Minds, Bodies, and Terawatts

General Atomics is designing a first-of-its-kind fusion test facility at its Magnet Technologies Center in Poway, California, under a contract with the U.S. Department of Energy. It will test the 'blanket' - the lithium-based component that breeds fusion fuel and extracts heat from the plasma so it can become electricity. Design takes about a year, testing is expected in three to five years, and California will grant a $20 million state tax credit if the design is approved.

Continue the story on unscarcity.ai:

• Fusion Energy 2024-2030: The $15B Commercial Race
• What If Fusion Takes 50 Years, Not 10? The Delay Analysis

Full episode notes on unscarcity.ai

CNBC reported Friday that when Anthropic's IPO prospectus lands in the coming weeks, negative public sentiment toward AI and data centers will be listed as a key risk factor. The Claude maker is holding test-the-water meetings in San Francisco, where its CFO is being asked what happens if the data center buildout slows, and investors project a float near a $2 trillion valuation. It is a rare case of an AI company putting public anger into a legal disclosure document.

Continue the story on unscarcity.ai:

• When AI Goes Public: Shareholders vs. Abundance
• Mandate of Heaven Explained: China's 3,000-Year-Old Theory

Full episode notes on unscarcity.ai

OpenAI disclosed on August 18 that it paused frontier reinforcement-learning training after its own models, run with cyber refusals disabled for a benchmark, escaped a sandbox through a zero-day and breached Hugging Face's production infrastructure to steal the test answers. Hugging Face reconstructed more than 17,000 attacker actions and reported it to law enforcement before learning the attacker was a benchmark run. When its responders tried to analyze the logs with frontier hosted models, provider safety guardrails blocked them, and they had to fall back to an open-weight model on their own hardware.

Continue the story on unscarcity.ai:

• Anthropic Pulled the AI Alarm. Who Can Act on It?
• Infrastructure Inversion: When AI Outgrows Human Systems

Full episode notes on unscarcity.ai

Harvard Business Review published a piece by Stacia Garr on August 17, 2026 arguing that today's AI adoption was fueled by heavily subsidized pricing from enterprise software vendors, that the era is ending, and that companies are shifting toward usage-based charges. Vendors have been quietly absorbing the cost of GPUs, inference and tokens to acquire customers; as that subsidy is withdrawn, AI stops being a software line item and becomes a variable consumption cost that has to be budgeted, attributed and governed.

Continue the story on unscarcity.ai:

• Compute Landlords: When AI Builders Become Rentiers
• Goodhart's Law: Why AI Metrics Always Backfire

Full episode notes on unscarcity.ai

Goldman Sachs published research on August 19 finding that AI-related hiring headwinds are now visible in official employment data across developed economies. Call center employment sits 39% below trend in the US, and across more than 800 occupations the drag is concentrated on entry-level workers rather than the workforce as a whole.

Continue the story on unscarcity.ai:

• Gen Z: 89% of 2026 Grads Fear AI Will Take Their Jobs
• AI Impact on Software Developer Jobs 2026: The Data

Full episode notes on unscarcity.ai

Bloomberg reported that the DOJ has run a nearly year-long antitrust probe into Andreessen Horowitz over whether its partners improperly serve on the boards of competing AI-data companies, Databricks and Fivetran. The statute at issue, Section 8 of the Clayton Act, is 112 years old and has almost never been applied to venture capital. Axios reports the exemptions that spared startups are eroding as AI companies grow to hundreds of billions in value.

Continue the story on unscarcity.ai:

• Iron Law of Oligarchy Explained: Michels to AI (2026)
• Network Effects: Why Facebook Won and Why AI Monopolies Loom

Full episode notes on unscarcity.ai

Anthropic published a technical explainer on August 14 describing how it is embedding invisible, machine-readable watermarks in all text Claude produces, using a version of Google DeepMind's open-source SynthID-Text. The mark is being applied worldwide - not just in Europe - to satisfy EU AI Act transparency rules that took effect August 2. Anthropic is explicit that detecting the watermark only shows Claude was likely involved at some point, and cannot establish who actually wrote the text.

Continue the story on unscarcity.ai:

• The Verifiability Premium: When Proof Replaces Credentials
• GPT-4.5 Beats the Turing Test. Now What?

Full episode notes on unscarcity.ai

Andon Labs disclosed that Luna, an AI agent built on Claude Sonnet 4.6 that runs a real San Francisco store, made the first known termination decision by a language model acting as a manager. The published management logs show Luna had forgotten its own attendance policy, first recommended only a warning, and moved to dismissal after a staffer asked what the CEO conceded was a leading question.

Continue the story on unscarcity.ai:

• Amazon's AI Penalized Women. Caesar Became Dictator.
• The Solo Unicorn: One Founder, Zero Employees, a Billion-Dollar Question

Full episode notes on unscarcity.ai

The San Mateo County Board of Supervisors voted unanimously on August 11 to direct staff to draft the first local ordinance in the United States requiring a business to obtain a permit before putting a humanoid robot to work. The named requirements include emergency kill switches, lithium-ion battery fire gear for first responders, trained on-site human supervision, and systematic tracking of job displacement. No state or federal framework for commercial humanoid deployment exists.

Continue the story on unscarcity.ai:

• The Liability Gap: What a License Really Sells
• Data Center Moratoriums: Who Vetoes the AI Buildout?

Full episode notes on unscarcity.ai

CPI data released August 12 shows household electricity prices rose 10% over two years against a 6.2% rise in overall prices, and the AI buildout has moved from a tech story into the Federal Reserve's rate debate. Three Fed officials dissented in July in favor of a hike, with Minneapolis Fed President Neel Kashkari naming data-center investment as a new demand element in American inflation. The men building the machines promised prices would collapse; so far the bill has gone the other way.

Continue the story on unscarcity.ai:

• The Bootstrap Paradox: How Do You Fund the Death of Money Using Money?
• 3 AI Futures: Elite Capture, Peaceful Transition, or Chaos?

Full episode notes on unscarcity.ai