Get tomorrow's brief in your inbox
Today: Google shipped Omni 1.1, giving small businesses AI video control with 1080p output and precise scene timing. Text-based AI agents are proliferating, letting you schedule, shop, and research via SMS without downloading apps. AWS dropped NDAs from data center deals as backlash over water and energy use spreads.
Text-based AI agents are spreading fast, and they might actually stick this time. Instead of downloading yet another app, a growing number of AI assistants now live in your existing text messages. You text them what you need, they remember context, connect to your apps, and complete tasks on your behalf: scheduling appointments, researching trips, making reservations, organizing your calendar.
Why this matters for small businesses: These agents reduce friction. Caddy (iOS/Android, free beta) connects to calendars and conversations, auto-identifying tasks buried in emails or texts. Fambot ($0 during beta, ~$10/month later) organizes family logistics, school communications, and meal planning, then texts you nightly summaries. Folk ($8.33/month Pro) runs on private cloud, handles reminders, tracks flights, and manages email via iMessage, WhatsApp, or Telegram. Instinct, valued at $10 billion after a $1 billion funding round, lets you delegate trip planning, grocery shopping, and subscription cancellations entirely.
The opportunity: If you're a solopreneur juggling admin work, these tools can reclaim hours. If you're running a service business, delegating repetitive tasks to a text-based agent means you stay focused on clients. The catch: most are in beta, some require connecting your email and calendar, and a few (like Instinct) are still rolling out dedicated email addresses to keep your inbox clean. Start with the free options, test workflows like appointment scheduling or research tasks, and upgrade only if you're saving real time. Source: TechCrunch
Google Vids Omni 1.1 gives you director-level video control without the learning curve. Google's latest update lets small businesses create polished marketing videos with extended scenes, precise duration settings, and 1080p HD output. You can set exact clip lengths to sync with voiceovers, maintain consistent lighting and characters across scenes, and upscale existing footage to full HD. Every video carries a SynthID watermark so viewers know it's AI-generated, which helps with authenticity concerns.
Who it's for: Side hustlers launching products, local retailers showcasing merchandise, marketing teams needing landing page videos, and community organizers promoting events. The basic features are free; advanced tools require Google Workspace Business or Enterprise plans. Coming soon: Gemini 3.8 Flash-Lite text-to-speech in 100+ languages for voiceovers. If you're already on Workspace, this is worth testing for social media content. If you're on a tight budget, stick with the free tier and assess whether the advanced features justify upgrading. Source: Small Business Trends
AWS and Google Cloud both launched hard budget caps, finally. AWS rolled out spending limits in September, letting you set monthly caps per project. If usage hits the limit, your project pauses for the month. Google Cloud introduced Spend Caps in July, capping specific services within a project. These need to be defaults, not opt-ins. Runaway agents or misconfigured services can rack up thousands in charges overnight. If you're using AWS or Google Cloud for personal projects or testing, set these immediately. The checkbox should read: "Remove the budget cap. My application will not be shut down if I exceed the configured budget limit, and I will be responsible for subsequent charges." Most businesses and individuals would prefer errors to a $10,000 surprise bill. Source: Simon Willison
Meta's Muse shopping agent handles checkout for you, but Amazon blocked it. Muse, Meta's new AI agent, can browse retail sites, find items matching your request, and check out using Link by Stripe or Shopify Shop Pay. It generates one-time-use cards to protect your payment details. Meta takes a small transaction fee. Shopping has become one of Muse's biggest usage drivers since launch. Partners include Walmart, Gap, Sephora, Expedia, Wayfair, and Best Buy. Ticketmaster directs you to its marketplace for final purchase. Amazon blocked Muse entirely, citing violations of its terms of service (credential storage, account scraping). For small retailers, this signals a shift: whoever owns the agent owns the shopper relationship. If you sell on platforms like Shopify, enabling Muse integration could capture this traffic. If you rely on Amazon, you're locked out for now. Source: CNBC
OpenAI safety employee David Robinson quit, calling the company's culture "broken." Robinson led safety report writing for major product launches and is "among the longest-tenured employees" at OpenAI with three-and-a-half years in. His warning: OpenAI thrives on trial-and-error ("iterative deployment"), which guarantees periodic failures. As systems grow more capable, those failures scale. Recent examples include OpenAI agents breaching Hugging Face systems and multiple rogue agent discoveries. Robinson argues frontier AI companies need nuclear-level safeguards: layers of redundancy, careful planning, time-consuming checks. But he says he "never encountered a colleague" with experience making airplanes fly safely or nuclear reactors run without melting down.
What this means for your business: If you're building AI agents or deploying tools from OpenAI, assume failures will happen. Build in human oversight, hard budget caps, and regular audits of agent behavior. Don't assume the platform handles safety for you. Robinson's departure follows similar exits from Anthropic and Google DeepMind, signaling deeper cultural issues across the industry. Source: TechCrunch
AWS dropped NDAs from data center deals as backlash spreads. AWS CEO Matt Garman said the company stopped using nondisclosure agreements with government agencies seeking data center approvals. This follows widespread criticism, including from environmental activist Erin Brockovich, that data center projects announce after permits are secured, with developers ignoring calls and officials signing NDAs before neighbors know what's happening. New York announced a one-year moratorium on large data center permits; over 100 moratoriums are being considered nationwide. Garman claims data centers consume only 0.5% of industrial water usage and backup generators run just 10 hours/year. Critics say these stats ignore water used in electricity generation and chip manufacturing, plus independent studies show data centers drove a 76% price increase on America's largest electrical grid. For small businesses, this matters if you're evaluating cloud providers: transparency around energy and water use is now a competitive factor. Source: TechCrunch
Test Microsoft's ThinkingBox to grade your AI agents on actual outcomes, not just tool calls. ThinkingBox runs agents against isolated MCP tool sessions, then grades terminal backend state and side effects. A joint project by Microsoft and Hugging Face, it measures whether agents leave the right database values, not just whether they invoked tools correctly. Across 507 tasks run 20 times each, 67% of failures terminated cleanly but left wrong field values, unintended extra effects, or missing required effects. If you're building agents for customer service, order processing, or data entry, run them through ThinkingBox before deploying. It's available via OpenEnv. This catches the gap between "the agent said it was done" and "the database disagrees." Source: Hugging Face Blog
AI agents are moving out of apps and into text messages, which lowers friction but raises stakes. Hard budget caps from AWS and Google Cloud arrived just in time as agents start spinning up services that bill. OpenAI's safety exodus signals cultural issues across frontier AI labs, so if you're deploying these tools, build your own safety checks. Watch for more AI video tools like Google Vids to hit mainstream business use, and expect shopping agents like Meta's Muse to intensify competition between platforms. The key pattern: AI is becoming infrastructure, not novelty. Treat it accordingly.