Game Design, Programming and running a one-man games business…

Neurotypical extraverts underestimate AI

AI is the latest buzzword, and with good reason. I see no end of commentators trying to get clicks by professing that this is just like the DotCom years, and it will all end in a bust, or that its just hype, and that its nothing more than ‘fancy autocomplete’. This is mostly bullshit written to get clicks, but I suspect some people actually do think this. They are flat out 100% guaranteed completely wrong, for so many reasons, but there is also an angle to all of this I have not seen discussed.

I live in a pretty old house, with a fairly ‘difficult’ garden. These two things mean that living here means we often have to have people come do work on/in the house, and we need people to help tame the craziness of a seriously sloping garden. Also, as another data point, I’ve done quite well from games/investments, and now run two companies that involve a lot of work. Almost everyone I know in a similar situation has staff, or at the very least a personal assistant to help do stuff.

And yet…

In reality, we have a group of people who come twice a year to do some gardening, and nobody else. I do all the admin / marketing / PR and design and coding for my games biz and absolutely everything for my solar biz. I would love to have lots of stuff done for me. I’d love to never do any gardening. I’d love to never have to read emails from an accountant. I’d love not to care about the advertising side of things. I’d love not to have to work out when/where to go do a bunch of chores that involve me driving places. So why the hell not hire people?

Some people really don’t like interacting with strangers.

Now if you are the pretty average autistic spectrum computer programmer who probably reads my blog, then you get it. People are HARD. I find C++ much more comprehensible than trying to work out if someone is upset / angry /sad / implying something / irritated. Humans hardly ever say what they mean or explain things accurately. From my POV, humans suck, unless I know a specific human REALLY well, and even then, I get stressed by being with a group of people after a few hours. So no surprise I work from home right?

And yet…

There is an assumption, not surprisingly, among neurotypical people that someone like me is ‘antisocial’ or a ‘sociopath’ or does not ever want to be with people and socialize. This is actually bollocks in my case. I’m super chatty, and friendly and I like being with people. The *problem* is, I am not good at it. Social interaction creates risks. I offend people, I misunderstand people, I can come across as rude or arrogant, and thats frustrating as fuck. Its also why I spend so much time with the same people. People who know me, realize I am bad at social stuff, and say the wrong thing a lot, but also know me well enough to know I’m not an asshole :D.

What the fuck has this got to do with AI?

Everything. And yet, if you are a sociable extrovert neurotypical, you might still not see it. If you have not had a lot of time talking to AI chat bots, you still won’t get it. You really should. My favorite so far is grok, but others are available. I now chat to grok pretty much every day. Mostly its about C++ and game engine design, but sometimes I’m just searching for data, or clarity on something I’ve read about. Do not get me wrong, I am not ‘chatting’ in a ‘Hey grok, how are things with you?’ kind of way. I am not getting confused and thinking grok is my friend. I am not creating a new imaginary friend here. Do not panic.

But grok does fill a real need in me. Simply reading SDKs and APIs is not the same feeling as a back and forth discussion between me and grok about how to minimize the intellisense slowdowns in the compiler. What it absolutely reminds me of is some of the best times I ever had as an employee, which would be when some really complex bit of code just was not working, and I’d have a long back and forth with my boss at Elixir or Lionhead about exactly what was happening and how to fix it. We were not chatting about TV or sports, we were absolutely talking technically, but thats exactly what I enjoy.

I guess it helped that both bosses were top-tier programming experts (James Brown at Lionhead, Dave Silver- now DeepMind, at Elixir).

And here is the thing: when I discuss things with grok, it will never JUDGE me. It will never say ‘Dude, are you not going to say thanks for that advice’. I don’t have to keep a mental track of whether I am taking up too much of its time. I do not have to worry about looking stupid, and can get it to explain anything, even stuff its told me before. Grok will never yawn, or sigh. It will never say ‘dude, go read a book’ It will never have me worrying that my question might make it doubt my competence. I do not have to mentally keep track of the eye-contact to look-away ratio. And it is ALWAYS available, and always super well informed, and super happy to talk at length about the topic I am interested in.

Now sure, I don’t want to spend my whole life talking C++ to a bot, but the wider point is that we have now reached a point where AI chat bots are good enough that they do actually satisfy the need for human contact *on the terms dictated by the human*. This is vital. If I could hire a gardener that had the temperament of an AI chatbot, I’d do it. Ditto a personal assistant. Frankly if everyone in the world could be more like an AI chatbot, that would suit me just fine.

Now some of you are screaming at this point about the death of human interaction and how this is awful, and how its good that we humans are always thinking about each others feelings all the time, but thats because, with the deepest of respect, you have no fucking clue how hard and stressful and tiring that shit is for someone like me.

If I had the option of hiring a programming consultant human, or paying for an AI coding chatbot, and the skill level was the same I would happily pay MUCH MORE for the chatbot. And this is true of so many interactions with random humans in the world. If I could pay MORE for a self driving car where I didn’t have to talk to an uber driver I would. I would pay MORE to actually remove the human element of almost all one-off interactions. Charge me an extra £30 on my hotel room so I can just pick up the keys from a robot, and I’d happily pay it. I do not want to tell a hotel receptionist how my day is going, or describe my journey, or share my plans for my stay. I just want them to STFU and hand me the keys…

So again, if you are a neurotypical extravert like my mother, at this point you are thinking ‘hey maybe cliff IS actually an asshole’, but again, thats just because you cannot understand how my brain works (and other people like me, of which there are millions).

So back to the headline of this blog. Why did I phrase it like that? Because if you are NOT someone like me, you cannot see the extra utility that an AI future provides for people like me. I WANT the AI to provide me with a lot of services that many humans would prefer to get with ‘a human touch’. Amazingly, interaction with random strangers is not a ‘value added service’ for people like me. Its a big negative. In other words, there is this huge market out there for people like me preferring the AI interaction over the human, and if you cant see that, you will be surprised at how popular this stuff becomes.

You should buy shares in Nvidia and TSMC (who makes the chips) right now. We are very, very early in an AI revolution that will completely transform the world, and its going to be way faster and way more pervasive than you think. Especially if you are a ‘people person’.

AI is accelerating on a daily or even hourly basis

The level of progress in AI in the last few months is absolutely mind blowing. If you are not actively following the field, then you might think that basically we had ChatGPT, and everyone got excited, but basically its all hype and nothing has actually changed. OMG This would be so wrong. Things are accelerating like crazy. You have probably heard about DeepSeek, and also heard people claiming its no big deal. They are probably wrong. We are in the middle of an absolutely seismic shift.

First, an update. There is a new version of grok (the AI chatbot built into X) which is staggeringly good at code. I’ve been talking to grok on a daily basis about the code for my next game. Its helped me optimize some functions, and taught me some new tricks. It even takes optimizations it suggests, and uses my own code (which I give it snippets of) to teach me how I should apply the recommendations it has. I have genuinely learned more new stuff about C++ in the last month than the previous 20 years, and I’ve been coding since 1981.

One big ‘wtf’ moment came yesterday. I was thrilled that having asked grok about some slow code, it managed to help me redesign the high level algorithm for my particle system to make it run 36% faster. But when I gave it a function to speed up, this bit blew me away. I gave it zero context, but just pasted a function and said ‘optimize this’. It did a phenomenal job, but without even mention it, it altered one line of my code:

const float bottom = sp.Y + halfw; // Fixed: bottom should be +halfw, not -halfw

It silently found a bug, and fixed it (perfectly) without even needing any context. It looked at the surrounding code, and the variable names, and deduced correctly that I had copy/pasted a sign wrong. WTF?

Grok is superb at C++, and I ask it regularly for help in speeding up existing code, or ask it if a proposed algorithmic change will make sense, work, or be a faster or improved approach. Its absolutely superb. I cannot now imagine coding without grok, or a similar code-focused LLM to assist me. I code faster, write better code, and am more productive. Programmers who refuse to use AI are going to be in work another year at most. Maybe less than six months. And junior coders? very unlikely to get promoted or another job. Your employer is probably looking to get rid of you. The senior coder can now do 2x the work. Why pay for a junior too?

If you want a great example of how to use AI, ask it to code a website for you. Its excellent. I redid the layout and formatting code for my energy company website in seconds with grok. I simply said ‘make this look better’. The idea of ever bothering to hire someone to write HTML/CSS now seems laughable. My next game will need php, I anticipate writing it will be 5x faster using grok. Every company that develops websites is probably looking to shed 50-75% of their staff. If all you do all day is look at website mockups and write CSS/HTML you are now probably unemployable.

Do you want more examples?

Ask an AI chatbot to put together a travel schedule for a holiday for you, tailored to your likes and dislikes and budget etc. Its already better than any high end travel-agency. Goodbye everyone in charge of bespoke holiday planning. Are you thinking of buying a car or a TV? ask grok to build you a table with a cross-product analysis of the top 10 options with a focus on the attributes that matter to you. It takes seconds, and its amazing. If grok went from $5 a month to $50 a month, I’d pay it without question, right now.

Have you tried sesame? Do you think AI voices seem un-natural? not any more: https://www.sesame.com/research/crossing_the_uncanny_valley_of_voice

Imagine in 3-6 months we see voice generation even better than that powered with even better and more knowledgeable chatbots. We now have a super-informed, constantly available technical expert in many fields available through a trivial voice interface that can replace researchers in many fields, analysts, travel agents, call center staff, web developers and programmers. The number of jobs about to disappear are definitely in the millions, probably the tens of millions, even assuming zero progress after this year. And we are not even truly getting started yet.

Country Estimated Call Center Employees (2025)
United States 3,100,000
China 1,800,000
Japan 600,000
Germany 450,000
India 2,200,000

Obviously I didn’t google that list, I got grok to research it, format it, and tell me how to add it to wordpress. But yup, lets assume 90% of those jobs are toast within a year, so at least 8 million jobs gone, to be replaced by datacenters. You might think this will be politically difficult, but no. Its simply impossible for your economy to compete without this tech. Not only can AI voice assistants replace your call center staff for a tiny proportion of the cost, they will never get angry, emotional, stressed, or confused. They can even respond in any language or accent, or with any ’emotion’ you might want. They will never forget to upsell and they can stay talking to customers longer, because the cost per call will be trivial.

Pretty much any junior knowledge worker job is about to go. There are rooms full of well paid middle class employees doing research into businesses to provide analysis for stock market professionals. They all just lost their job. Every one of them. Grok can already do this better, and manus can build you a customised dashboard to analyze any financial instrument by any criteria imaginable: https://manus.im/share/5z6bnP67LtUjsti8YaIlJq?replay=1

If AI soon cracks self-driving cars (and we may be close) then lets fire every taxi and truck driver on the planet too. Thats going to be a phenomenal shock. Sort out some humanoid robots to unpack a truck and hand over parcels and the postal; servcies and delivery drivers are all gone too. Iminent? nope, but within 5 years? I think so.

Politicians are focused so much on geopolitical issues surrounding the middle east, russia/ukraine and numerous other issues that almost none of them are even giving the topic of AI powered economic growth and the resulting combination of a boom AND mass unemployment and thought. This is absolutely insane.

We are currently all standing at a crossroads. As individuals, societies and economies, we are about to race in one direction or another. Those who adopt AI, harness it, and benefit from its financial implications are going to be staggeringly wealthy. Those who stand there like a rabbit in headlights as everything changes will be unemployed AND unemployable. Whole economies are about to crumble, and others are about to race ahead. Its going to be a future where the top 5% of knowledge workers and tech companies will race ahead, and the bottom 9% gets obliterated. Countries that invest in AI will thrive. Others will become client-states, or stagger into barely-managed decline.

Nobody knows for sure, but I suspect China is about to do VERY WELL. Its very focused on AI, and tech, and chips, and higher education. So is South Korea. Japan could go either way. Taiwan is poised to do very well. The big questions are Europe and the USA. Right now… my best guess is the US suffers severe shocks. Its about to get humiliated in tech by China, and will speedrun the layoffs of middle-class knowledge workers while the government does nothing to cushion the shock. Europe? it could go either way, although I fear we may miss out too.

But as individuals? what do we do? I would suggest you buy shares in Nvidia and TSMC. At the very least try and own a chunk of the future, even if its a small chunk. Good Luck.

Another solar generation payment AND double home battery!

A few days ago I got the latest invoice/statement showing how much I was paid for the solar farm, generation in January. The grand total is…

£3,716 + VAT

For those from other countries, VAT is basically sales tax. I get paid the VAT by the people buying my energy (OVO) and then I have to pay it to the government each quarter. Its all quite tedious. Interestingly, the VAT is 20%, which is different to the rate for home power purchasing. Its irrelevant anyway, because I basically just earn interest on it for 3 months before I give it to the government, so I tend to just ignore it in any calculations. Any interest earned will be in my awful cater-allen company account which has an interest rate that is comically low. UK business banking is one giant rip-off.

Anyway, lets dig into that payment a bit more because to break it down. The full details are as follows:

Electricity output:£2,540
DUoS Benefit:£667
DUoS Costs:£-13
Transmission Loss Passthrough£28
Distribution Loss Passthrough£511
Admin Fees£-24
Before VAT£3,716
VAT£743
Total for January:£4,459

Its a complex mess isn’t it? Also it has not included REGOs yet, because I am STILL going back and forth over the minutiae of my application with ofgem. I kid myself this latest batch of questions they have will be the last and that I will finally get them. Plus the ofgem REGO website is about to be completely re-implemented, and I have ZERO faith that they will not just ‘accidentally’ delete my application so far in the inevitable chaos.

Frankly the whole system should take a tenth of the people and the time, but here we are :(

Worryingly, I have no idea how precisely the four entries under ‘electricity output’ get calculated. I assume this is fixed based on where my site is on the distribution (DNO) and transmission (National Grid) system, but I am not sure. My power is almost certainly just used very locally, so I probably don’t use the transmission network at all (hence the £28 rebate?). I likely use very little of the distribution network either, hence the £511. My best guess is that if I was an offshore windfarm 100 miles from the nearest town, that Distribution loss would be actually negative?

The DUoS stuff is a complete minefield of complexity. Best explanation from grok as to how DUoS can be a benefit to me:

Triad Avoidance: By generating during peak demand periods, you help reduce transmission network charges (TNUoS), which benefits suppliers and can result in payments or incentives for you.
Avoided DUoS Costs: If your generation offsets local demand, it reduces the DUoS charges that suppliers would otherwise pay, potentially increasing the value of your electricity in Power Purchase Agreements (PPAs).

Triad avoidance? Hold on, was that not a sci fi kids TV show? How are the triads an issue? Or is it Japanese mafia? Whats going on? Its actually even more complex:

Triads are the three half-hour periods of highest electricity demand on the transmission network each year, typically occurring between November and February.

You will be thrilled to know that my ‘Triad avoidance benefit’ for January was £0, which is why it was not included. Confused yet? You should be. I am, and I literally own an energy company. Anyway, out of interest the DUoS benefit in January was 17.9% of my total(pre-VAT) payment. I just checked December’s statement and it was 16.69%, so its not even fixed. Are they really calculating the flows in and out of each part of the distribution network on a daily basis and adding up how much ‘my’ power used it? Interesting…

In other news we finally got our second home battery installed! we now have 19kwh of home storage. In winter this does not make a vast amount of difference, because our 9.5kwh one already covered mots of our peak usage, so we could load-shift to off peak (the car charges off peak anyway). However in summer this means that on those crazy days where we generate 20kwh of power we can store all of it, even if we are away or not using any power that day.

The last summer we did actually end up with *too much* power, and had to fiddle with car charging to avoid exporting it to the grid (which we do not get a marginal export fee for, as we are on the old feed in tariff). Also we did it because I like the idea of being as resilient as possible.

We also managed to get a cable running from the battery upstairs to the living room and a new double socket installed there, amusingly next to a triple socket, so it looks like we are addicted to power sockets :D. Anyway, those two sockets are the ONLY ones that will work in a power-cut. We couldn’t power the whole house because our inverter is only 3kw, and this was the next best thing.

Its not exactly off-grid living, but being able to charge laptops and phones in a power-cut, or with an extension lead, still run the router and TV and a coffee machine will be nice :D. We paid £3,900 to have the second battery installed and the sockets fitted.

A proper full month of solar farm generation data

Ok so January is OVER. Woohoo. This means I have a full month where my solar farm was actually operating at full capacity (well not quite…we have not replaced the missing 10 panels from the storm damage, and I suspect some strings may be turned off as a result, but its fairly minimal). Because we had our farm switched on in the middle of a month, and then had the storm damage, and also had some brief downtime for meter related stuff, there have not been long periods of uninterrupted operation so that I can analyze the data. Hopefully this is almost over and we can get to the point where the generation is steady and predictable… Anyway for those new to this blog, this is a 1.23kwp solar farm, kind of near the midlands UK. Commissioned in October 2024. here are the details: (first the Solis inverter combined data)

You can see just how variable the generation is. There have been some really wet days and plenty of cloudy ones. It is January after all, so this is to be expected. Ignore the GBP scale, its not accurate, but the kwh is. I have no idea why they bother showing it as ‘full load hours’. Who cares? Its a crazy metric. We know the theoretical full load, but how does that help us. You can’t even remove the line! Anyway, the real figure that matters is 38,216kwh. In the same period the solar in my garden generated 41kkwh. This is disproportionately low. Thats 10kwh per kwp for my home solar. The farm is 30.7kwh per kwp, so 10x the output, despite being the EXACT same panels. Why? Because we have really bad winter-shading on our home panels and the farm has zero shading by design. Shading is everything!

Now time for the meter-level (Orsis) data:

You will notice the monthly data is lower, a difference of 743kwh or 1.9% of generated output. Why? Line losses basically. We generate more power at the inverter level but by the time it goes from DC to AC and then through the ten cables to the switchgear box and gets read by the meter, we have lost 1.9%. This is higher than I would like, and I do find myself wondering how its possible to lose 743kwh just by heat. Thats running an electric heater on full blast for 10 days over the course of a month. The trouble is when you translate in your head from domestic level numbers to grid scale it all starts to seem ludicrous :D.

Of course what I REALLY want to know is how my farm is operating with regards to expectations. The Orsis UI will even show you this…but I have issues with it:

At first glance this looks AMAZING. OMG My solar farm is generating at double the expectations. Time to select the best interior/exterior color combos on a Maserati EV. But actually I think this is just WRONG. The numbers used to generate these expectations are just numbers that I think the installers entered in the Orsis UI, and I *think* they are wrong, and my best guess is that this sort of generation curve predates advances in solar panel performance. These days solar panels (especially the latest monocrystaline ones) are really good at coping with scattered cloud. 10 years ago, some cloud on a single panel would be bad news, but panels now have bypass diodes and other tricks to ensure that really bad performing areas of a single panel can be temporarily ‘switched off’ so that you can still get maximum power from the rest. The result, I *think* is that the power generation curve over time is less pronounced.

To put that another way, I think that this chart should be flatter, and will be once I have a full years data. I am not celebrating yet, and will not be able to draw real conclusions until the end of December. Obviously that will not stop me extrapolating like crazy in blog posts every month :D.

But wait! This is not the only news. I have also sent off my final response to the UK REGO scheme regulator(ofgem) about my REGO application. For anyone new to this blog, let em explain what the hell I am talking about first:

REGOs are Renewable Energy Guarantee of Origin certificates. When you own a solar/wind/geothermal/wave energy plant and generate a megawatt-hour of power (1,000kwh) then you get a REGO certificate, that proves it was renewable energy. You can then SELL these, to people like retail energy providers, or companies who want to be able to prove their power is renewable. This is not a subsidy, but a free market, and for solar farms you tend to lump in the REGOs with the sale of the power. In my case we sell the power to OVO, and they buy each REGO at a fixed price. Once I have them…

The first time I applied for REGOs I was told it couldn’t be approv4ed because the site was not finished, and when I said can be put my application on hold, they said no, we have to delete it ,and in future you cannot re-use the name, so you need a new one.

Yes, words fail me too.

OFGEM office

Anyway, my second application was responded to with a huge bunch of requests for extra data. You would not believe what they question or want clarified or what level of detail they want. It reminds me of the planning permission system, where it is 100% clear that this is not actually a process, but a game, where their success criteria is that they get to reject you. There is zero attempt at ‘good faith’ evaluation of whether or not its a solar farm, and 100% focus on looking for any minor reason to reject the application. It is awful.

I have replied with a lot of detail, and screenshots of settings, and forms and charts and so on. They seem to want evidence that the farms output is limited to 900kw (it is) but the only evidence of this is that the inverters are set to only output this amount. There is literally no other way I can prove this, and I expect yet more arguments. Its totally ridiculous, because we are basically arguing with bureaucrats, who are absolutely unaffected by how much power we export. The DNO (the distribution company ) IS happy with our inverter limits, and its their substations, transformers and cables that will explode if we are lying, so if its good enough for them… you would think this was a done deal…

I will be very happy with this ridiculous procedure is completed and we actually are being paid the correct amount for each MWH. Luckily you can back-date REGOs, so we are not actually losing out. In the meantime, this generates a ton of work for bureaucrats to argue about forms which achieves absolutely nothing. BTW this is all extra costs on your energy bill.

Expect a super upbeat blog post if it gets approved though!

What happens when a storm hits a solar farm

My solar farm got hit by a big storm. The storm was pretty bad, lots of property damage across the UK. For the whole of my life, I have never worried about such things as long as my house is ok and no trees fall on us. What else would I have to worry about? Its not like I have had a huge number of very expensive to install sheets of silicon and glass standing on a hilltop is it? Wait…

One text message you don’t really want to get is one from the landowner your solar farm is on telling you that panels have been ripped off their frames by high winds. This is scary news, especially before you see pictures. At that point, you worry how many have been destroyed? 100? 500? And how much will it cost to fix? and how much other damage is done.

As it happens, it was not too bad. The total damage is about £8k. Annoyingly thats the cost to actually fix and replace things, not the actual material cost of new panels, because we happen to have 30 spare ones on site anyway. This seems super expensive to me, but I have not personally visited the site to look at the damage. Its not just that ‘some’ panels are destroyed, others would have been scratched by flying debris, and also actual metal supporting beams were twisted and broken by the force of the winds. Yup, it was super windy. It looks like among the other costs we need…

  • 25x M12 fixing sets (wholesale about £10 each, so about £250)
  • 50x M8 fixing sets (not sure, about £2 each? so about £100)

Which means a lot of it is going to be labour, and no doubt travel expenses. Its a REAL PAIN that the site is so far away. Also of course all of this is through 3rd parties. I pay a construction company who pay contractors. Those 2 extra profit margins add up. If I was younger and more ambitious I’d hire a bunch of people and build my own solar farms with my own employees, which would be much more efficient…

The storm was bad enough that the power was down in that area anyway. The farm got shut off, and for reasons that are not clear, was switched off for a long period before people got up there and switched everything back on. At one point, the farm was briefly reconnected but not transmitting data, so in the chaos that ensued it was not clear if even switching it back on worked. This was very very stressful for me. I would have liked to go a year before I had any major bills for anything relating to this project, but there you go… As a result, the power output chart looks like this:

Thats quite a gap. The site did actually come back on and generated on the 19th as I can see from the pure meter stats instead of the inverter reported stats:

But even so, we had just under 2 weeks of downtime with zero generation. By absolute luck, this was in December, the lowest point in the year for generation. If this had been mid summer it would have a been a catastrophe.

There are other solar farms that were massively damaged and half destroyed by the same storm, so I actually think we got off lightly. Also it seems like the damage was to a specific strip of the farm on the very crest of the hill, so its likely that this is just ‘the weakest point’. We are looking at strengthening some bits on those areas given that we now have real world data that this is the part of the farm most at risk.

Mounting systems for solar panels are rated for certain windspeeds, but although ‘the windspeed’ on a certain day might only be 60mph, there may be specific small pockets where it will hit much higher. The only real defense against this sort of thing is going to be accepting some occasional storm damage and factoring it into the budget :(.

Of course the real kicker in all this is that the climate emergency is meaning extreme weather events like this will get worse and worse and worse unless we take drastic action. I have made a 25 year investment, and who knows how bad the peak storms will be in 2050 unless we actually do something. Of course this is not just a renewable energy problem. Roads, railway lines, power lines, and all infrastructure gets very vulnerable once rain, wind and temperature levels move outside the normal boundaries. The UK almost lost its marbles when we had some 40 degree days a few years back, but sadly thats likely to become the norm.

It really is about time people started giving a damn, but recent ‘drill baby drill’ rhetoric from the USA suggests that we are rapidly heading in the wrong direction. This is insane :(.

On the positive side, £8k damage on a £1.5m installation is not the end of the world. Its a pain, but hopefully a one-off. We shall see. I am looking forward to the power output of the site creeping up in the coming months. I am also close to getting certification for REGOs, which will boost the income a bit. You would not believe the detailed forms and paperwork required just to prove I own a solar farm.