In 60 seconds:
- DeepSeek V4 Pro dropped: 1.6T params, 1M token context, MIT-licensed, ~10× cheaper than the US frontier.
- Microsoft Playwright MCP makes browser agents production-ready — no more screenshot hallucinations.
- Claude × Blender + Autodesk Fusion connectors landed — 3D / CAD generation in plain English.
- Grok Voice Think Fast 1.0 is live at Starlink running real customer support today.
- Google, Microsoft, Meta, Amazon committed $725B in 2026 AI infrastructure capex.
Last week’s first drop got way more love than expected. Thank you! Turns out executives are hungry for signal over noise. Let’s keep going.
This week the frontier kept sprinting. Here’s what you need to know.
🐉 DeepSeek V4 dropped (April 24)
China’s breakout lab just released V4 Pro (1.6 trillion parameters, 1M token context, MIT licensed). Coming scarily close to GPT-5.5 and Claude Opus 4.7 on coding and agentic tasks. Price? $1.74 per million input tokens. That’s roughly 10x cheaper than the US frontier.
🤔 So what: The gap between US frontier models and top open-source ones is now measured in months, not years. Cost of experimentation just collapsed. 10x! Again.
🎭 Microsoft Playwright MCP Server (April)
This one’s nerdy but massive. You know how Claude takes screenshots of websites, clicks, and starts again. Well this fixes it. Agents can read websites in a structured, clean manner with zero ambiguity. No more screenshot-based agents hallucinating clicks.
🤔 So what: Automating Salesforce, internal tools, any browser-heavy workflow just became production-ready. If your team is still copy-pasting between tabs, the efficiency rug just got pulled.
🎨 Claude × Blender + CAD connectors (April 28)
This one made me say it out loud: NO WAY.
Anthropic dropped official MCP connectors for Blender, Autodesk Fusion, Adobe, and more. 9 creative tools total. Describe what you want, Claude writes the Python, controls the software directly, generates 3D models, scenes, CAD geometries. In plain English.
🤔 So what: Prototyping, product design, architectural viz. The barrier just dropped through the floor. Designers who adopt this become 5-10x faster. Anthropic also became a Blender Development Fund patron. They’re not just building tools, they’re funding the ecosystem.
🗣️ Grok Voice Think Fast 1.0 (April 25)
xAI’s new voice agent. Ultra-low latency, reasoning in the background, 25+ languages. But here’s the kicker: it’s already live at Starlink. 1 in 5 inquiries leading to a purchase are handled by AI. 70% of support resolved without a human.
🤔 So what: Voice AI just went from demo to deployment. Customer support, sales, field ops. This isn’t a prototype. Starlink is running it in production. Today.
💰 Big Tech Capex: $725 billion (Earnings week)
Google, Microsoft, Meta, Amazon. Combined AI infrastructure spend for 2026: $725B. Up 77% from last year. Microsoft says they’ll still be capacity-constrained. Google Cloud grew 63% YoY.
🤔 So what: This isn’t hype budget. This is war budget. Treat AI as core leverage or get lapped.
The real gap isn’t between companies that read the news and those that don’t. It’s between leaders who deep dive themselves, understand what it means, and those still waiting for the “right time.”
Bottom line for leaders: Browser agents and voice AI just crossed from “demo” to “deployed.” If your team still has someone copy-pasting between Salesforce tabs or a call centre handling tier-1 support manually, that workflow is already obsolete — you just haven’t met the replacement yet. Want a structured way to lead that shift? See Executive AI Fitness.
Which surprised you the most? DeepSeek’s price collapse? Claude doing CAD? Starlink running voice AI in production? 👇