Mark Laursen is an advisor, founder, and executive producer with 25+ years building technology companies, gaming platforms, and entertainment products.https://marklaursen.com/https://marklaursen.com/blog/ai-is-running-out-of-power-data-and-quality/https://marklaursen.com/blog/ai-is-running-out-of-power-data-and-quality/AI data centers now consume more electricity than Japan. Human training data may already be exhausted. Models are measurably degrading. These three crises are connected, and fixing one makes the others worse. Here's the research, and what might actually help.Wed, 15 Apr 2026 00:00:00 GMThttps://marklaursen.com/blog/1-bit-llms-could-make-gpus-obsolete/https://marklaursen.com/blog/1-bit-llms-could-make-gpus-obsolete/Microsoft's BitNet matches full-precision LLMs at 12x less energy using ternary weights on a CPU. If 1-bit LLMs scale, NVIDIA's $3 trillion valuation is built on an assumption that may not hold. Here's why AMD and Intel aren't rushing in.Wed, 08 Apr 2026 00:00:00 GMThttps://marklaursen.com/blog/why-your-multi-agent-ai-system-keeps-failing/https://marklaursen.com/blog/why-your-multi-agent-ai-system-keeps-failing/Multi-agent LLM systems fail 41-87% of the time. 79% of those failures come from coordination, not capability. Here is what the research says about building systems that actually work.Fri, 03 Apr 2026 00:00:00 GMThttps://marklaursen.com/blog/ai-agents-need-governance-not-guardrails/https://marklaursen.com/blog/ai-agents-need-governance-not-guardrails/SDK wrappers are guardrails agents can bypass. Infrastructure-level governance — where agents never hold API keys — is the only real enforcement. Here's why.Fri, 13 Mar 2026 00:00:00 GMThttps://marklaursen.com/blog/live-streaming-before-twitch/https://marklaursen.com/blog/live-streaming-before-twitch/In 2008, Nihilum partnered with Xfire and Dyyno to live stream World of Warcraft raids to 2,800 concurrent viewers — years before Twitch launched. Here's what happened.Wed, 28 Jan 2026 00:00:00 GMT