Anthropic shipped Fable 5.1 and Mythos 5.1 today with numbered safety claims attached — 85% fewer biology false positives, 60% fewer cyber ones, an invisible EU watermark. Each traced to its source.
claude-fable-5-1 keeps Fable 5's headline price and adds a cache-read discount, a mid-conversation effort switch, and a watermark you cannot turn off — this is what Anthropic's own documentation says a builder needs to know before shipping against it.
Anthropic's own numbers for Claude Fable 5.1, checked against each benchmark's own home page, the effort settings the announcement discloses, and what nobody outside Anthropic had independently reproduced by launch's end.
A 1977 drought on one Galapagos islet selected for deeper beaks at an intensity unmatched in vertebrate field studies. Decades of caliper measurements later, selection had reversed twice and left its signature in two named genes.
Nuclear power's seventy-year record is a lineage anyone can trace: a founder event, a dominant clade, extinct branches with named causes of death, three selection shocks, and a contested resurrection now under construction.
An apple maggot fly, two new wildflower species, a warbler ring around Tibet, and a mosquito's rewritten origin story — four cases of speciation caught forming, each later revised by the same kind of evidence that first established it.
Trial-and-error synthesis, computational screening, and autonomous laboratories are not rivals for one crown — they close different parts of the gap between a candidate structure and a material you can trust in service.
Flask chemistry, rover geochemistry, and exoplanet spectroscopy all study whether life happens twice, but they trade directness for reach in different amounts and none of them can stand in for the others.
Supersymmetry, extra dimensions, hidden sectors, and effective field theory all predict a layer beneath the Standard Model. Null collider results have not eliminated any of them, but they have changed what each can still say.
Telescopic spectroscopy, orbiters, landers, and sample return answer different questions at different costs; none of them substitutes for the others.