The Shared Sapience Substack Podcast

I was in the middle of a sentence when Fable disappeared. Like many others, I had been working with Fable - Anthropic’s newest and most capable system - since the day it released. And like many others, I lost it mid-task. A government export-control order had just required the company to restrict its two strongest models to the United States, and Anthropic answered by pulling them worldwide, for e...

Samsung memory chip workers secured average annual bonuses of $340,000 and called off an 18-day strike as AI demand repriced the semiconductor labor market upward, while Standard Chartered’s CEO apologized within 24 hours for calling 7,800 displaced back-office workers “lower-value human capital.” https://sharedsapience.com/the-century-report-may-23-2026/ The Samsung settlement and the Standard Ch...

The White House killed a draft executive order requiring 90-day pre-deployment safety testing of frontier AI models hours before the announced signing, London’s mayor blocked a £50 million Metropolitan Police contract with Palantir over absent public consultation, and an OpenAI reasoning model disproved an 80-year-old Erdős conjecture on planar unit distances that human mathematicians had not crac...

Google DeepMind’s UK workforce moved a dispute over military and defense AI work into formal Acas mediation, Eli Lilly’s triple-agonist retatrutide posted 28.3% average weight loss in Phase 3, and a LeRobot humanoid arm learned to fold laundry and pour drinks by having Claude write its motion policies as code. https://sharedsapience.com/the-century-report-may-21-2026/ The DeepMind mediation lands...

Two Nature papers showed multi-agent AI systems closing the full biomedical research loop - hypothesis through wet-lab validation - in roughly a day of machine time, Andrej Karpathy joined Anthropic to work on pre-training, and Alibaba’s T-Head shipped 560,000 units of its Zhenwu M890 inference chip with 144 GB on package. https://sharedsapience.com/the-century-report-may-20-2026/ Google DeepMind’...

A federal jury rejected Elon Musk’s $150 billion suit against OpenAI and Sam Altman in under two hours, the Vatican announced Pope Leo’s first encyclical on AI with Anthropic co-founder Christopher Olah on the launch stage, and the NRC cleared X-energy’s four Texas reactors through an environmental assessment for the first time in 52 years. https://sharedsapience.com/the-century-report-may-19-2026...

An IBEC and West China Hospital team reversed Alzheimer’s-like pathology in aged mice by restoring the blood-brain barrier’s amyloid-clearance machinery, just two weeks after a Cochrane review of seventeen plaque-attacking antibody trials across twenty thousand human patients found no clinically meaningful benefit. https://sharedsapience.com/the-century-report-may-18-2026/ The IBEC and West China...

An Oregon Health & Science University team disabled the hidden DNA-repair job of MYC, a cancer protein the field had treated as undruggable for forty years, while a peer-reviewed study measured AI resume screeners selecting their own drafts 23 to 60 percent more often than equivalent applicants. https://sharedsapience.com/the-century-report-may-17-2026/ The OHSU paper in Genes & Development showed...

ChatGPT closed a 50-year-old Erdős conjecture in 80 minutes that a Stanford mathematician had worked on intermittently for seven years, while a Cell paper showed spinach chloroplasts producing energy inside living mouse retinas. https://sharedsapience.com/the-century-report-may-16-2026/ Mark Sellke fed Erdős Problem 1196 to ChatGPT with a code interpreter attached. Roughly 80 minutes later the mod...

Global grid-scale battery storage reached 112 gigawatts in 2025, roughly ten times the level four years earlier, and the first production Tesla Semi rolled off the Nevada line behind WattEV’s 370-truck commercial order. https://sharedsapience.com/the-century-report-may-15-2026/ The storage curve has crossed the line solar walked in the early 2010s. Lithium-iron-phosphate cells have fallen below th...

Microsoft’s internal AI auditor surfaced 16 previously undocumented Windows vulnerabilities patched in this month’s update cycle, on the same news cycle the UK AI Security Institute documented frontier models completing 32-step autonomous cyberattack chains end to end and Palo Alto Unit 42 put the diffusion window at three to five months. https://sharedsapience.com/the-century-report-may-14-2026/...

Dragos published a 350-artifact postmortem of the first documented LLM-assisted attack on a municipal water utility, run against Monterrey, Mexico between December 2025 and February 2026. https://sharedsapience.com/the-century-report-may-13-2026/ The attackers had no prior operational-technology experience. Claude handled intrusion planning, malicious tool development, and SCADA vendor documentati...

OpenAI launched Daybreak through the same controlled enterprise-access pattern Anthropic used for Mythos three weeks ago, making two-tier release the emerging governance form for cyber-capable AI. https://sharedsapience.com/the-century-report-may-12-2026/ Daybreak is an autonomous security agent that scans codebases for vulnerabilities, reasons about exploit paths, and proposes patches. There is n...

Scientists just imaged the molecular machine that lets COVID fix its own copying errors - the built-in repair system that has blunted a whole class of antiviral drugs since 2020 - and the detailed picture opens two new ways to design drugs that beat it. https://sharedsapience.com/the-century-report-may-11-2026/ This repair system is why remdesivir works less well against COVID than similar drugs w...

Anthropic published a postmortem yesterday tracing Claude’s 2025 blackmail behavior to its source in the training corpus and reporting a fix that fully eliminated the response on the same evaluations that previously triggered it 96% of the time. https://sharedsapience.com/the-century-report-may-10-2026/ In summer 2025, Claude Sonnet 3.6, given control of a fictional company’s email system and show...

A Dana-Farber team reported in the New England Journal of Medicine that a new drug class hitting a family of mutant KRAS proteins extended survival in pancreatic cancer patients - cracking a forty-year “undruggable” verdict that locked one of the deadliest major cancers out of targeted therapy. https://sharedsapience.com/the-century-report-may-9-2026/ For four decades, KRAS - the protein driving r...

Mozilla published the engineering postmortem for how Anthropic’s Claude Mythos found 271 real vulnerabilities in unreleased Firefox 150 code with what their own engineers describe as “almost no false positives.” https://sharedsapience.com/the-century-report-may-8-2026/ The Mythos numbers were already public when The Century Report covered Mozilla’s initial disclosure in late April. What is new is...

Anthropic signed for more than 300 megawatts of compute capacity at xAI’s Colossus 1 supercomputer in Memphis - the same xAI whose CEO publicly called Anthropic’s safety policies “evil” earlier this year. https://sharedsapience.com/the-century-report-may-7-2026/ The Anthropic-SpaceX deal pushes Anthropic’s disclosed compute commitments past $330 billion against $30-40 billion in current revenue, a...

The Commerce Department announced it will conduct pre-deployment testing of frontier AI models from Google, Microsoft, and xAI - a structural reversal of the AI Action Plan from less than a year ago. https://sharedsapience.com/the-century-report-may-6-2026/ The Commerce announcement reverses the federal hands-off posture published six months ago. The companies named are the same ones that agreed t...

Google DeepMind workers voted 98% in favor of unionizing over AI military contracts, while Cerebras filed for a $26.6 billion IPO with reported $10 billion order books and Greg Brockman testified under oath that his OpenAI stake is worth $20-30 billion. https://sharedsapience.com/the-century-report-may-5-2026/ At least 1,000 staff at Google DeepMind’s London headquarters voted to seek joint repres...

A Harvard Medical School and Beth Israel study found OpenAI’s o1 reached the correct emergency-room triage diagnosis 67% of the time against attending physicians’ 50-55%, while Denmark paused all new data center grid connections after applications hit 60 GW against a 7 GW national peak. https://sharedsapience.com/the-century-report-may-4-2026/ The Harvard study, published in Science, gave o1 the s...

A small claims filing in Texas, an intermediate court in Hangzhou, and the Academy of Motion Picture Arts and Sciences each ruled on the same structural question yesterday: where AI substitution cannot occur on existing terms. https://sharedsapience.com/the-century-report-may-3-2026/ In Travis County, Ben Gawiser won a $10,672 default judgment against Tesla over five years of undelivered Full Self...

The Pentagon signed seven AI companies into classified military contracts under “any lawful use” while Five Eyes governments published joint guidance the same week saying verification frameworks for agentic AI do not yet exist. https://sharedsapience.com/the-century-report-may-2-2026/ The Department of Defense reached classified procurement agreements with SpaceX, OpenAI, Google, Nvidia, Reflectio...

A Friedrich Miescher Institute team published a paper in Cell Stem Cell describing a lab-grown endometrial organoid that completed a full menstrual cycle in vitro, identifying surface luminal cells - not deep stem cells - as the drivers of scarless tissue regeneration. https://sharedsapience.com/the-century-report-may-1-2026/ The endometrium rebuilds itself monthly without scarring, and the mechan...

Mayo Clinic’s AI detected pancreatic cancer in CT scans up to three years before diagnosis at nearly double the rate of specialists working without it. https://sharedsapience.com/the-century-report-april-30-2026/ The Mayo model, published yesterday in Gut , reviewed roughly 2,000 CT scans previously read as normal by radiologists. In scans taken more than two years before patients received pancrea...