Red Hat built a $34 billion business on Linux. IBM bought it. What the deal validated was a hypothesis that had held for four decades: that companies extracting enormous value from shared code would, in self-interest, keep funding the projects they depended on. That hypothesis is now under stress. Not because anyone decided to stop funding open source. Because the industry that funded it most — SaaS — is being dismantled by the industry that depends on it most — AI.
In a previous piece, I explored how eight independent research teams converged on the same insight: instead of building memory systems around the model, train the model itself to manage memory as a learned skill. Post-memory training — using reinforcement learning in the post-training phase — produces agents that decide what to store, delete, consolidate, and retrieve, all optimized against task completion.
Post-memory training has been a hyperfocus of mine over the past few months. If you have followed my recent writing on context management, memory architectures, and the recurring question of why agents degrade after turn 50, this article is where those threads converge. The original pattern was clear enough. Eight independent research teams arrived at the same conclusion: stop building memory systems around the model and train the model itself to manage memory as a learned skill. That convergence was significant.
As we look back on 2025, one thing is clear: artificial intelligence and data are no longer experimental tools sitting at the edges of organizations. They have moved decisively into the core of how businesses operate, compete, and create value. The pace of adoption tells the story. By the end of 2025, roughly one in six people worldwide had used generative AI tools, according to Microsoft’s AI Diffusion Report. In enterprises, momentum was even stronger, with nearly 70% of global organizations deploying generative AI in at least one business function by mid-year. What began as isolated pilots has rapidly evolved into embedded capabilities affecting decision-making, customer engagement, and operational efficiency.
While AI has unlocked vast possibilities for the industry, large-scale implementation remains challenging. Overall, only a minority of retailers have successfully operationalized personalization at scale, and many organizations are still constrained by gaps in talent readiness and change management, slowing their transformation journeys.
In sport, data and AI are primarily associated with on-field performance: player analytics, tactical modelling and injury prevention. Technology has expanded the boundaries of athletic achievement, enabling athletes to push beyond previous limits. So why aren’t more sports organisations applying the same thinking to the business of sport? Sport is an intensely competitive entertainment industry where marginal gains in areas such as fan engagement, content, operations and commercial decision-making can matter as much as results.
The technology is real but immature. The trajectory is clear but the timeline is not. Most organizations deploying long-running AI in 2026 will learn expensive lessons. A few will gain genuine advantages. The difference will come down to three things: where they deploy, how they govern, and whether they understand what "autonomous" actually means in practice.
In late 2025, a technique with the absurd name "Ralph Wiggum" went viral among practitioners. The name comes from a Simpsons character—deliberately silly, because the core idea is almost embarrassingly simple.
In March 2025, a research organization called METR published a finding that got less attention than it deserved. They had been measuring something unfashionable: how long AI systems could work on tasks before they broke down. Not what they could do in a single interaction. METR wanted to know how long they could sustain coherent, useful effort.
