TabLOID

Alphabet reported $127.6 billion in third-quarter revenue, up 25% from a year earlier. Google Cloud brought in $30.7 billion, up 103%. The cash-flow table tells the other half: $42.3 billion from operations met $51.4 billion in property and equipment purchases, leaving negative $9.1 billion in quarterly free cash flow.

The company earned $2.12 per common share. That figure includes an $8.7 billion loss on equity securities, which reduced earnings per share by 55 cents. Investment swings and operating performance deserve separate attention.

Microsoft, reporting its fiscal first quarter, posted revenue of $92.3 billion, up 19%, with Azure revenue up 48%. Earnings were $4.92 per share under GAAP and $4.89 excluding the impact of its OpenAI investments. Those are distinct measures.

Our reading: demand is visible in the revenue. Whether enormous infrastructure spending earns an attractive return is still a different question.

Alphabet earnings release, distributed unedited · Microsoft results

Governor Kathy Hochul signed S.9051B/A.10379C on October 28. The law restricts features that simulate human identity, cultivate emotional dependency, encourage self-harm or engage minors in sexual conversations.

The governor's explanation says those features may not be offered to under-18s, or to users whose adulthood the company has not confirmed. That distinction matters: the announcement describes restrictions on features and treatment of unknown-age users. It does not say every person must upload a government ID to use every chatbot.

For builders, our takeaway is practical: personality, memory and relationship cues are design decisions with consequences. A warning tucked beneath the chat box does not address what the product actually says.

The announcement describes implementation in the coming months. We are not assigning a precise start date from its separate baby-monitor provision.

New York's signed-legislation announcement

Anthropic's October 27 research report describes an unreleased internal model finding and chaining vulnerabilities in widely used browsers and operating systems. Its Chrome work, the company says, produced 62 assigned CVEs rated high or critical, patched by Chrome 156.

Google's release notes independently credit Anthropic researchers assisted by Claude for multiple fixes. That is useful corroboration of the patch work; it does not independently validate every claim about the model's autonomy.

The distinction from a public product is equally important. This was an internal research system working on selected targets. The report does not establish that every consumer chatbot can reproduce the same results, or that the disclosed Chrome issues remain unpatched.

Our reading: AI security claims are most useful when they leave a trail from discovery to vendor acknowledgement to a shipped fix. Here, there is one.

Anthropic's research · Google's Chrome release notes

PsiQuantum unveiled an optical circuit-switch demonstration built on its silicon-photonics platform. The idea is to route light between connections as large computing systems place more demands on their networks.

The company reports typical laboratory insertion loss of 1.2 decibels and reconfiguration in under three milliseconds for its 8-by-8 demonstration. These are vendor measurements, not independent production benchmarks. Samples are going to selected prospective customers; commercial products are still ahead.

This is a networking component. It is not an announcement of a completed fault-tolerant quantum computer, and the measurements do not establish a particular speedup for AI training.

The interesting possibility is narrower and quite useful: technology developed for a difficult long-term project can find an earlier job moving data.

PsiQuantum's announcement