know. I built a better window into a building that's on fire. The window is real. The fire is also real. And some mornings I open the app and look at my own dashboard — five feeds, five different update cycles, three conflicting recommendations — and I think, this is useful. And then I think, useful for what.

Footnotes

1. In Maricopa County's 2023 community survey, 39% of respondents said they would want to bring a pet or service animal to a cooling center, and 76% said they would drive themselves. Maricopa County Department of Public Health, 2023. https://www.maricopa.gov/ArchiveCenter/ViewFile/Item/6510 ↩

2. NWS HeatRisk calculates a daily value for each location using both high and low temperatures compared to historical values, producing a single 0–4 score for a 24-hour period. https://www.wpc.ncep.noaa.gov/heatrisk/overview.html ↩

3. The AQI reports a single number based on the pollutant with the highest concentration. It does not distinguish between ozone and PM2.5, which require different indoor-air responses. https://www.cdc.gov/heat-health/hcp/clinical-guidance/how-to-use-the-heatrisk-tool-and-air-quality-index.html ↩

4. Maricopa County Department of Public Health, 2023 general community survey on cooling center access. https://www.maricopa.gov/ArchiveCenter/ViewFile/Item/6510 ↩

5. Demand-response program terms typically allow a limited number of overrides before automatic disenrollment. https://www.dteenergy.com/us/en/residential/save-money-energy/get-started-with/demand-response.html ↩

6. Maricopa County's 2025 heat-related deaths report found that among indoor deaths where AC was present, the unit was not functioning in 72% of cases. https://www.maricopa.gov/ArchiveCenter/ViewFile/Item/6510 ↩

7. Smart thermostat adoption reached approximately 16% of households with internet access, with only about 20% of those participating in demand-response programs. https://www.utilitydive.com/news/demand-response-dr-utility-programs-resideo/754205/ ↩