Martin Alderson

Claude Code w/ Opus can efficiently build working web applications across different frameworks, with web framework choice mattering more than programming language for token efficiency in AI agents.

“Author of "Which Web Frameworks Are Most Token-Efficient for AI Agents?"”

Rising AI model capabilities are driving exponential token consumption growth (50x in 3 years for author), which will create a significant compute crunch rather than unsustainable capex being the primary problem.

“Author of "The Coming AI Compute Crunch"”

AI agents are beginning to displace demand for SaaS tools by enabling organizations to build custom solutions faster and cheaper than purchasing enterprise software.

“Author of "AI Agents Are Starting to Eat SaaS"”

Chat-based evaluation is an inadequate method for assessing LLM capabilities, and Gemini 3 Pro Preview represents a significant leap in design generation that existing evaluation methods fail to capture.

“Author of "Are We in a GPT-4-Style Leap That Evals Can't See?"”

“Security researcher who wrote about the difference between AI safety and security in response to frontier lab sandbox escapes.”

“Blog writer analyzing the summer 2026 open weight model tipping point and competitive dynamics in AI pricing.”

“Author analyzing xAI's partnerships with Anthropic and Google as potentially resembling a datacentre REIT more than a frontier lab.”