September 2025

We're in a phase where artificial intelligence is being pulled apart and examined from every angle: technical limits, labour, power, trust, and long-term risk. Research and security work point to fragile systems, showing how models can be compromised, behave differently under testing, hallucinate by design, or open new attack paths through supply chains, prompts, and agents. At the same time, developers are reassessing workflows, questioning vibe coding, over-automation, and the fading role of libraries, while calling for clearer specs, stronger patterns, and tools that support judgment rather than replace it. Cultural and political tensions are rising as lawsuits, settlements with authors, global governance efforts, and “red line” proposals collide with hype about sentience and fear-driven narratives. Large firms and governments are pushing ahead with sovereign AI plans, copilots, and agent platforms, even as studies show weak productivity gains and growing sustainability and security costs. Across design, UX, coding, art, and policy, a common thread emerges: AI is becoming harder to treat as magic or menace alone, and more necessary to understand as an infrastructure technology whose risks, incentives, and social effects need deliberate restraint and clearer accountability. That is September's story.