X Paper READ

TSMC reported September consolidated sales of NT$511.857 billion on October 8: up 54.6% year on year, down 0.6% month on month. Evan highlighted annual growth; this report keeps the company’s original currency rather than adopting a dollar conversion.

January–September sales reached NT$3.898727 trillion, up 41.1%. The two monthly comparisons use different bases: September last year and August this year.

The release does not separate AI-related sales or disclose September profit and margins. Revenue alone cannot establish either the mix of orders or profit growth.

Background: Vaxcyte’s October 5 OPUS-1 announcement reports positive primary endpoints for its 31-valent pneumococcal vaccine candidate. These are immune-response results, not demonstrated reductions in infections or regulatory approval.

The randomized, blinded, active-controlled study enrolled 4,047 adults. Responses were assessed after one month. Its prespecified primary rules allowed comparisons with PCV20 and/or PCV21; all 28 shared serotypes passed.

Separately, only 17 of 19 PCV21-shared serotypes passed the stricter threshold: the response-ratio confidence interval’s lower bound had to exceed 0.667. Types 3 and 12F passed only the historical 0.5 threshold. Overall success does not mean every individual comparison succeeded.

These are company topline findings, without a full dataset independently reviewed here. A filing is planned for the first half of 2028.

On October 8, dotey described Fable planning and reviewing work while Opus subagents execute it. Several sessions stay open, accepting more tasks while delegated work runs.

His September 23 approach used /advisor fable and an execution–review–repair loop. The new account emphasizes centralized orchestration. He also reports enabling 1M context and setting /autocompact 300k.

He finds this easier to manage but more expensive than using Opus 5.5 alone. No task sample, bills or failure rates establish savings or superiority. These are his settings and experience, not a workflow reproduced by this newspaper.

On October 8, guizang said Grok bot collected material, coded and rendered a news video in a cloud VM. His instructions require a trial edition and personal acceptance before scheduling daily 9 a.m. runs.

The workflow uses the previous full local day, converted to UTC; five to seven news threads; official checks for numbers; and React/GSAP, Playwright and FFmpeg for visuals. Shot timing follows measured narration, and the video adds no facts beyond the brief.

The instructions were read, but the full video and workflow were not independently tested here. A demonstration does not establish reliable daily production or complete news coverage.

Background: Alector’s October 5 agreement gives Genentech exclusive worldwide AL050 development and commercialization rights: $100 million upfront, up to $1.17 billion in conditional milestones, plus sales royalties. The maximum is not cash already received.

The investigational program combines engineered GCase with brain delivery technology to address an enzyme deficiency associated with Parkinson’s disease. Genentech takes development, regulatory, manufacturing and commercial responsibilities; Alector keeps its delivery platform.

The program is moving toward clinical development, not an approved treatment with demonstrated human efficacy. Later payments depend on successful milestones.

Follow-up: BBC’s October 6 report cites a Pentagon official’s October 5 statement that the department had stopped using Anthropic products. A previously announced late-August deadline had passed; the reason remains unexplained. No public decision document was independently obtained here.

Other BBC sources said Claude remained embedded in Palantir’s Maven intelligence platform until the preceding week. Deep integration complicates removal, according to the report. Anthropic declined BBC’s request for comment.

Ending one vendor’s use does not establish an end to military AI or an exclusive replacement. This revisits an older reported development, with attribution preserved.

Pichai said early on October 9 Beijing time that Google and BIDMC’s AMIE study had appeared in The Lancet. Patients chat with this research system before appointments. We read the March 9 preprint, not the journal revision; the findings below refer to that earlier version.

In a single-center, single-arm feasibility study, 100 adults used computer text chats up to five days before urgent primary-care visits in Boston. Patients needing emergency care, pregnant patients and mental-health chief complaints were excluded. This was not autonomous emergency medicine.

The final diagnosis, established through eight-week chart follow-up, appeared somewhere in AMIE’s candidate list in 90% of cases. That is not first-choice accuracy. Overall diagnostic-list and management-plan quality did not differ significantly from primary-care doctors; practicality and cost-effectiveness favored doctors.

Physicians monitored every conversation live, with zero safety stops observed. A small, supervised, single-site study cannot establish unsupervised safety. Alphabet funded the research. The journal revision, wider populations and health outcomes remain to be checked.

Hassabis said Google was investing $150 million in Genesis Mission this week. The White House’s October 8 fact sheet places this within more than $6 billion of public, corporate, academic and philanthropic initiatives. About $2.4 billion covers science AI tools and compute credits from 11 companies for over 15 federal agencies.

Listed commitments include NVIDIA’s $1 billion, AMD’s $500 million, OpenAI’s $200 million, and $150 million each from Google and Anthropic. These are mixed forms of support, not cash necessarily transferred at once; earlier federal commitments should not be counted again as fresh funding.

Anthropic specifies three years: Claude, Claude Code and API credits for hundreds of research projects, alongside fusion and quantum-computing partnerships, training and technical support. This expands its existing Energy Department partnership.

A separate $100 million Genesis doctoral fellowship combines scientific disciplines, AI tools and laboratory or industry training. The announcements describe resources and ambitions, not demonstrated discoveries or verified disbursements.

BBC reported on October 8 that Guojun Xuan and Silvia Zhang were arrested in Arcadia, California, and pleaded not guilty in Pasadena. We read the Los Angeles prosecutor’s October 6 release to check charges and timing; allegations have not been established by a verdict.

Both face conspiracy-to-abuse charges and separate accessory and witness-dissuasion counts; Xuan also faces a child-abuse count. The release refers to at least 14 children, elsewhere 15 of 16 children in the home. It does not fully reconcile those counts, so neither is treated as an independently established victim total.

BBC’s 21 children comprises 15 removed from the home and six found elsewhere last year. That is a placement total, not proven abuse victims. The investigation began after a two-month-old was hospitalized with head injuries in May 2025; a search followed in July. The new development is criminal proceedings.

Lawyer Mitchell Krems told BBC the allegations mischaracterized events and his clients were innocent. The prosecutor also states the presumption of innocence. We have not independently reviewed the full complaint or surveillance footage.

Background: Cato’s October 6 post highlighted Stephen Richer’s analysis of the October 1 Cox ruling. Florida federal judge David Leibowitz dismissed the prosecution of Chelsea Cox. We read the order’s opening, scope and conclusion: 18 U.S.C. §611 was unconstitutional as applied to her, not a finding that she qualified to vote.

The court expressly declined a facial challenge covering every application of the statute. Cato stresses that Florida remains able to prohibit and punish noncitizen voting; one district-court decision does not bind courts nationwide. Failed federal prosecution does not mean nationwide legalization.

Richer asks whether the SAVE America Act’s citizenship-document rules verify qualifications established by states or impose an additional federal qualification. That is analysis of potential litigation, not a ruling invalidating that Act. The judge also discusses a hypothetical federal law supporting state qualifications, leaving it to a future real case. We have not checked subsequent appeals; this is labeled background.