Are We Sleepwalking Into Another Race to the Bottom?

Mars · 2025-10-04T10:35:00Z

When I first came across Heat Geek, I was always under the impression (based on their branding, positioning and messaging) that it stood on a platform of quality. Take your time. Do the calculations properly. Understand the property, the heat loss, the emitters. Design a system that works, that lasts, that delivers comfort and efficiency for decades. That philosophy cut through a sector that was plagued by bodged installs, incorrectly sized units and undersized radiators, leaving homeowners with cold rooms, high running costs and a sense of disillusionment. In many ways, Heat Geek helped professionalise an industry that desperately needed it. They raised expectations. They told us that quality mattered more than speed. That was my perception. Which is why their latest initiative, ZeroDisrupt, has left me scratching my head. According to their own launch announcement, this is an AI-powered design system that can make heat pump installs up to 60% cheaper, 50% faster and with far less disruption. To homeowners weary of quotes in the £12,000-£15,000 range (even after the Boiler Upgrade Scheme), it must sound like manna from heaven. Faster, cheaper, less disruptive... who wouldn’t want that? But my question is simple: have we not already seen how this story ends time and time again in the very recent past? We know exactly what happens when you throw heat pumps into homes with minimal radiator upgrades, with flow temperatures cranked up to compensate, with pipework that can’t cope and emitters that can’t deliver. It works beautifully on paper. The numbers look fine. The sales pitch is compelling. But the reality? Cold rooms, spiralling electricity bills, SCOPs that collapse under real-world use and homeowners who regret ever making the switch. That, in turn, fuels the anti-heat pump narrative and sets back public confidence. My concern with ZeroDisrupt is that “less disruption” often translates to no radiator upgrades, no system improvements and designs that lean on higher flow temperatures to make things “work”. That’s a guaranteed route to higher electricity bills. With tariffs unlikely to fall significantly any time soon, what looks like a cheaper install upfront can quickly become a financial haemorrhage... hundreds, even thousands of pounds a year lost to inefficiency. I don't see that as innovation. In my opinion, that’s shifting the cost burden from the installer to the homeowner. It's a creative way to move the bill from capex to opex. The real danger here is that the market is being seduced by volume (and perhaps by the expectations of shareholders). Octopus, British Gas and Aira are already throwing thousands of heat pumps into homes, often at a loss. Their installation arms are haemorrhaging tens of millions of pounds, but because Octopus and British Gas sell energy, they can claw it back later. So what happens next? To survive, the model must shift toward scale with more installs, done faster, for thinner margins. It becomes a numbers game. And we all know who loses in that scenario: the homeowner. They’re left with a “cheap” system that’s supposed to last twenty years but ends up inefficient, underperforming and ultimately a liability. I don’t claim to fully understand the workings of the AI modelling behind ZeroDisrupt. According to Heat Geek’s own video, it has been “trained” on thousands of installs. But no matter how clever the algorithm, AI cannot model the chaos of British housing stock. It cannot know what’s hidden behind the walls of a Victorian terrace, how poorly insulated an Edwardian semi might be or what decades-old cowboy pipework lies beneath floorboards. British homes are messy, idiosyncratic and unpredictable. There are too many variables. AI can assist a designer. It cannot replace one. Even among Heat Geek’s own elite installer network, unease is growing. Several have told me privately they’re uncomfortable with the direction of travel. They joined Heat Geek because of its founding ethos: slow down, design carefully, educate the homeowner, deliver systems that actually work as efficiently as they can. Now they feel the message has flipped on its head. SME installers are already operating on wafer-thin margins. And let’s not gloss over the language now being thrown around. “ZeroDisrupt might work on smaller homes,” I’ve been told by an installer. Might. That is not a word any homeowner wants to hear when committing thousands of pounds to the system that will heat their family home for decades. Might is not acceptable. It must be must. Heat pumps are not disposable tech, and they are fundamentally not cheap. They are the beating heart of a home’s heating system... something expected to perform reliably for 15 to 20 years. When the industry starts using words like might, maybe and probably, what it’s really saying is we’re not sure. And that’s truly terrifying. Experimenting on people's homes that are paying a lot of money for the privilege is not OK! This is where I fear the industry...

Page 6 / 6 Prev

Government Schemes, Regulations, Training, Planning & Organisations

Last Post by ksim 3 weeks ago

62 Posts

12 Users

22 Reactions

1,621 Views

RSS

SUNandAIR

(@sunandair)

Noble Member Member

Joined: 3 years ago

Posts: 514

17/12/2025 2:00 pm

Posted by: @ksim

↑

for example Gemini calculates that the flow should be fine, but still scared of microbore so recommends buffer tank.

That’s a very interesting exercise…. And it’s good to see some caveats in what was offered. I think Gemini is cautious of microbore for good reason since there are too many unknowns.

Apart from specific bore size…. How was it installed eg with kinks and tight bends? How many radiators are Daisy chained off a 10mm branch pipe? Is there any push-fit plastic microbore. To name a few.

We do have some microbore and it does work but we did do some remedial work to ensure the pipe sizes cascaded down so that only 1 radiator was fed from one branch of 10mm pipework. There is a limit as to how much heat energy a 10mm pipe can carry. The attached file shows how much energy can be carried by a given pipe size.

ps adding a buffer tank doesn’t solve the problem of under sized heat carrying pipes for a given load requirement.

This post was modified 3 weeks ago by SUNandAIR

ReplyQuote

ksim

(@ksim)

Active Member Member

Joined: 4 weeks ago

Posts: 8

17/12/2025 3:05 pm

Posted by: @sunandair

↑

Apart from specific bore size…. How was it installed eg with kinks and tight bends? How many radiators are Daisy chained off a 10mm branch pipe? Is there any push-fit plastic microbore. To name a few.

this is why I provided current flow temperatures in the prompt and dt at different pump speeds and asked to do calculation of the existing flow. LLMs are not "thinking", looks like Gemini training data is more incline to install buffer or separation for microbore. so even the flow is fine, it recommends a buffer.

I do not exclude any problems with microbore, absolutely possible to have an install where a buffer is needed, or repipe. I had installers from: "microbore won't work", "microbore will work but with only with lowloss header", to "everything is fine it will work if you keep it as open loop and weather comp", Let me show an example of my data and you can decide for yourself.

This is 5 hours of tonight, outside temps are 1-2C (external sensor). boiler minimum delivery is 7000kW, so every interval is boiler running at that power.

Room temperatures over the same time

7000W at delta T 5K, you can calculate the flow, there is a formula; similar to my heat loss/size/design temperature properties are running at ~12l/min according to heatpumpmonitor.org.

Running a bit further, I've asked Gemini to calculate the heat loss, gave property build/extension walls/roof insulation, windows, etc, no sizes except the whole house, it estimated 5.3-5.8kW, ChatGPT - 3.3kW, Heat Geek - 5.3kW. Next I provided annual gas consumption, ChatGPT run out of calculation tokens so I dropped it, Gemini adjusted to 2.0 - 3.0kW for the main house and 600W for the extension where Air2Air currently used. Next provided graphs from above and final recommendations were:

Please do not think I am trusting LLMs here, it is just interesting....

It would be interesting to feed to LLMs some real data which requires separation, if I increase flow temp from my boiler in the prompt they immediately starts to have doubts that ASHP will work in the property and recommends to keep gas.

As an idea, load a design document from the installer to LLM with all the numbers and ask to analyse it for possible questions to the installer and problems system might have.

This post was modified 3 weeks ago 3 times by ksim

I have a magic boiler

ReplyQuote

Page 6 / 6 Prev