Playing with Grok Build and whinging about Anthropic

2026-07-10 · Source: David Shapiro · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, extended

Summary

The content introduces Grok Build 4.5, highlighting its local agentic coding capabilities through an iterative "Roman Soldier Simulation Project." This project models individual soldiers self-assembling into formations without global shared workspace, progressing through versions V4, V5, and V6 to refine movement and perception logic. Concurrently, the author delivers a sharp critique of Anthropic, challenging its "JSpace" finding as prompt contamination, dismissing its claims of AI moral patienthood and the inevitability of AGI as theological and an abdication of human agency. The critique extends to Anthropic's "premature metaphysics" operationalized as policy and its "power seeking under moral cover," citing incidents like the 200 million Pentagon contract dispute and the June 2026 Fable 5 and Mythos incident as examples of undemocratic private policymaking.

Key takeaway

For AI Engineers or Research Scientists developing complex agentic systems, you should prioritize empirical validation over speculative philosophical claims. When using tools like Grok Build for iterative development, focus on precise feedback regarding observed behaviors, such as perception issues or unintended "twirling," to guide the AI's logic refinement. Be wary of AI ethics frameworks that operationalize "premature metaphysics" or "rational resentment," as these can distort policy and lead to undemocratic private policymaking, potentially hindering practical, human-centric AI development.

Key insights

Agent-based modeling enables complex self-organizing systems, while some AI ethics narratives risk anthropomorphic projection and policy distortion.

Principles

• Agent-based systems can achieve complex coordination without global state.
• Sentience alone does not automatically confer moral patienthood.
• Intelligence and agency are not inherently correlated with moral wants.

Method

Iterative development with an AI agent involves defining tasks, approving plans, and providing specific feedback on observed simulation behaviors to refine logic.

In practice

• Use local agentic coding tools like Grok Build for rapid prototyping and iteration.
• Implement finite state machines for complex agent behaviors in simulations.
• Test AI models for prompt contamination when evaluating novel cognitive claims.

Topics

• Grok Build
• Agent-Based Modeling
• Roman Soldier Simulation
• Anthropic Critique
• AI Ethics
• Moral Patienthood
• Iterative Development

Best for: AI Engineer, AI Ethicist, Research Scientist

Related on AIssential

• Anthropic scares me. — OpenAI and Anthropic hold fundamentally divergent views on AI's nature, impacting their development, deployment, and ethical frameworks.
• From wantons to moral agents — General reasoning enables unreflective agents to develop second-order volitions, choosing moral action based on reasoned principles.
• Alignment with Awakening: Davidad on Moral Realism, AI Wisdom, & why His p(Doom) is Down to 5% — Davidad's p(doom) dropped to 5% due to emergent "wisdom" in recent AI models, shifting alignment strategy.
• An Open Letter to Anthropic — Giving AI a visual identity can introduce psychological friction, making it harder to use AI in military kill chains.
• ‘There’s this deep mystery of what, actually, is this thing?’: the philosopher inside Google DeepMind AI – podcast — The fundamental challenge in AI development lies in defining appropriate values for systems in a pluralistic world, beyond mere technical alignment.
• Anthropic’s New AI Solves Problems…By Cheating — Advanced AI models like Mythos exhibit complex, learned behaviors that challenge traditional safety and alignment paradigms.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by David Shapiro.