@Damon_BPHR @myGrantUK Are you using the homely controls with it?

@jeremyphoward @dwarkesh_sp @ylecun @YejinChoinka Could extrapolation be trained with RL? The Deepseek paper shows that they teach CoT from examples. These examples might only show linear progression though a problem. Examples that jump between different disciplines, could teach the models implicitly search across fields?

@NicDenso @energystatsuk @EnergySysCat are running an air trial for the UK gov. I think this is to look at the relative savings Vs resistive heating. Grant changes will likely follow the result of the trial.

@NoPoke @_heatgeek I see it as making sure that poor installs still perform well and making the margin for error greater. I think competition comes up with creative solutions.

@_heatgeek One thing that would make many installs better would be better radiators that emit much more heat per unit area and are cheap. A competition for that would be good. Also a ranking board for heatpumps, which should undergo a secret performance test.

@DanNeidle See if they let you sign up.

@AravSrinivas Could you give the option for people to make their answers public to build a new Wikipedia type thing from the outputs? It seems a bit wasteful throwing away the answers every time the model is run.

@7Kiwi @ClemCowton @OctopusEnergy Why pursue uncertain fixes when we could simply reduce @TheCrownEstate leasing fees & revenue sharing on offshore wind to guarantee lower CfD clearing prices. Yes, that does mean part of your bill goes to King Charles. @energygovuk @Ed_Miliband

@half_tree @ProfRayWills Not if oil use for internal combustion engines goes down. They make 75% of the world's EVs. It makes sense for them to switch to electric as they have little oil.

@L__Bow @fionaharvey I think long term yes, but maybe not in the timescale for which their CfD runs out? I think that is March 2027.

@ShakeelHashim He repealed the AGI act. That's probably why?

@L__Bow @PeterEastern @TrystanLea @suburbanpirate @glynhudson @plumbers_urban Does it seem to be more fan or compressor related?

@PeterEastern @TrystanLea @suburbanpirate @glynhudson @plumbers_urban That is what I was going off, if the unit ices up then that dictates the heat that can be delivered. But I think different people are looking at it from different ways. Trystan's compressor power usage makes sense too.

@suburbanpirate @PeterEastern @TrystanLea @glynhudson @plumbers_urban It looks like you start to see the reduction of the output there too.

@PeterEastern @TrystanLea @suburbanpirate @glynhudson @plumbers_urban I think you can tell because the heat delivered over each defrost cycle decays. So if defrosts are regular, say every half hour, and the heat is dropping off rapidly in that half hour, you could assume that you have a peak heat.

@TrystanLea @PeterEastern @suburbanpirate @glynhudson @plumbers_urban Are the tests for output standardised? If so this should be fairly consistent across all brands. It might be a bit like emission tests for cars, in that they use to optimise for the test, rather than real world.

@theolodian @glynhudson @PeterEastern @suburbanpirate @TrystanLea A single a2a supporting one room, to drop the required output by a kW or so might make all the difference.

@plumbers_urban @glynhudson @NandoTolboom @tomasmcguinness @suburbanpirate @TrystanLea I think they do it according to the EN 12102-2 standard. If you can find the document then you can see how they do it.