2025 has been a transformative year for artificial intelligence. We witnessed the emergence of reasoning models like o1 and o3, which demonstrated genuine problem-solving capabilities rather than mere pattern matching. Claude gained the ability to use computers autonomously, DeepSeek proved that frontier-level performance could be achieved on modest budgets, and AI coding assistants became genuinely useful collaborators rather than autocomplete on steroids. The rapid iteration between OpenAI, Anthropic, Google, and emerging players created an almost dizzying pace of advancement, with each month bringing capabilities that would have seemed implausible just a year prior.
Looking ahead to 2026, I expect AI agents to finally deliver on their long-promised potential. We'll likely see models that can reliably execute multi-step tasks over extended periods — managing projects, conducting research, and handling complex workflows with minimal human intervention. The cost of inference will continue to plummet, making sophisticated AI accessible for personal use cases that previously seemed economically absurd. More intriguingly, I suspect we'll witness the first serious applications of AI in scientific discovery: not just analysing data, but formulating hypotheses and designing experiments. The question is no longer whether AI will reshape how we work and create, but how quickly we can adapt to the new landscape it's building.
Perhaps most fascinating is how the conversation around AI has matured. The early hype cycles have given way to more nuanced discussions about capability, safety, and societal integration. We're beginning to understand that the path forward isn't about replacing human judgment but augmenting it — creating tools that extend our cognitive reach while preserving the creativity and intuition that remain distinctly human. As we enter 2026, the organisations and individuals who thrive will be those who learn to work with these systems fluidly, treating them as capable collaborators rather than either infallible oracles or mere toys.