Episode 14 - AI brain builds 100k-line compiler, now in Excel and PowerPoint

Episode 18 - Governance Beats Speed Copilot Cowork and Google Gemini

Will you let an AI agent act on your data with approvals and an audit trail—or keep it as a brainstormer? Microsoft Copilot Cowork in 365 uses Claude to run tasks across Outlook, Teams, OneDrive, SharePoint, and Excel, with governance baked in; Frontier and E7 previews. Google Gemini Embedding 2 maps text, images, video, audio, and PDFs into one searchable representation, with cross-file drafting in Docs and Sheets. The takeaway: governance and execution determine AI value, not speed.

Episode 17 - Claude Tops US App Store; Test Gemini for Cost Savings

Your biggest decision this week isn’t which feature to ship—it's whether the default AI layer will save you time or cost you rework. Claude hit #1 in the US App Store amid a government-terms dispute, nudging you to draft a one-page AI vendor narrative your board would accept. Google Gemini 3.1 Flash-Lite is a cheaper, high-volume model for small tasks; test taxonomy cleanup and escalate only if it misses. In 2026, winners prove reliability, governance, and cost control at scale.

Episode 16 - Your Budget Wins When Google Gemini 3.1 Handles Ad Renders

Will you let the same system that allocates bids generate and test your ads? Google Ads now lets you run Gemini 3.1 Flash Image inside the platform, delivering 4K renders 25–50% cheaper with watermarking and provenance. Notion gives you Custom Agents to run AI-driven workflows on schedule, with one agent triaging intake. You’ll see AI move from stand-alone tools to core platforms, so your governance and defendable budgets decide who wins.

Episode 15 - Google's Gemini 3.1 Halves Ad Image Costs, in Google Ads

Your next board decision hinges on whether you can run AI-generated ad images inside Google Ads at half the cost. Google rolled out Nano Banana 2—Gemini 3.1 Flash Image—into Ads, delivering 4K images, cleaner text, and SynthID provenance at 25–50% less than pro-tier. Perplexity Computer launches a $200/month digital worker that orchestrates multiple models for multi-step tasks. You’ll walk away seeing AI execute inside your stack, demanding governance like infrastructure, not a creative experiment.

Episode 14 - AI brain builds 100k-line compiler, now in Excel and PowerPoint

You’ll see an AI brain that built a 100k-line compiler in two weeks for $20k now living in Excel and PowerPoint. Anthropic’s Claude upgrade lets you see agents close live tickets and ship complex software with minimal oversight. OpenAI Codex for Mac now lets you run multiple agents on a schedule with memory and a real terminal. You’ll walk away with a repeatable workflow to turn months of numbers into three forecasts and a five-slide AI-drafted deck.