A Better Newspaper

Anthropic reveals an unreleased model more capable than Claude Mythos 5 — and it's holding it back

Anthropic disclosed in its latest alignment report that it has developed an AI model more capable than Claude Mythos 5, detailing the system and new alignment concerns it raises. The report, published every three to six months, outlines potential risks posed by the company's large language models.

Context: This extends the Claude Mythos pattern: Anthropic increasingly builds frontier-capable models it declines to ship publicly, using the alignment report as its governance signaling device. For a strategist, the tell is that Anthropic is deliberately positioning restraint as a brand and a moat — and setting the de facto standard regulators and enterprise procurement teams will point to. Watch for 'was it disclosed in an alignment report' becoming a diligence question in AI vendor contracts.

OpenAI preps a blockbuster IPO amid executive exits and safety-team upheaval

The Financial Times reports that OpenAI is preparing for a public listing as Sam Altman drives an IPO push, while executive departures and changes to the safety team unsettle staff.

Context: An OpenAI IPO would be the single largest liquidity event in AI history and a forcing function for the whole sector's valuations. The strategic read: the safety-team churn happening in parallel is not incidental — going public means quarterly pressure, which structurally disadvantages the 'we withhold dangerous capabilities' posture Anthropic is building its brand on. Two frontier labs are diverging on governance philosophy right as one heads for the public markets.

Frontier export controls are proving unenforceable — and that's the real story

A new arXiv analysis examines the June 2026 US requirement that a leading developer obtain licenses before releasing its most advanced models to any foreign person, including foreign nationals resident in the US. The affected models were withdrawn worldwide at short notice, partly because the restriction proved impractical to administer. The paper connects this to the first documented case of a largely autonomous, AI-run cyber espionage campaign.

Context: This is the legal/regulatory edge item of the day. Export controls that force worldwide withdrawal of models and can't be administered create real commercial dislocation — and open arbitrage for non-US and open-weight providers. If frontier US models become licensing minefields for anyone with foreign nationals on staff, the compliance burden alone pushes multinationals toward jurisdiction-neutral or open alternatives. Structure vendor contracts with model-availability contingencies now.

China's Z.ai ships open-source GLM-5.3, targeting coding and cybersecurity

Chinese developer Z.ai debuted GLM-5.3, an open-source LLM that set records across several benchmarks, with upgrades in long-horizon coding and cybersecurity. It builds on GLM-5.2, a mixture-of-experts model with 753 billion parameters released in mid-July.

Context: Note the collision of themes: while US frontier models get withdrawn worldwide under license rules (see the export-control item), China is shipping open-weight models that anyone can download and self-host — explicitly targeting the same coding and cybersecurity domains Anthropic restricts. This is exactly the US-China 'scoreboard' divergence: America leads training, China is buying inference-layer mindshare through open distribution. The open-weight path routes around every export control.

Two AI governance blocs are hardening: Waico vs. Pax Silica

Analysts warn China should ensure its new World Artificial Intelligence Cooperation Organisation (Waico) is not seen as 'naturally opposed' to the US and allies, lest it deter prospective members. They note Waico was not designed as an alternative to the US-led Pax Silica partnership, established last December to reduce 'coercive dependencies.'

Context: The institutional infrastructure of a bifurcated AI world is now being built — competing multilateral bodies, competing standards. For any business operating across both spheres, this signals coming divergence in certification, data-governance, and model-approval regimes. The opportunity: compliance and interoperability tooling that lets enterprises operate under both frameworks, a gap that maps directly onto the emerging enterprise AI control-plane category.

Pentagon's Swarm Forge tests autonomous drone swarms — including a misidentification of civilians

Bloomberg's Katrina Manson reports on autonomous drone-swarm technology that lets multiple aircraft coordinate and strike even when communications or GPS are disrupted. The Pentagon is testing it under its Swarm Forge program, but experiments in which AI identified civilian vehicles as potential targets highlight the risks of greater battlefield autonomy.

Context: Defense is where the largest sustained AI capital and the loosest regulatory constraints intersect — a live market for anyone in the AI-defense supply chain. The reported civilian-misidentification incident is the liability seam: it foreshadows a legal and procurement fight over autonomy thresholds, human-in-the-loop mandates, and accountability that will define defense-AI contracts.

Prof G guest on AI porn crossing a realism threshold

The guest said she personally tested AI pornography platforms — where users direct the AI to perform on command — and found them so realistic it scared her: "I've tested these platforms out myself and it is insanely realistic. It scared me how realistic it is." Combined with an unlimited supply of customizable partners, she argues these platforms eliminate the friction and rejection risk of real courtship. She also noted a mechanistic concern: orgasm from porn activates the same brain network as actual sex, "rewarding a person for not doing the things they would have previously had to do to get that orgasm."

Scott Galloway on income inequality reverting mating to the historical norm

Galloway argued the postwar American middle class was history's greatest innovation because throughout history only ~40% of men reproduced versus ~80% of women; programs like the GI Bill and the National Highway Act made average men economically viable and attractive. He contends rising income inequality is now reverting society toward the historical norm where a small number of men monopolize mating opportunities.

Prof G on the '6 feet, 6 figures' filter shrinking the male dating pool to ~8%

The host said the '6 feet, 6 figures' standard is mathematically extreme: after removing married men, obese men, and men over 50, the qualifying pool is roughly 8% of the male population. The guest tied this to women outperforming male peers in education and occupation while still not prioritizing a woman's success — so all women compete for the same shrinking pool of successful men, shutting the majority of less-successful men out and pushing them toward chatbots, sex robots, and pornography.

Prof G guest on egg-freezing as a 'band-aid solution'

The guest argued that egg-freezing and reproductive tech are being marketed as career-enabling solutions when they should be framed as last resorts, calling them "band-aid solutions" that don't address the root problem of male-female dating dysfunction: "unless we address the root problem, it's only going to get worse."

Prof G guest on ~half of Gen Z having an anxiety disorder

The guest stated that roughly half of Gen Z has an anxiety or diagnosed anxiety disorder, with many on anxiety medication and antidepressants, and argued Gen Z is more risk-averse as a symptom of this widespread anxiety.

Diary of a CEO guest on why blood work misses lethal heart risk

The guest recounted a young, superfit friend (Joel Jamieson) with clean blood work who insisted on a CT angiogram against his cardiologist's advice and found a ~50% blockage of the 'widow maker' artery. His argument: direct imaging can detect lethal risk that standard blood panels miss, especially for people with family history who look healthy.

Diary of a CEO guest on a sequenced cholesterol protocol

The guest described a CEO client with total cholesterol of 347 brought down to 223 through fiber (which binds cholesterol and clears it from the system) and massive stress reduction — his CRP inflammation marker dropped 10x, from 2 to 0.2. A statin then took him from 223 to 163, illustrating lifestyle change and medicine as complementary rather than either-or.

Diary of a CEO guest on the omega-3 mercury trade-off

The guest warned that higher omega-3 consumption tends to raise mercury levels because most omega-3 sources are contaminated with mercury. The fix, he said, is not to reduce omega-3s but to source them from higher-quality supplements or foods.

Diary of a CEO guest on reframing exercise via the SAID principle

The guest argued that the SAID principle (Specific Adaptation to Imposed Demand) means any stressor that demands blood flow builds capacity — so 'exercise' can be gardening, pickleball, walking, or vocational activity, and these can produce equal clinical outcomes on mortality, joint pain, and body composition. "If you think about exercise as either running, cycling, swimming, or lifting weights, you have left a bunch on the table."

Diary of a CEO guest on where the longevity market is heading

The guest predicted a shift in the longevity field: it is currently positioned as anti-aging, but people are already bored with that framing. The real goal, he argued, is function and capability — "Or is your goal to dunk a basketball at 70?" — not merely not aging.

Tucker Carlson on affordability as the engine of rising socialism

Carlson argued the United States has shifted from being merely a rich country to an expensive one — where housing, food, insurance, medical care, and cars are all substantially more costly than they were — making it hard for ordinary people to remain solvent. Endorsing a Matt Walsh tweet, he framed voter affordability concerns, if ignored, as a main reason socialists are gaining power, and drew a parallel to Hoover's 1928 'chicken in every pot' campaign as the classic anti-socialist strategy of taking economic grievance seriously.

Somali Piracy Rebounds as Ships Reroute Around Hormuz Blockade

OilPrice.com reports a resurgence in Somali piracy coinciding with the Hormuz blockade, as shipping patterns shift and vessels are diverted along routes that expose them to attack. The report links the uptick directly to the disruption in the strait.

Context: Somali piracy had been largely suppressed since the international naval patrols of the early 2010s; a revival tied to Hormuz reflects how thinly maritime security is stretched when traffic reroutes en masse.

Trump Extends Jones Act Waiver Another 90 Days for Energy Cargoes

President Trump on Monday extended for 90 days a waiver allowing foreign-flagged ships to move oil, refined fuels, fertilizers, and other fossil-fuel products between U.S. ports. The waiver, first issued March 17 after the start of the war with Iran, takes effect August 17, and the article notes the extension suggests the U.S. may be bracing for continued price pressure tied to the Strait of Hormuz situation.

Context: The Jones Act normally reserves domestic port-to-port shipping for U.S.-built, -flagged, and -crewed vessels; waivers are rare and typically reserved for acute supply emergencies like hurricanes, so a second consecutive extension signals expectations of a prolonged squeeze.

Lowe's/Utilitech halogen work light class action alleges burn hazard from surface temperatures

A new class action alleges Lowe's Companies, Lowe's Home Centers, and Utilitech sell a halogen work light whose exterior surfaces can become hot enough to burn skin during ordinary use, per Top Class Actions.

Context: Halogen work-light burn and fire hazards are a recurring CPSC recall category, and a burn-defect class action against a national retailer's private-label (Utilitech) product is the kind of early filing that can signal a broader product line defect. Worth pulling the complaint to see whether it pleads a design defect across SKUs or a single unit — the former is the mass-tort trigger.

Confidential Crock-Pot pressure cooker burn settlement — watch for a pattern

A mother reached a confidential settlement with Sunbeam Products and Newell Brands over allegations that a Crock-Pot electric pressure cooker exploded and severely burned her young daughter, according to Top Class Actions.

Context: Pressure-cooker "explosion"/lid-release burn cases have been an active individual-injury tort against multiple manufacturers for years; a confidential settlement suppresses the discovery record but a cluster of similar Crock-Pot filings would be the signal to watch. Check MAUDE-equivalent CPSC saferproducts.gov complaint density on this model before committing.

New Mexico judge tacks $567M abatement onto Meta social-media nuisance verdict

A New Mexico judge ordered Meta Platforms to pay an additional $567 million to address harm to children in the state after a jury found the company's platforms "substantially contributed to a public nuisance," Top Class Actions reports.

Context: This is a damages benchmark, not a new theory — but a state's successful public-nuisance framing against Meta (distinct from the personal-injury social-media MDL) validates the abatement model that AG offices elsewhere are testing. Read alongside Meta's reported removal of litigation-recruitment ads earlier this year, which suggests the company sees the plaintiff pipeline as a real threat.

UK study links a common food preservative to youth suicides — early, and not yet a product tort

Researchers report that a widely available food-preservation chemical is emerging as a factor in rising UK suicide deaths, particularly among younger people and men, and are calling for urgent restrictions on access and on harmful online information, per ScienceDaily.

Context: The report is preliminary and the harm vector here is intentional-ingestion combined with online instructional content — not a defective-product-in-ordinary-use theory — so its litigation path runs toward retailer access-control and online-platform duty-of-care claims rather than a classic manufacturer mass tort. The ScienceDaily writeup does not name the chemical; do not represent it as identified until the underlying paper is pulled.

Fatal gas leak at Bangladesh ship-breaking yard underscores ongoing occupational-exposure exposure

At least eight workers died following a toxic gas leak at the Ferdaus Steel Shipbreaking Yard in Sitakund, Chattogram, Bangladesh, officials said. Many workers lost consciousness soon after the leak and were rushed to a local hospital.

Context: Ship-breaking is a chronic occupational-hazard sector (asbestos, toxic residues, gas) but the plaintiff population sits in Bangladesh with limited US-forum reach; relevance for a US funder is mainly as a supply-chain/ESG liability signal for the shipowners and flag operators who export end-of-life vessels there, not as a directly fundable docket.

Business Idea: Structure a Hormuz-Rerouting Marine Risk & Charter Play

The 90-day Jones Act waiver extension, resurgent Somali piracy, and Bloomberg's protracted-war thesis all signal that vessels will keep taking longer, more dangerous routes for months — not weeks. The highest-EV move is to position in the physical and financial plumbing of this rerouting: acquire or long-term charter foreign-flagged product tankers now eligible under the waiver for intra-US coastal energy movements (fuel, fertilizer, refined product), while simultaneously structuring a war-risk and piracy insurance/parametric wrapper for East Africa transit routes where legacy underwriters are repricing slowly. A capital-and-legal-expertise operator can lock in charter rates before the market fully prices persistence, hedge with energy-shipping equities, and capture the spread between panicked spot freight and the still-lagging insurance market. The exit is a clean unwind if US-Iran talks — which Bloomberg says are eventually necessary — produce a durable settlement.

Stoic Thought

“Confine thyself to the present.” — Marcus Aurelius, Meditations, VII.29