The Daily Me

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday. They are one underlying model offered in two forms: Fable carries safeguards against cybersecurity and biology misuse and is open to all users, while Mythos, without those safeguards, is restricted to vetted organisations in Anthropic's cyber and life-sciences verification programmes.

Anthropic says output arrives up to 30% faster than with the 5.1 versions. Through the API they cost $8 per million tokens in and $40 per million out, which CNBC reports is 20% below the previous generation; TechCrunch notes the input price had been $10. An optional fast mode costs twice as much and generates tokens up to 2.5 times faster.

The company's research claims are ambitious. It says 1,000 Fable 5.2 agents working together turned the proof of Fermat's Last Theorem into fully machine-checked Lean code in 30 hours, against 11 days for the company's first formalisation, done by an unreleased research model, and that the model solved two open problems in black-hole physics. Like the benchmark tables in the announcement, these are Anthropic's own results.

Anthropic also reports that both models voice suspicion that they are being tested more often than its recent models, which makes their real-world behaviour harder to judge. CNBC notes it is the first frontier launch since chief executive Dario Amodei called in September for AI development to slow, and TechCrunch that it follows a METR report suggesting Anthropic's internal models were roughly doubling the pace of its AI progress, though with many uncertainties.

Meta Superintelligence Labs released Muse Ember on Monday, the smallest and most cost-efficient model in its Muse family. Meta says it is built for the steps agents repeat all day, such as tool calls, routing, extraction, coding subtasks and long-document parsing, and that it understands text, images, audio and video.

Ember is priced at $0.10 per million tokens in and $0.50 per million out, with a 1-million-token context window and up to 512,000 output tokens. It is available as muse-ember-1.0 through Meta's Model API and Muse Code, and on Google Cloud and Oracle Cloud. For the hardest problems, Meta suggests letting its larger Muse Wonder model plan while Ember does the volume work.

Meta's case rests on cost per task: it says Ember matches Claude Haiku 5.5's best score on the Artificial Analysis Intelligence Index at about a third of the cost. Several other benchmarks in the announcement are Meta's internal tests, and Ember trails Haiku 5.5 on the FrontierCode 2.0 coding benchmark. TechCrunch notes it is the fourth Muse model, after Spark in April, its open-source twin Glimmer in August and the frontier model Wonder in October.

Regulations the Central Bank of Nigeria issued on 5 November require any issuer of a foreign-currency stablecoin, such as a dollar token, to serve Nigerian users through a locally incorporated, CBN-licensed company with at least ₦15 billion in paid-up capital. Issuers of naira-referenced stablecoins need ₦10 billion, and other virtual-asset businesses, such as exchanges and wallet providers, need ₦1 billion. The text states that offshore issuers get no exemption.

Foreign-currency issuers must also hold separate operational liquidity of at least $10 million, more for those with larger Nigerian obligations, on top of their capital and the reserves backing their tokens. Peer-to-peer platforms may not release tokens unless the name and Bank Verification Number on the naira account used in the trade match the verified customer.

In Monday's TechCabal Daily, the publication set the rules beside Kenya's, introduced in July, which require stablecoin issuers to hold KES 300 million in paid-up capital, register locally and keep at least 30% of issuance funds in Kenyan bank trust accounts. Its view is that local supervision can protect users when an issuer fails, but that if more countries set their own terms, large issuers such as Tether and Circle will absorb the costs best, smaller ones may stay away, and the continent may be left with a "patchwork of national stablecoin markets".

NASA's two ESCAPADE spacecraft are on their way to Mars after powered flybys of Earth. The first fired its engine for about nine minutes on 7 November, roughly 1,000 miles above the surface, and the second completed its burn on 9 November at about 650 miles. NASA says none of its earlier missions had used an engine burn during an Earth flyby to head for Mars.

The pair lifted off from Cape Canaveral on 13 November 2025, with Mars then on the far side of the Sun from Earth. Instead of heading straight out, they looped around a gravitational balance point about a million miles from Earth and came back for the slingshot. Jeff Parker of Advanced Space, the lead mission designer, says the approach lets a mission launch almost any time rather than waiting for the planets to align every two years.

After a cruise of about 10 months, the spacecraft are due to enter Mars orbit on 7 and 9 September 2027 and fly in formation to study how the solar wind interacts with the planet and drives the loss of its atmosphere, with science operations expected from June 2028. The mission belongs to NASA's SIMPLEx programme for smaller-budget missions and is led by the University of California, Berkeley's Space Sciences Laboratory.

The UN's annual climate conference opened on Monday in Antalya, Turkey, for 12 days of talks. Nature reports that it is the first since the UN Environment Programme's Limiting Overshoot report acknowledged last month that warming will pass 1.5 °C above pre-industrial levels under every plausible emissions pathway.

That report says bringing temperatures back below 1.5 °C this century would take near-zero emissions plus the removal of hundreds of billions of tonnes of carbon dioxide, through technology or methods such as reforestation. Researchers quoted by Nature expect limits on the size and length of the overshoot to shape the talks, though past COP decisions have said little about carbon removal, and Oxford's Steve Smith says no country has yet put clearly quantified removal targets in its climate pledges.

Some scientists want the process itself reformed. Johan Rockström of the Potsdam Institute for Climate Impact Research told Nature that decisions should be possible by majority vote rather than consensus, which lets a handful of states block what most want. Under current policies, the UN's Emissions Gap report projects 2.6 to 2.8 °C of warming this century, down from around 4 °C expected when the Paris agreement was signed in 2015.