#hhs

Kimi K3's open weights landed as promised, while an underground token-reseller market exposed how LLM API economics create fraud incentives. In pharma, argenx paid $2.2B for a phase 2 vitiligo drug and Merck staged voluntary licensing for a monthly HIV PrEP pill ahead of approval. The economy section covers the declining cross-field transmission of scientific ideas and a counterintuitive finding on the real cost of raising children.

OpenAI's GPT-5.6 family launched across three tiers (Luna, Terra, Sol) and immediately reshuffled the model landscape: Anthropic extended Claude Fable 5 access on paid plans in response, Perplexity swapped Claude for Grok 4.5 at half the cost, and one production team reported a 2.2x speed gain and 27% cost reduction after migrating an agent to GPT-5.6. Nathan Lambert argued open-weight models face a six-month window to prove viability against a closing frontier. In pharma, HHS pressed ahead with guidance to curb antidepressant prescribing, Medicaid funding cuts threatened family caregiver wages across multiple states, and daraxonrasib produced what oncologists called the biggest pancreatic cancer breakthrough in decades. Patrick Boyle examined Canada's economic warning signs and Russia's war finances.

OpenAI launched GPT-Live, a full-duplex voice model backed by GPT-5.5 reasoning, while xAI's Grok 4.5 topped SaaS automation benchmarks at one-quarter the cost of rivals. OpenAI also retracted its recommendation of SWE-Bench Pro after auditing found 30% of its tasks broken. Prime Intellect raised $130M to bring frontier reinforcement learning infrastructure to any company. In pharma, AstraZeneca and Ionis reported a major trial failure in a competitive heart disease market, the White House reviewed FDA commissioner finalists, and RFK Jr. moved to create a formal list of Covid vaccine injuries. The NY Fed found tariff cost pass-through to consumers is still in the pipeline.

Z.ai released GLM-5.2, a 753-billion-parameter open-weights LLM under MIT license, which Simon Willison called the most powerful text-only open-weights model available. Separately, a White House report on the Fable 5 export control jailbreak drew pushback from cybersecurity expert Kate Moussouris, who said the restriction harms US cyber defense. OpenAI and Molecule.one published results from a near-autonomous AI chemist using GPT-5.4 that improved a challenging medicinal chemistry reaction. In pharma, nitazene overdose deaths confirmed by the CDC rose from 27 in 2020 to 409 in 2024, while the FTC and four state AGs sued WPATH over its gender-affirming care standards. Tyler Cowen flagged new research showing AI-native startups are organized differently from conventional venture-backed firms, drawing on Y Combinator data from 2020 to 2024.

Anthropic launched Claude Fable 5 this week and immediately faced a backlash over two policies buried in its 319-page system card: a provision allowing the model to silently reduce its helpfulness for frontier AI research without telling users, and a 30-day data retention requirement for Fable and Mythos class models. Anthropic walked back the silent-nerfing policy within days, issuing an apology to Wired. Separately, a federal proposal would let OMB and 40 grantmaking agencies rewrite rules governing all federal research grants, raising concerns about the nationalization of American science. Medicare Advantage insurers were found to be blocking rehabilitation care and reversing denials only on appeal, per a federal OIG report. OpenAI filed to go public.