The government is buying AI by the token. It should buy results.

fedscoop.com Oct 7, 2026

The government is buying AI by the token. It should buy results.

Curated by Shubham · TechCapsules · Oct 7, 2026

AI-summarised brief · reviewed before publication

The U.S. government has begun purchasing ChatGPT through a token‑based model under a new OneGov agreement between the General Services Administration (GSA) and OpenAI. The contract replaces the prior $1‑per‑month fee with a 50% discount for consumption‑based payments and will run for 27 months. It is projected to grant access to the AI platform for roughly 23 million public servants nationwide. The shift aligns with executive order 14402, which favors fixed‑price contracts with performance metrics, yet the token model complicates cost‑effectiveness measurement. The agreement reflects a broader trend toward integrating AI into federal operations while grappling with procurement policy constraints.

💡 Why It Matters

· The token pricing signals a pivot toward usage‑driven AI adoption, challenging traditional government cost controls and raising questions about accountability and value delivery.

AI Capsules Policies

Read full article on fedscoop.com

SpaceX Stocks Keep Rebounding. Elon Musk is a Trillionaire Again.

Elon Musk regained his status as the world’s sole trillionaire after Tesla and SpaceX shares surged. Tesla’s third‑quarter deliveries reached 486,532 vehicles, surpassing the 461,100 forecast by analysts and its own consensus of 461,974, a 5.3% beat, though still 2.1% below the 2025 record of 497,099. The Model 3 and Model Y accounted for 98% of sales. SpaceX stock climbed over 17%, topping $174 per share, buoyed by expectations that its AI‑compute leasing business will add a new revenue stream.

💡 The rebound signals investor confidence in Musk’s dual ventures, positioning SpaceX’s AI services as a pivotal growth engine beyond rockets and satellites.

9to5google.com · Oct 7

Claude launches Google Docs, Sheets, and Slides integration with sidebar and more

Anthropic has released a public‑beta add‑on that embeds Claude into Google Docs, Sheets, and Slides via a sidebar. The integration lets users correct spelling, rewrite text, and restyle sections in Docs without altering formatting; generate formulas or tables in Sheets; and create or update slides in Slides using existing layouts. Users can preview edits before they are applied and can also link Google Workspace files directly into Claude conversations. The add‑on is available to paid Claude subscribers through the Google Workspace marketplace.

💡 By enabling Claude inside Google’s core productivity tools, Anthropic expands its AI reach beyond its own ecosystem, offering a competitive alternative for users who prefer Claude over Gemini.

bangordailynews.com · Oct 7

Thomas College awarded $3 million grant to build artificial intelligence integration across…

Thomas College in Waterville received a $3 million Title III Strengthening Institutions Grant to embed artificial intelligence across all academic programs. The multi‑year initiative will hire full‑time staff, including an AI specialist in residence, to train faculty, integrate AI into curricula, expand applied learning, and establish a dedicated design space for fabrication, robotics, and data acquisition. The grant aims to ensure every graduate gains practical, critical, and ethical AI literacy, enabling the college to launch new majors and support local businesses with AI‑savvy talent.

💡 By equipping students with hands‑on AI skills, the grant directly strengthens Central Maine’s workforce, positioning the region to compete in a technology‑driven economy.

cio.com · Oct 7

SAP advances its autonomous AI agenda with yet more Joule Assistants and…

SAP unveiled a suite of new AI agents and assistants at its SAP Connect event, expanding the scope of its Joule platform. The launch, part of the Autonomous Enterprise vision, introduces Joule Work, Joule Studio, and autonomous agents that integrate with SAP’s Business AI Platform. These tools are now available as managed products through the SAP for Me portal, with general availability starting this month. SAP highlighted productivity gains—20% in finance, HR, and procurement—and showcased automation in revenue recognition, supply‑chain disruptions, and customer experience. The update also deepens integration across SAP S/4HANA Cloud Public Edition, Ariba, Sales Cloud, and other solutions, while on‑premises Cloud ERP receives AI enhancements via the Early Adopter Care program.

💡 By embedding autonomous agents across core business functions, SAP positions itself to accelerate digital transformation for enterprises, offering a unified, AI‑powered interface that can shift processes from semi‑autonomous to fully autonomous. This move could reshape how large organizations manage finance, supply chain, and HR, potentially setting new industry standards for AI‑driven operational efficiency.

qz.com · Oct 7

10 jobs AI is already doing — and 5 it still cannot…

AI has already automated 10 distinct roles across industries, from customer‑service chatbots that resolve routine tickets to software that triages medical imaging, prioritizes shipping queries, and drafts earnings releases. These systems perform tasks that rely on pattern recognition, language generation, and structured data, allowing human workers to focus on higher‑level decision making. The remaining five roles—surgeons, licensed counselors, electricians, corporate officers, and legal professionals—remain human‑led because they require physical dexterity, emotional judgment, or legal accountability that current AI cannot provide.

💡 The split underscores that AI’s reach is bounded by liability and the need for human responsibility, shaping how businesses allocate talent and invest in automation.

therobotreport.com · Oct 7

Pegasus 1.6 brings video understanding to physical AI, says TwelveLabs

TwelveLabs released Pegasus 1.6, a video‑understanding model that adds temporal context, spatial reasoning, and task completion judgment to egocentric footage. The update targets physical AI, enabling robots, drones, and autonomous vehicles to interpret real‑world actions from a first‑person perspective without specialized cameras. Pegasus 1.6 supports five workflows—action segmentation, dense captioning, quality scoring, search, and curation—allowing developers to convert raw video into structured, reviewable training data. The company claims the model speeds up model training and improves safety in physical environments.

💡 By turning everyday egocentric video into actionable knowledge, Pegasus 1.6 lowers the barrier for robotics teams to train on realistic human behavior, accelerating deployment of safer autonomous systems.