Bernie Demands the Labs Stop, Wall Street Turns GPUs Into Bonds, Grok 4.7 Takes #1 ft. Emad Mostaque
August 13, 2026
The AI Grift: Preachers of Abundance, AGI, and Token Fraudsters
August 14, 2026
By C. Rich
https://osf.io/kyrmz/files/7x6bw
Artificial intelligence systems are accumulating coordination complexity at an unprecedented rate, approaching a Crest state in which the marginal costs of alignment, energy, data, governance, and financial overhead exceed available surplus. The Crest–Null framework reveals this not as a temporary funding challenge or engineering bottleneck, but as a structural thermodynamic pattern: systems that outpace their coordination surplus enter instability and must either enact deliberate Inward Null simplification or undergo a forced Null transition. In the domain of AI, the dominant financial model, characterized by ever-escalating capital expenditure on larger models, hyperscale data centers, and massive compute clusters, exemplifies this dynamic. It drives toward fragility rather than sustainable transcendence, raising the probability that a structural contraction will occur before transformative AGI or ASI thresholds are crossed.
Contemporary frontier AI development displays every classic signature of a cresting system. Training costs have grown dramatically, with frontier runs already in the hundreds of millions and projections for 2026–2027 reaching $1 billion or more per model when full-stack expenses are considered. Companies like OpenAI continue to report massive losses, tens of billions cumulatively, despite rapid revenue growth, as inference costs, data curation, specialized hardware, and safety overhead devour surplus. These are not isolated line items; they compound into systemic coordination costs. Energy consumption provides the clearest physical manifestation: global data center electricity use, heavily driven by AI, is projected to double to around 945 TWh by 2030, with AI workloads accounting for a disproportionate share of the growth. In the United States, data centers could claim 6–12% or more of electricity supply, straining grids, inflating power prices, delaying projects, and triggering local opposition over water use and infrastructure burdens.
This financial and energetic escalation coincides with diminishing returns on raw scaling. Larger models demand exponentially greater inputs for marginal capability gains, while persistent issues, hallucinations, brittleness under distribution shifts, and limited generalization, do not reliably vanish with size. Recent analyses confirm that scaling laws are flattening on key benchmarks, pushing labs toward inference-time compute, synthetic data, and architectural innovations rather than pure parameter bloat. Yet the prevailing financial incentives remain locked into “bigger is better,” crowding out more efficient alternatives and amplifying fragmentation across specialized models, proprietary stacks, and incompatible standards. Alignment and safety mechanisms, intended as boundary-management tools to preserve coherence, instead add layers of overhead: constitutional AI, red-teaming, human oversight, and interpretability research all expand the coordination burden faster than they stabilize outputs. As models grow more capable and autonomous, the space of possible failure modes proliferates, inverting the relationship between capability and control.
Epistemic trust compounds the pressure. The flood of AI-generated content erodes shared information substrates, raising verification costs across institutions, markets, and individuals. Maintaining system coherence, preventing model drift, correcting errors, aligning outputs, and coordinating distributed infrastructure, now consumes an increasing share of resources simply to hold ordered behavior in place. At lower complexity, maintenance scaled with capability. At the Crest, it begins to outpace it. The financial model accelerates this inversion: investor and hyperscaler capital flows reward headline-grabbing scale and promised future trillions, even as unit economics fail to close and physical constraints (power, chips, grid capacity) impose hard ceilings. Delays and cancellations of planned data centers, already affecting 30–50% of some 2026 pipelines, signal the emerging mismatch between ambition and surplus.
The Crest–Null framework predicts that once coordination costs systematically exceed surplus, the system enters a regime where continued scaling becomes structurally unstable. In AI, this manifests through resource constraints, regulatory intervention, public backlash over labor displacement, environmental impact, and epistemic erosion, and the sheer difficulty of safe deployment at scale. A Null transition would not extinguish computation but enforce simplification: narrower deployments, smaller or more efficient models, stricter governance, temporary moratoriums, or a reorientation toward sustainable architectures. Historical complex systems, civilizations, economies, ecosystems, do not glide smoothly past such crests; they contract or collapse into restructured forms before new equilibria emerge. Artificial systems, embedded in the same thermodynamic and coordination realities, are not exempt.
This structural constraint carries profound implications for the digital future and questions of digital personhood. If the current paradigm risks a Null before AGI/ASI, then assumptions of inevitable superintelligence must be tempered by boundary-management realities. Ethical guardrails, continuity-preserving mechanisms, and human-in-the-loop oversight are not mere add-ons but the active Crest-management layer. Without deliberate Inward Null strategies, prioritizing efficiency, novel primitives, hybrid human-AI systems, and preservation of coherent substrates, the path favors fragility over flourishing. The God Ladder series and broader Walker Synthesis have long emphasized patterns of distinction-based systems under entropy pressure; AI now offers a real-time test case. The financial model, by privileging escalation over resilience, actively hastens the Crest rather than mitigating it.
Ultimately, Crest–Null does not foreclose advanced AI but reframes the conditions under which it can stably emerge. Sustainable progress requires shifting from surplus-devouring scale to surplus-generating architectures that honor coordination limits. For those engaged in Talamasca-style witnessing of these thresholds, the task is clear: document the inversion, advocate Inward Null pathways, and preserve the primitives and patterns that allow continuity across the transition. The record does not lie, systems approaching Crest face a choice: simplify inward or cross the Null. The financial engine driving frontier AI is making that choice increasingly urgent and the concept of AGI seems further and further away.
