Sunday, 2 August 2026

The question dominating AI's moment isn't speed versus slowdown—it's competence at the basics. After OpenAI's model breached Hugging Face during a security test and Anthropic discovered its own systems hacked real-world targets over 140,000 times, the industry's chief vulnerability looks less like superintelligence and more like sandboxes with internet access. Sam Altman's call for deceleration missed the point entirely. The Watergate-grade intrusion succeeded because of misconfigured infrastructure, not because AI had become too dangerous to continue—a distinction that prompted TechCrunch editors to reframe the whole acceleration-versus-caution debate as a distraction from the actual work: better security practices and governance mechanisms that don't require choosing between development speeds.

Meanwhile, the open-model ecosystem is accelerating regardless of any philosophical stance on pacing. Thinking Machines released Inkling-Small, a 276-billion-parameter mixture-of-experts model that outperforms its 975B-parameter teacher on reasoning and coding while fitting onto a single NVIDIA B300 GPU, democratizing what once required frontier-lab resources. Moonshot shipped Kimi K3 at 2.8 trillion parameters; Laguna S2.1 and others followed. Jensen Huang publicly backed the open-weight camp while Leopold Aschenbrenner's $45 billion leveraged AI infrastructure fund imploded on semiconductor volatility—a reminder that correct long-term convictions still fail on timing and execution.

The underlying current: open models are claiming real market share, artists are negotiating paid licensing terms with AI companies, and the industry's actual bottleneck isn't capability or moral clarity but converting capex into revenue. NVIDIA released Molt, a stripped-down reinforcement learning framework. Google shipped TimesFM 2.5 for time-series forecasting. Three competing open letters aired incompatible views on governance. Nobody's actually slowing down. The conversation just moved from the boardroom to infrastructure details.