Oh, So AI now runs your entire enterprise? 🏢 - Edition 006

Today: Microsoft rolls out Agent 365 and the E7 Frontier Suite today, giving enterprises a real control panel for every AI agent in their stack. OpenAI tells all macOS users to update before May 8 after a supply chain hack. And AI Overviews now appear in nearly half of all Google searches.

💬 Daily Quote

“The work you keep skipping is usually the work worth doing first.”

📌 TL;DR

Today: Microsoft launches Agent 365 and E7 today, the first new enterprise license tier since 2015. OpenAI forces a macOS update after a third-party supply chain hack, deadline May 8. Fresh data shows AI Overviews now appear in 48% of Google searches, up from 35% just four months ago.

🏆 LLM Leaderboard #AI #Models

The most popular models this week.

GPT-5.5 still owns the top of the intelligence chart, but almost nobody uses it as their daily model. Xiaomi’s MiMo-V2-Pro leads real usage with 4.65 trillion tokens a week on OpenRouter. Chinese models now account for over 45% of total traffic on the platform. Meta Muse Spark enters the benchmark chart this week at 52, fresh off its April launch. The gap between what scores highest and what developers actually reach for has never been wider.

Source: OpenRouter Rankings + Artificial Analysis Benchmarks

📰 Oh, So AI did this #AI #News

1. Microsoft launches Agent 365 and E7 today. Enterprise AI governance is now a product.

Microsoft’s first new enterprise license tier since E5 in 2015 goes live today. E7 costs $99 per user per month and bundles Microsoft 365 E5, Copilot, Entra Suite, and the brand-new Agent 365 into one SKU. Agent 365 is the part that matters most. It is a control plane that lets IT teams deploy, monitor, and set guardrails for every AI agent running across the org. Not just Copilot. Any connected agent. Per Microsoft Blog. Before today, enterprise AI agent governance was a patchwork of hope and manual checks. Now it is a licensed product with a dashboard.

Read more at Microsoft Blog #AI #News

2. OpenAI tells all macOS users: update your apps before May 8 or they stop working.

A widely used JavaScript library called Axios was hacked in a supply chain attack back in March. A malicious version of Axios got into OpenAI’s macOS code-signing workflow and gave attackers a potential path to app certificates. OpenAI says no user data was accessed and no systems were breached. But it is revoking old certificates and forcing everyone to update. ChatGPT Desktop, Codex, and Atlas all need to be updated before May 8. Per AI Magazine and BleepingComputer. If you use any of these on a Mac, open the App Store and update right now.

Read more at AI Magazine #AI #News

3. Novo Nordisk wires OpenAI into its entire drug pipeline.

The pharma company behind Ozempic is partnering with OpenAI to integrate AI across its full business. Drug discovery, clinical trials, manufacturing, supply chains, and commercial operations. Full deployment is planned by end of 2026. Per Reuters. This is the largest announced use of OpenAI in a regulated industry so far, and it shows clearly where this technology is heading next.

Read more at Reuters #AI #News

4. Meta drops its first proprietary model, breaking a decade of open-source tradition.

Meta Superintelligence Labs, led by Alexandr Wang, launched Muse Spark this month. It is the first model Meta has built that is not open-source. Muse Spark scores 52 on the Artificial Analysis Intelligence Index and has a Contemplating mode that runs multiple agents in parallel for harder reasoning tasks. The company that gave developers Llama for free now says it “hopes to open-source future versions.” That is a very different thing from actually doing it. Per TechCrunch. For anyone who built workflows on Llama as a free foundation, this is a shift worth tracking.

Read more at TechCrunch #AI #News

🪄 Oh, So AI can do that?! #AI #Tools

1. Microsoft Agent 365 is available as a $15/month standalone add-on today.

You do not need the full E7 bundle to get Agent 365. Microsoft is selling it separately from today at $15 per user per month. Plug it into your existing Microsoft 365 environment and you get a dashboard to see, manage, and set limits on every AI agent connected to your stack. If your team is already running Copilot agents or third-party bots inside Teams or SharePoint, this is the visibility layer that makes that manageable. Per TrustMarque.

Read more at TrustMarque #AI #Tools

2. Gemini 3.1 Flash Lite is now the cheapest serious multimodal model on the market.

Listed on OpenRouter this week, Gemini 3.1 Flash Lite costs $0.25 per million input tokens and $1.50 per million output tokens. Multimodal by default. That puts it directly against Claude Haiku and GPT-4o Mini for high-volume, cost-sensitive production apps. If you build features where cost per API call is a real budget line, this one is worth a benchmark against your current setup. Per OpenRouter.

Read more at OpenRouter #AI #Tools

3. Meta Muse Spark is now available via API with built-in parallel agent reasoning.

Meta’s new proprietary model is live for developers. The Contemplating mode runs multiple agents at the same time on complex problems, similar to GPT Pro and Gemini Deep Think. It scores 52 on the Artificial Analysis Intelligence Index, above most mid-tier models. Works with standard API formats, so no full rewrite needed if you are switching from Llama. Per VentureBeat.

Read more at VentureBeat #AI #Tools

⚡ Oh, So I can do this #SEO #WebDev

1. AI Overviews now appear in 48% of all Google searches.

New data from March 2026: Google’s AI-generated answer boxes now show up in nearly half of all searches, up from 34.5% in December 2025. That is a 40% jump in four months. Informational queries trigger them 39.4% of the time. Per Stackmatix. If your traffic strategy is built around owning organic positions, you are now playing in a world where half of all searches never scroll to a result.

Read more at Stackmatix #SEO

2. Only 38% of pages cited in AI Overviews actually rank in the top 10.

New Ahrefs data from February 2026: just 38% of pages cited in AI Overviews also rank in the top 10 for the same query. Seven months earlier that number was 76%. Google is pulling AI answers from pages it trusts, separate from its regular ranking signals. Structured content, original data, and expert takes are what get picked up. Per Digital Applied. Ranking on page one and getting cited in an AI answer are now two different games, and you need to play both.

Read more at Digital Applied #SEO

3. Native CSS in 2026 is doing things that used to need JavaScript.

New CSS features landing in major browsers this year handle custom dropdowns, scroll animations, carousels, and interactive transitions without any JS. The fully customizable select element is here. Scroll-driven animations are stable. Native CSS nesting is in every major browser. Per LogRocket. If you are still reaching for a library every time you need a hover effect or a modal, native CSS might already handle it.

Read more at LogRocket #WebDev

🎣 Top Hook Ideas

Hook 1: Microsoft just launched a control panel for every AI agent in your company. Set limits. Monitor activity. Revoke access. Enterprise AI governance is officially a product now. #AI

Hook 2: OpenAI just told every Mac user: update your apps by May 8 or they stop working. A supply chain hack hit their code-signing workflow. Here’s exactly what you need to do. #AI

Hook 3: Google’s AI Overviews now appear in 48% of all searches. Four months ago it was 35%. Your SEO strategy from last year is already running on old information. #SEO

Hook 4: Meta just broke a decade of open-source tradition with its first proprietary model. The company that gave developers free AI for years is going closed. Plan accordingly. #AI

Hook 5: 38% of pages cited inside Google AI Overviews don’t even rank in the top 10. You don’t need page one anymore. You just need content Google trusts. #SEO

Curated with ❤️ ByHarshal

Creative Director orchestrating AI workflows for founders' teams. Writing about productivity, design, and AI systems.