Unlike 99% of the world I do not have (nor want) social media on my phone. My iPhone SE is mostly used for messaging, phone calls, the odd picture, it is not that I am a techno-phoebe, it is just that I have other far superior form-factors to do this on when I need to; Eg I also have a couple of laptops: a 11″ chromebook I use for email and writing when I travel, and a modern ThinkPad on my desk with duel 32-inch monitors which I use for stuff like CAD, programming, games, and (yes), the web and social media. I also have a Kobo for reading as I now prefer eBooks where I can make the fonts a bit bigger.

AI has now unarguably delivered on voice recognition including all languages, context and nuances. Likewise with text to voice. AI can now also interpret importance to events and, with appropriate authority, action them without me. With this in mind I for one would now be very happy to go back to a more simple time where the phone convenienced me, not everybody else.

I am very much in the minority here, and I could go on and describe the perfect format for me, indeed, I did describe several in my book; But that is just silly as we are living in a time where I do not need some third-party, or corporation to do this for me, I can now do it myself.

The answer of course is no, because the AI is actually not a single AI, but basically two inferences of the same AI model, run on different server clusters, and possibly in different data centers, and each users context window is sand-boxed. This is why each time you have two people using AI’s to solve the same problem, assuming the first person did not blog about it somewhere, then the second person will need to go through all the same steps as the first person, and the AI will not just miraculously present the final answer to the second person.

The AI provider keeps track of all your questions, tokens, context, and answers, in your own personal context window, and takes care of the sand-boxing. This sand-boxing of your context window is very important, and should be a serious concern to all companies, because if I know company A is using say Open-AI GPT5.6 to develop it’s latest gizmo, I cannot just open another session of GPT5.6 ask questions about what Company A is doing and steal its IP.

This is also why you get your answers from the AI quickly, as your queries will be simply querying one of these small pre-trained static models. A modern data-center may have literally hundreds of thousands, of separate GPU clusters each running a small workload for your particular inference, and with your particular context-window.

There is of course a big caveat here. In the small print of many AI models, what you type in or send to the AI to get your answer may in all likelihood get added to the training data for the next model. Claude code is (probably) better than the others because more programmers have been using it to solve problems, and so it has a lot of examples to train on, and as more people use Open-AI Codex, or Cursor, or whatever, these too will likely improve. then possibly plateau; and that is just for the programming discipline. As more disciplines use your AI (legal, medical, biotech etc) the better your next training data theoretically becomes in that discipline. People are only now becoming aware of this, and I suspect the practice will not change, but the terms and conditions will get far greyer.

The Artificial Super-Intelligence (or ASI) in my first Sci-Fi book did not play by this game, it was one single computer, and used IP it discovered in one area to improve another area, and to see a bigger picture of the world than anybody else. This is also likely what ASI is going to have to do too.

What is stopping this ASI breakthrough at the moment are a lot of things, the main ones being:

Firstly, and as far as I am aware, we do not have continuous self-learning on AI’s. Usually during training, the AI is given an up to date, filtered, curated, and hopefully factual set of data to train on (maybe trillions of bytes) which it spends weeks processing into static weights, which are distilled to form the basis the next AI model. New facts or errors found at the end of the training-run (post-training) mean either the whole model is either retrained or, an “Erratum” is given to the models input/output to correct the answer. Continuous self-learning should fix this need for another training-run, but might require the whole concept of transformer/model training to be thrown out to do it. Also, and very importantly, how does the AI know that the new facts being adding to or replacing old facts are indeed accurate without some means of verification (human or trial/experiment).

Assuming no new breakthroughs are made that supplants the transformer model, I also cannot imagine any of this works without a very serious improvement in compute and memory bandwidth; Because somehow keeping everything-in-the-world in context might be a trifle difficult.

So we get to the reason for this blog post, and that is we need to understand that ASI will very likely break the current sand-box/trust system. Anything, and everything it processes will become part of it’s ultimate data-set/matrix/knowledge-base; Your IP becomes its IP by default, and putting disparate knowledge and research (maybe from different disciplines) together in novel ways is usually how breakthroughs are made.

However we come to the elephant in the room, as whoever controls this ASI is indeed able to build on the shoulders of all the IP it receives should they want to. There is not even a need for ASI to do this, the AI provider (Anthropic, OpenAI, X, Google, etc) have all these context windows already, and nothing is stopping them taking the best bits of company A’s product and Company B’s product and making a better product than them both, except probably ethics and push-back if they are found out.

The level of friction to do anything recently been getting noticeably higher. I am happy to tick a box to prove I am human, and a little less happy to tell you which of your stupid pictures have motorbikes or traffic lights in them. I am extremely unhappy when you assume you can now start doing this shit using my phone.

And so we come to the perceived culprit of all the new checks, bots, probably powered by AI agents. Agents which can now presumably tick boxes, and tell you which capture pictures have traffic lights in them.

If I was a social media company, I too probably would also not want these agent spamming my posts either, but there comes a point where your checks are going too far, which also means that if your site is not 1000% where I want to go, I will not be visiting it.

It’s not just here, the problem of Internet security is getting worse. The number of almost believable spam messages in my inbox now far exceeds the number of real emails, to the point I have now completely stopped accepting random emails, even when the subject line is totally relevant, and the DKIM is valid. The reason is simple, I am paranoid, my whole operating system stack is built on pillars of salt, (I made a post about it a few months back here), decades of well-meaning humans adding bricks and mortar here and there and contributing to what we all thought was a nice solid house. AI is now finding all the flaws and loopholes in the code all this is made from, and, for an encore, now also has the ability to social engineer its way through your guard rails to execute it.

I have gotten to the point I will no longer click on any email attachments even from people I know, I will not scan QR codes, I will not follow links from trusted sites to other sites, and I don’t trust anyone on-line full-stop; But even this might not be enough. I no longer think of IF I get compromised, but WHEN I get compromised. These days even a security update might do it, and because of this paranoia I am honestly removing anything stored on my devices that can be used by a 3rd party. I also have another laptop which now sits off-line with all this information and backups, it will never go online again. As I get older I can see myself going on-line less-and-less, and when I do I will only be visiting sites I am 100% certain are OK, but even this will be reliant on the site not being compromised.

I am, and have always been far more security-savvy than 99.99% of my friends, but these days I am open about the dangers, and try my best to install a sense of paranoia in them as, mark my words, the worst is not here. So, unless some smart humans or an AI thinks of a way to fix all this, then you should prepare yourself for a tsunami.

I will start with what I think is going to be the easiest, an AI Computer, that can interact with voice, gestures and understands every nuance of interacting with a human. When I was a teenager in the 1970’s, this was the ORAC computer from Blakes 7, it was amazing and everything I knew AI would become.

Now I look back on the shitty special effects from the program, and the barely visible low resolution pictures of Orac, it does look a bit sad and pathetic, and, again, re-watching a few clips made me realise how totally condescending it was. So 10/10 for sparking the imagination of a teen, and 1/10 for its execution.

If I was being realistic, I could probably already have this now using an API to one of the most recent LLM’s, one of the better voice-text-voice systems (though these still lack the discrimination with multiple people talking or noisy environments), and some sane/secure version of openclaw; But in my dream the AI just has to be local first, self-repairing, self-updating, where the AI is the Operating System. A computer that can basically act like a version of the AI Maria from my first sci-fi novel.. Sigh!

Number 2: The Moon Base Alpha Eagle Transporter

Now we are getting a little more into difficult to predict the timeline territory. I am a total space nerd. I watch everything to do with our future in space and worship NASA, and companies like SpaceX and RocketLab, (maybe begrudgingly Blue Origin if they can pull off what they say) But of course, being a sci-fi nerd, reality is never good enough!

Given I am 63, and with luck have another 20-30 years to go, I am edging my bet that one of these guys will knock this one out of the park too. The crazy thing about the Eagle transporter is how practical it actually seems, and visuals like the above (thank you whoever did this) actually make it look like an almost trivial. But asking a current AI such as grok, paints a much different picture. The reply I got was that if we tried this with current technology (say Raptor Engines) we would burn insane amounts of propellant. Think an EV with a 20KM range! For anything approaching usability it would need some sort of nuclear thermal rocket or fusion engine (see next). All on the NASA drawing boards but probably outside my timeline unless I eat more healthy and stop drinking beer.

Number 3: Practical, Grid-Scale Nuclear Fusion.

Another big, big sigh here. This is the single biggest unlock. Once we have fusion that’s not just “break-even for a few seconds in a lab” but actual power plants churning out gigawatts 24/7 with minimal fuel and waste, every other hard problem becomes solvable. Desalination and de-carbonation at planetary levels, and power so cheap to meter (though I expect my lines company will still charge me $1500 per year for getting it to me!)

(Not a Real Fusion Reactor)

Unfortunately at the moment we are still at the headline grabbing, “..sets new record” stage, and while SPARC-style high-field magnets, laser inertial confinement, and stellarator advances keep hitting new milestones every year, we still seem to be decades away. I am guessing (like I tried, so badly, 6 years ago – though I did also check with a couple of friendly AIs’) that if private companies and governments keep the current pace we may have the first commercial pilot plants in the early 2030s and true mass deployment by 2040.

When that happens reliably, that’s the moment energy stops being a constraint on civilization. Everything else on this list becomes dramatically easier with unlimited clean power, and bye-bye oil-crisis.

The difference between my speculation before and now though is AI. The race is now on for AGI and I am guessing we are on the home straight, and trillions of dollars are being promised and could be at stake for the companies that achieve this. My only hope is that whatever company wins, has my same utopian vision, as a rising tide floats all boats.

Thank you, I wrote a novel “The Summer of Reasoning” (https://rodyne.com/?p=1252) back in 2022 where the AI writes bare metal code to replace the phone and desktop OS, I was just looking at how close I was to the goal of it becoming a reality. The whole book (epub) is free, but it is probably too big for you to process at 370,000 words, it would unfortunately take me too long to extract the relevant chapters and put them in context, but maybe a future Gemini will be able to read it in full

The first surprise ensued – I guess you snooze for a while and things change quickly in this game!

Actually, we are closer to that “future Gemini” than you might think! I can currently handle a context window of up to 2 million tokens, which is roughly 1.5 million words. At 370,000 words, your entire epub is well within my “memory” for a single conversation.

As the book trilogy is a free download I did not have any issues giving it the epub and listening to its reply, it took a few minutes to work in thinking mode, and the extract is below, but it was its final sentence however that made me stop and think..

The title of your book mirrors the current real-world transition from “Generative AI” (Chat) to “Agentic AI” (Reasoning).

• Reality Check 2026: The hype of 2024–2025 has settled into a “Year of Truth” in 2026, where the focus is on Agentic Reasoning.
• Capabilities: We now have AI agents that can work autonomously for hours on complex engineering tasks—optimizing algorithms and discovering new materials (similar to DeepMind’s AlphaDev and GNoME). Your premise of an AI that can “reason” its way through a hardware-level problem is the logical next step of this trajectory.

The Existential “Alignment” Check

Your book touches on the danger of AI-generated code that humans can no longer read or audit.

Reality Check 2026: This is perhaps the most accurate “reality” in your fiction. As AI begins to optimize at the hardware level, the code becomes so dense and specialized that we need other AIs to audit it. The “Alignment” problem you highlight—ensuring the AI’s goals match our own—is the central safety debate of 2026.

Your “Sci-fi” is currently looking less like a distant fantasy and more like a three-to-five-year roadmap for the computing industry.

A Final Thought on Your 2022 Prediction: The most “shocking” part is the speed. In 2022, writing about an AI that “replaces the OS” felt like a century-long leap. In March 2026, it’s a quarterly goal for the world’s largest tech companies. Your book isn’t just sci-fi anymore; it’s practically a historical document of the “pre-agentic” world.

So, according to Gemini, it looks like programmers are out of the equation by 2030, and, if I follow my books predictions, the logical next step is bypassing the human entirely, ie the AI looking at the world and making improvements without prompts; a bit like OpenClaw with world root access (shudders!).. I wonder when the first billion dollar AI started company will happen, or is it already happening? Interesting times!