xAI just shipped three things at once, and the boring one is the most useful. Grok 4.7 is out, replacing 4.6 with better performance on long coding tasks and complex knowledge work, at identical speed and price. Fine. Nice upgrade. But the thing that actually matters for small teams is Team Bots. Right now if you're at a small agency and you want your whole team using the same Grok Bot, everyone has to go set it up themselves, upload the same files, type in the same context, and hope they did it right. Team Bots fixes that. One person configures the bot once, the whole team shares it, and everyone's individual conversations stay private. Think of it like a shared folder on a network drive, except the folder knows your business and can answer questions. Also bundled in the same announcement: Grok Voice Transcribe 2.0, which doubles accuracy over the first version and adds speaker diarization (meaning it can tell who said what in a multi-person recording) with zero code changes required if you're already using the API. Same price. Existing integrations just get better overnight.

Vercel AI Gateway added four new models this week, and two of them are free right now. If you don't know Vercel, it's a platform developers use to build and deploy web apps, and its AI Gateway acts like a universal plug-in strip for AI models. The new additions include Glyph Cluster, a reasoning model aimed at coding and long-document analysis (up to 256,000 tokens, which is roughly a novel-length document), free for Pro and Enterprise teams during its early release period. Then there's Tako Search, which is the one that should catch your attention: a single API call that pulls in live web search results plus structured data like market prices, company info, and weather, all with citable sources. No stitching together three different services. Free through October 15. Microsoft also added two voice models and a streaming transcription model. The voice models generate expressive speech in 23 languages, and the Flash variant is tuned for low-latency voice agents, meaning it responds fast enough to hold a real conversation. If you've been thinking about building a customer-facing voice assistant, this week is a cheap week to experiment.

Plaud just made meeting recordings actually automatic. Skills and Routines are new inside Plaud Agent, and they solve a specific, annoying problem: you have to explain what you want every single time. Skills let you define a template once, exactly how a summary should be structured, what tone, which sections, how much detail, and the agent follows it every time without being reminded. Routines schedule that skill automatically. Something like "every Monday at 9 AM, summarize last week's meetings and list the open items" just runs. Your existing transcription plan and pricing stay the same. This is the kind of thing that sounds minor until you realize you've been spending 20 minutes every Monday doing that manually.

Tools Worth a Look

  • Grok Team Bots: Share one configured Grok Bot across your whole team, same files and context, with private individual chats. Built for small agencies tired of everyone setting up AI from scratch.

  • Grok Voice Transcribe 2.0: Speech-to-text API, 2x the accuracy of v1.0, speaker diarization included, existing integrations upgrade automatically. Same price as before.

  • Tako Search on Vercel AI Gateway: One API call returns live web search plus structured market and company data with sources. Free through October 15.

  • MAI-Voice-2.1 / Flash (Microsoft on Vercel): Expressive text-to-speech in 23 languages; Flash variant for voice agents that need fast response times.

  • Glyph Cluster on Vercel AI Gateway: Reasoning model for coding and long-context analysis, up to 256K tokens, free for Pro and Enterprise teams during its stealth period.

  • cathrynlavery/diagram-design: 42 ready-made diagram templates for AI coding tools like Claude Code and GitHub Copilot, self-contained HTML and SVG files, no external dependencies. (825 stars today)

  • addyosmani/agent-skills: A collection of production-tested engineering skills you can drop into AI coding agents so they handle real dev tasks reliably. (677 stars today)

  • trycua/cua: Open-source tools for running AI agents that control computers, across multiple operating systems, with benchmarks for testing how well they perform. (228 stars today)

  • EpicGames/raddebugger: A graphical debugging tool from Epic Games for digging into what's happening inside running software. (90 stars today)

  • manaflow-ai/cmux: A Mac terminal app with vertical tabs and built-in notifications designed for managing multiple AI coding agents at the same time. (44 stars today)

Also Worth Knowing

-Scott

SmartOwner is published (almost) daily by the team at DigitalTreehouse. Want AI consulting or automations for your business? Reply to this email.