It’s the second week of October 2026. GPT-6 Astra is a month old, Claude Opus 5.5 is a fortnight old, and somewhere this week a business with ten-year-old kit on the floor and ten-year-old software in the office is being told to buy new. If that’s you, stop and listen to this. I want to come at it from a strange direction…video games industry, because this month has all been about the speed of reverse engineering, technological up-cycling, as the first part of creative evolution for business. The IP itself stays owned. The gap between that and what a new startup can build with the knowledge inside it (the R&D in mechanics, physics, engineering and human behaviour) is getting wider.
The business world always does this…
Reverse engineering is older than software. Businesses hire talent from competing institutions. They reverse engineer; innovations, strategy, engineering, consumer insights, and content. They use other people’s conversations as the starting point to drive new creative executions. You could argue that’s taking the rules and complexity of one person’s IP, or one company’s IP, but that’s exactly what the business world does. So for me, what’s happening now is the same thing, it’s just hyper localised.
On 3 September OpenAI published the system card for GPT-6 Astra, and in it there’s a benchmark called SRE-Bench: can the model reverse engineer a software binary with no access to the source? Given four attempts, Astra solved 99.2% of the challenges. Its predecessor managed 68.7%. For scale, a group of volunteers decompiled Super Mario 64 by hand and finished in 2019, after about two years. That’s now a four-attempt job.
So we’re looking at a world where the distillation of that complexity, and the re-use of it, is the starting point for personal software in effective real time, and for personal creative capability too.
So, back to games
The games industry is the obvious place to start, because the moat was always the difference between the source code and the binary that ships. Grand Theft Auto V. Or Call of Duty, with its approach to unique simulations and physics, plus the fluidity of human ergonomics for the first-person shooter genre, nailed in a way nobody else has ever really matched. These are the standouts: games with release pipelines spanning decades without a real competitor sitting alongside them.
For Grand Theft Auto it’s not just the physics, the rendering and the real-time events. It was their attention to the narrative and storyline that made it a compelling piece of game creation (GTA V came out in 2013 and Take-Two’s own filings put it at over 230 million copies, with no successor until next month).
Grand Theft Auto since 1997, Call of Duty since 2003, Mario since 1985. All of these franchises have been the benchmark for many a game developer, and have become the human expectation for the bar and level that most modern console and PC gamers expect. I’m not talking about the IP. I’m talking about the mechanics under the hood, these were the hardest part to master, and that’s the part that just got cheap to recover. Just like hiring the chief animator from Treyarch, reverse engineering the code and the physics and the ability to get that fluidity is now just hours away.
This isn’t a conversation about creating your own games, though. This is a conversation about creating your own experiences and your own media. Think about it this way. If you can put Spider-Man into a Batman world, then realistically, are we that far away from putting Spider-Man into a Batman film? Are we that far away from creating our own worlds to experience our own things within them? I’ve spoken about this on many occasions, we do this all of the time in the stories we tell, the memories that we make, and the experiences that we communicate - become hyper personalised to our own tendencies and memories.
The innovation trampoline
Isn’t this just piracy with a better vocabulary?
IP and piracy belong in this conversation, but realistically we’re not talking about copying Spider-Man’s IP into a Batman game. We’re talking about mechanics. Physics, routing, the pipelines and flows that make a thing work. Microsoft Word versus Apple Pages versus Google Docs: is there an IP claim between those because the software works very similarly? It’s a word processor. Walk down any supermarket own-brand aisle and you’ll find the everyday version of the same question, right up to the point where it goes to court (M&S and Aldi settled over a caterpillar cake in 2022). The line exists, and most of commerce does its business just the right side of it. It’s a factor of evolution that we improve, hand in hand, upon our past experiences and our collective wisdom, but what we’re seeing at the moment is the speed of getting to ‘table stakes’ being so much faster.
So when we extrapolate that into a business setting, we’re not just talking about changing the CRM from an Excel spreadsheet managed poorly and being forced into using Salesforce, or Adobe Creative Suite, or some proprietary model or routing tool because it’s the only one available, or the only one you’ve paid for. We’re talking about taking components that work really well that suit the personal natures and the business practices of SMEs and bridging them between the proprietary elements that you already have in place. I’d call it an “innovation trampoline”: someone else’s finished mechanics as the thing you bounce off, rather than the thing you’re locked into. It skirts close to the line, and I’d rather say so than pretend otherwise.
Finance, where the commercial model relies on ‘lack of visibility’
The financial industry has grown on the basis that its customers don’t move. Money sits where it landed. I used to call that laziness, and it’s partly that, but it’s more accurate to say it’s the difficulty for consumers to engage with the products, or to translate them. Sky do this. The mobile operators have done this. Ofcom have done their best to regulate and simplify these markets (broadband switching finally went to a one-touch process in September 2024). But energy providers, utility providers and financial institutions have built profits and margins off the back of making it difficult to switch, or to properly control what already belongs to you.
If you’re clever you can find accounts in 2026 paying seven or eight per cent. Most people don’t, or can’t. Now, put an AI agent on the customer’s side of it. We’re looking at a world where agent-run savings are a real risk to the financial industry, because you gain the capability and the control to flex the financial industry with your own money at the speed of a Morgan Stanley or a BlackRock. I haven’t seen a launched UK product that does this yet, but the thinking has reached the institutions. Apollo’s chief economist asked at the end of September whether an “agentic bank run” is coming: if every household used AI agents to optimise the return on its cash, banks lose the cheap deposits they lend against.
This is the bit I’d put in front of the board. It’s the movement of the market by individuals adopting an AI opportunity that forces industries to rethink what the value of its product or service actually is. Which is exactly the same reckoning marketers, businesses, creatives and consultants have had to go through, where the traditional commercial models aren’t standing the test of time. Finance is just the latest to consider it.
Ten-year-old cameras
The third example is my favourite, because you can do it on Monday. I saw a post this week of someone using existing CCTV footage, no custom cameras, no specific computer vision model, just old ten-year-old cameras, to build interactive 3D heat maps of their store: watching what people were interested in, where they lingered, what they converted on, and where to put products for the highest potential footfall.
This kind of analysis of physical retail in a virtual way, using old existing hardware, is one of the greatest opportunities commerce will see in the next two to three years. That tier of shopper analytics has been walled off (RetailNext sells it to several hundred large retailers), and here it is, open to a business with one shop and a DVR in the back office. I’m not suggesting you go out and buy that software. I’m suggesting you could build that exact thing to suit your own environment, your own CCTV, your own analytical needs - my point is it’s up to you!
Back to the kit on your floor
Which brings me back to the business being told to buy new. It works both ways.
A business running ten-year-old hardware in a manufacturing landscape is looking at depreciated kit it will have to refinance in order to update, and the case for updating is usually new features and new capabilities rather than failure. The software version is the one I’ve seen at close range. I’ve had a number of clients who are indoctrinated into specific bespoke or proprietary tooling kits that sit very closely between the operational elements of their business, so verbose and so specific that they can’t be serviced in a typical update cycle, which means every change triggers a custom build. What if, instead of replacing that working software, it becomes a headless connection: linking between the existing capability and new functionality in other tools, and the AI is the bridge. The old system keeps doing what it’s good at, and it stops being the only thing in the room.
There’s a marketing frontier inside this too, which I’ll keep for another day: a consumer who wants more interactivity and gets harder to convert because everyone else is using the same methodology, the way digital once rearranged bricks-and-mortar retail.
So, the ability to amplify the potential of the information and capability you already have, to transform the data you can capture and use to better serve what you’re doing, whether that’s solving the complex dynamics of lighting in your computer game, or rearranging shelf units to maximise customer conversion, or getting the most out of your old timesheet software so it cuts three hours of wasted time a week. All of it is not in buying new software. It’s learning how to use AI to use the old stuff that you already have in a more efficient or effective way.
