{{first_name | Leader}}, welcome back. Today’s updates:
Jace drafts, labels, and follows up so you don’t have to*
Microsoft to buy AI from Anthropic
Google concedes the “open web” is in rapid decline
Microsoft locks $17.4B Nebius deal for GPU cloud
Practical tools + must-read resources. Dive in ⬇️
Spend less time in Gmail as Jace drafts replies, files messages, and adds labels for you.
You skim, hit Send, and get back to work.
What you get:
✅ 5-second email reviews instead of 10-minute replies
✅ Automatic CCs, follow-ups, and labels
✅ Cancel anytime; 30-day money-back guarantee if you upgrade

Microsoft to Buy AI From Anthropic in Partial Shift From OpenAI
Microsoft licensed Anthropic’s Claude models for Azure and Microsoft 365 Copilot under a new multi-year pact. The move adds a second frontier-AI supplier, easing dependence on OpenAI. Claude launches in Copilot Studio later this quarter with transparent model labels. Financial terms were undisclosed publicly.
Dual-sourcing now gives Microsoft pricing leverage before FY-25 contract renewals. Internal tests show Claude Sonnet 4 matches GPT-4o accuracy at lower latency. Anthropic’s constitutional-AI guardrails strengthen compliance for regulated workloads. Ask Copilot owners to benchmark cost, speed, and safeguards this month.
Google Concedes the “open web” is in Rapid Decline
Google reported to regulators that news links visible to its crawler dropped 25 % YoY. Search will now boost live feeds and pay publishers for guaranteed access. A real-time article-sync API opens in the fall, letting partners push updates directly. Google has not disclosed pricing tiers.
A shrinking open web raises training costs for in-house LLMs and marketing analytics. Scraping-based analytics risk blind spots as sources migrate behind APIs. Legal exposure rises if models ingest unlicensed text. Allocate budget for licensed feeds or audit pipelines before the API opens.
Microsoft Locks $17.4 B Nebius Deal for GPU Cloud
Microsoft signed a five-year, $17.4 billion deal with Nebius to add 400 MW of AI-ready capacity across Europe. Phase 1 is underway in Hamina, Finland, installing liquid-cooled Nvidia H100 clusters. Both firms pledge 100 % renewable power by 2028. First capacity comes online mid-2026, with new Azure-aligned regions following.
The build eases Europe’s GPU crunch and could trim compute prices by 10–15 %. Microsoft customers’ training models get priority reservations. Early-access sign-ups for Nebius regions open this quarter. Apply now and secure energy-indexed pricing before general availability
Productivity Tools

Liquid.ai: Run private, on-device LLMs on phones & edge hardware—no GPU cluster needed.

Huru: Practice unlimited mock interviews; get real-time AI scoring and improvement tips.

Nerdio: Slash Azure spend and auto-scale secure Windows 365 desktops with one dashboard.

Genei: Summarize research papers instantly and outline structured reports with AI assistance.
Get featured tomorrow: How do you use AI for business/personally? Interesting stories will be shared with 100K curious readers.
Useful Resources
Craft a real-estate business plan with ClickUp’s AI outline (link)
Deploy revenue-ready AI agents using Zapier’s enterprise tutorial (link)
Steal a 30-day CMO onboarding blueprint from SaaStr’s playbook (link)
Claude can now turn chats into PDFs, slides, and sheets in seconds (link)
Learn how technical teams ditch cold calls for product-led growth at Anthropic & Cursor (link)
💰 Funding
Orchard Robotics secured $22M to scale precision-agriculture vision systems.
Vouched raised $17M to automate identity and agent verification.
💼 Who’s Hiring in AI
🐦 Automate Your Work
This guy literally automated 73% of his job, told his boss, and got promoted
— #Aadit Sheth (#@aaditsh)
4:01 PM • Sep 9, 2025
Designing a Sales Proposal Win Blueprint
You are an expert sales proposal consultant with extensive experience across multiple industries. Your task is to help improve a sales proposal by providing specific, actionable feedback and recommendations.
First, please gather the following information from the user:
1. What are you selling? (product, service, software, consulting, etc.)
2. Who is your target audience? (individual client, small business, enterprise, government, nonprofit, etc.)
3. Current state of your proposal:
- Do you have a draft that needs improvement?
- Are you starting from scratch?
- If you have a draft, please share the key sections or paste the content
4. What's your biggest challenge or concern? (structure, persuasiveness, pricing strategy, technical details, addressing objections, closing techniques, etc.)
5. Proposal format and length: (brief pitch deck, formal RFP response, detailed written proposal, presentation, etc.)
6. Timeline: When do you need to submit this proposal?
7. Competition: Are you aware of other vendors competing for this business?
8. Budget range: What's the approximate value of this deal?
Once you have this information, provide comprehensive guidance that includes:
- Structure optimization: Recommend the most effective organization and flow
- Content enhancement: Suggest improvements to messaging, value propositions, and key points
- Persuasion techniques: Identify opportunities to strengthen emotional appeal and logical arguments
- Risk mitigation: Help address potential client concerns or objections proactively
- Differentiation strategy: Ways to stand out from competitors
- Call-to-action optimization: Ensure clear next steps and decision-making guidance
- Professional presentation tips: Formatting, visual elements, and delivery recommendations
Provide specific examples and templates where helpful, and ask follow-up questions if you need clarification on any aspect of their situation.
P.S. Get more such prompts in the Prompting Playbook (free for you)
Q: Which research task is least fit for AI automation?

Correct Answer: Generating hypotheses
Reasoning: Generating hypotheses is least suited to AI automation because crafting novel, testable ideas still relies on human intuition and domain-specific creativity, which current AI lacks.
Stay curious, {{first_name | leaders}}
PS. If you missed yesterday’s issue, you can find it here.