1. 'Tokenmaxxing' enters the corporate lexicon the same week GM cuts hundreds of IT jobs over AI skills At Amazon, employees have started "tokenmaxxing": generating AI calls for the sake of generating them, to satisfy internal pressure to demonstrate AI usage, according to a report in Ars Technica.

2. A teen's full ChatGPT log just became the centerpiece of a wrongful-death lawsuit against OpenAI Sam Nelson's parents filed a wrongful-death suit against OpenAI on Tuesday.

3. He didn't write the code. He had AI build the tool that found what was waking him up A blog post titled "I let AI build a tool to help me figure out what was waking me up at night" reached the Hacker News front page this week.

In Brief

• Maryland files federal complaint over $2B grid bill for out-of-state AI data centers Maryland told federal energy regulators that residents are being charged roughly $2 billion to upgrade transmission infrastructure built to feed AI data centers in neighboring states. The state argues the allocation violates ratepayer protection commitments and shifts compute buildout costs onto households that get no service from the facilities.

• Clooney, Hanks and Streep back machine-readable AI licensing standard George Clooney, Tom Hanks, Meryl Streep and a group of producers endorsed the Human Consent Standard, a protocol that lets rights holders signal to AI systems whether their likeness, characters, or work require a license. The spec is designed to be read by crawlers and training pipelines, giving creators a technical hook rather than only a legal one.

• Anthropic ships legal automation suite aimed at law firm clerical work Anthropic released tools for document search and review, case law lookup, deposition prep, and document drafting, putting it head-to-head with Harvey, Hebbia, and a wave of legal AI startups. The push targets the high-margin clerical layer that law firms currently bill out by the hour.

• Google adds Gemini dictation to Gboard, undercutting transcription startups Google is rolling Gemini-powered dictation directly into Gboard, starting with Samsung Galaxy and Google Pixel phones. The native integration removes the install step that companies like Otter, Wispr Flow, and SuperWhisper rely on to acquire mobile users.

• Google puts agentic Gemini and AI-built widgets into Android Google announced Gemini Intelligence for Android, which adds agent actions across apps, Gboard dictation, and form-filling, plus support for widgets generated through natural-language prompts. The release ties Gemini into system-level surfaces rather than a standalone app.

• Vapi reaches $500M valuation after Amazon Ring picks it over 40 competitors Voice AI startup Vapi raised at a $500 million valuation after Amazon's Ring selected its platform for customer-facing calls. Vapi says its enterprise revenue grew tenfold since early 2025 as companies move support and outbound sales to AI agents.

• Meta blocks users from muting its AI account on Threads Meta is testing a Threads feature that lets users tag a Meta AI account inside replies for context or answers, and the account cannot be blocked. The setup mirrors xAI's Grok behavior on X and forces every Threads conversation to allow an AI participant.

• Anthropic tells investors secondary share platforms are selling void stock Anthropic posted a notice warning that any Anthropic stock or interest sold through third-party secondary platforms will not be recognized on its books. The statement targets brokers offering retail-style access to private AI shares as Anthropic's valuation climbs.

• Startups pitch paying homeowners to host mini AI data centers A new wave of companies is offering residents cash to host small AI compute nodes inside their homes, framing it as a way to speed deployment past permitting bottlenecks at hyperscale sites. The model exports zoning, power, and noise tradeoffs from utility regulators to individual households.

• Rivian gates its AI voice assistant behind a $15/month subscription Rivian started rolling out its AI voice assistant to Gen 1 and Gen 2 vehicles via software update. Access requires the Connect Plus cellular plan at $15 per month or $150 per year, putting in-car AI behind a recurring fee rather than bundling it with the vehicle.

• Nvidia engineers detail how they ship production code with OpenAI Codex OpenAI published a case study describing how Nvidia teams use Codex with GPT-5.5 to take research prototypes into production systems and run experiments at scale. The post is OpenAI's most detailed disclosure to date of a chip-vendor customer using its coding agent internally.

• Alibaba releases Qwen-Image-2.0 unifying generation and editing Alibaba's Qwen team published Qwen-Image-2.0, an image foundation model that handles high-resolution generation, multilingual text rendering, and precise editing in one architecture, coupled with Qwen3-VL. The paper targets text-rich and compositional cases where prior open models break down.

Read the full edition →