In partnership with
✨Welcome back, Entrepreneurs
AI is going local, clinical, and compute-heavy. Inside: Gemma 4 offline AI, Anthropic’s TPU deal, OpenAI’s superapp push, faster coding, AI sheets, and healthcare tools that are getting more usable.
🔋Gemma 4 brings offline multimodal AI to everyday devices
Google DeepMind has introduced Gemma 4, featuring 2.3B to 31B parameters, allowing devices like laptops, phones, and Raspberry Pi to run Gemini-grade models offline. This enables fast local inference, such as 40–80 tokens/sec on M-series Macs and real-time translation on iPhones, making on-device AI feasible for mainstream apps.
What’s new
Runs locally on laptops, phones, and Raspberry Pi
Multimodal: text, images, and audio in one model family
2.3B–31B parameter sizes for consumer hardware
Community tests show 40–80 tokens/sec on M‑series Macs
Why this matters for builders
Local AI unlocks privacy‑preserving apps, lower latency, and offline reliability — all without cloud costs. Gemma 4’s multimodal capabilities let developers build assistants, translators, planners, and security tools that run directly on user devices.
Replace your first 4 hires with AI. Free workshop on April 8th.
Most early-stage founders can't afford their first four hires. Sales, marketing, dev, and support alone can run hundreds of thousands in salaries.
On April 8th, AI thought leader Heather Murray shows pre-seed and seed founders how to build all four functions using AI tools. Live, with demos, for free.
Register today and get a free AI tech stack worth $5K+ including Claude, AWS credits, Make, and 90% off HubSpot.
🔥Big AI moves
Image source: Cursor
🔹Cursor debuts Warp Decode on Blackwell
Cursor's enhanced token generation for MoE models on Blackwell GPUs uses warp-level parallelism, reducing overhead and combining operations into two kernels. This results in 1.84× faster inference and outputs 1.4× closer to full-precision references, improving Composer 2, Cursor's top coding model with a 73.7% SWE-bench score.
🔹Sundar Pichai breaks his silence on the AI talent war
Sundar Pichai defends Google's AI leadership on Stripe’s Cheeky Pint podcast, citing DeepMind’s Demis Hassabis. He discusses upcoming memory-supply constraints from Micron and Samsung amid growing competition. The episode, featuring past guests like Elon Musk, releases amid AGI debates and talent battles.
Are you tracking agent views on your docs?
AI agents already outnumber human visitors to your docs — now you can track them.
🧲 AI just turned cardiac MRI into a one-click exam — full story in AIHealthTech Insider
Philips' SmartHeart system automates cardiac MRI planning in under 30 seconds, generating up to 14 views, reducing breath-holds by 75%, standardizing quality, and enabling specialist-level imaging in hospitals without cardiac MRI staff.
This issue covers AI reducing MRI scan times to 9 minutes, instant meal-scanning nutrition apps, AI-powered pantry remedies, and a sugarcane-derived protein protecting enamel after radiotherapy.
Interested in AIHealthTech Insider?
💰Anthropic locks in gigawatts of compute — and bets big on 2027
Anthropic has signed a new agreement with Google and Broadcom for multiple gigawatts of next‑gen TPU capacity coming online in 2027. The expansion supports Claude's frontier-model roadmap and meets rising demand, with run-rate revenue exceeding $30B and $1M+ customers doubling in two months.
Image source: ChatGPT/TheAIEntrepreneurs
What’s new
Multi‑gigawatt TPU commitment with Google + Broadcom
Capacity begins rolling out in 2027 across U.S. sites
Supports Claude’s rapid revenue and enterprise adoption
Deepens partnerships across Google Cloud and Broadcom
Why this matters for builders
Frontier-scale compute is crucial for AI companies. Anthropic employs a multi-cloud strategy using AWS Trainium, Google TPUs, and NVIDIA GPUs for optimized workloads and resilience.
📊 Genspark turns your messy spreadsheets into instant insights
Genspark AI Sheets transforms raw data into charts, summaries, and analysis in minutes. Drop in your CSV or paste a table, and the tool auto‑generates insights, visualizations, and explanations.
How it works:
1️⃣Upload a sheet or paste your dataset directly into Genspark
2️⃣Ask for summaries, trends, anomalies, or visualizations in plain language
3️⃣Generate charts, pivot‑style breakdowns, and column‑level insights
4️⃣Refine with follow‑up prompts to dig deeper into patterns
5️⃣Export results or embed them into your workflow for fast reporting
Top performance-driven ad channel in 2026
"Did this campaign drive that sale, or would it have happened anyway?"
Every marketer asks it. Attribution can't answer it. Incrementality can.
CTV now brings that reporting rigor to television:
✅ Smarter targeting
✅ Proof of incremental lift
✅ Ongoing optimization
Worth a look if you're spending on TV and need to prove it's worth it.
🟣OpenAI's $122B raise just rewired the AI race — read the full breakdown in Sunday Special
OpenAI’s $122B round pushes ChatGPT toward a unified AI superapp — merging coding, browsing, agents, and automation into one interface. With revenue hitting $2B per month and usage exploding across enterprise APIs, OpenAI is positioning ChatGPT as the workflow layer competing directly with Copilot and every agentic system.
In this issue, we cover Anthropic offering free Microsoft 365 connectors for all Claude users, major AI events for builders and operators, and McKinsey's insight that human skills are being reshaped within hybrid human-AI teams.
Would you like to get this exclusive Sunday Special in your inbox?
