Think Far, Act Close: The New Rules for Intelligence on Earth and Mars

Mars will not be settled by intelligence alone. It will be settled by teams that convert intelligence into reliable, verifiable, and accountable physical action - and the bottleneck is not compute. It is the path from uncertain evidence to an approved engineering result. In this grounded, evidence-driven book, engineer and harness-builder Bob J Mills presents the architecture behind that conversion: the Mars Harness. Not a chatbot, not a single agent - a staged intelligence layer (Discover → Engineer → Verify → Deliver → Evolve) in which specialized AI harnesses for science, engineering, and program execution work through explicit work-product contracts, ground-truth gates, and human approval. Inside, you will learn: - Why the seven functions of a harness - orchestration, instructions, tools, context, memory, subagents, and guardrails - matter more than which model runs underneath - Why skills outlast agents, and how the durable capability catalog turns people, models, solvers, and robots into interchangeable executors - How every claim carries an epistemic status - Observed, Derived, Assumed, or Unknown - with provenance recorded separately, so a plausible sentence can never silently become a physical fact - How the structured handoff (the Engineering Evidence Package) and the return path (the Research Request Package) close the loop between science and engineering - How mission memory compounds: preserving evidence, decisions, dissent, and lessons across people, tools, and model generations >=98% - to a modeled settlement-scale scenario of 592 tonnes over 730 days, with every modeled number labeled as modeled >Grounded in the author's Mars Society 2026 presentation and his real open-source harness projects (Hypatia Pro, Intelligent Engineer Pro), this is a practical architecture for making AI into structured engineering capacity - on Mars, and on Earth. Model = intelligence. Harness = mission capability. Fans of rigorous, systems-level thinking on AI and aerospace - readers of engineering-driven nonfiction on space settlement - will find here a book that respects the evidence as much as the ambition.