In Today’s Issue:

💻 Google is bringing highly capable multimodal edge AI

⚠️ The Iran conflict escalates drastically as the IRGC targets commercial cloud data centers

⏳ The authors of the pivotal AI 2027 forecast is predicting expert-level AGI is closer

🚀 Cursor 3 is signaling a massive shift in software development

✨ And more AI goodness…

Dear Readers,

Google just dropped Gemma 4, and the real story isn't the benchmarks, it's that genuinely capable multimodal AI now runs on a $200 phone with no cloud, no API, no data leaving your pocket. But while the industry celebrates edge AI liberation, the infrastructure powering everything else is literally under fire: Iran has begun targeting commercial data centers in the Gulf, a third of global helium supply is offline, and the Strait of Hormuz is choking oil flows that keep the lights on in every major cloud region.

Meanwhile, the authors behind the AI 2027 forecast just moved their timelines forward by over a year, Sam Altman is hinting that something "very big" is about to happen with next-gen models, and Cursor 3 is betting that developers will soon supervise fleets of coding agents rather than write code themselves.

Today's issue connects the dots between on-device breakthroughs, geopolitical chokepoints, and accelerating timelines, and honestly, the picture that emerges deserves your full attention.

All the best,

Kim Isenberg

Cursor 3 introduces a unified, agent-first workspace where multiple AI agents can collaboratively handle coding tasks across repositories, cloud, and local environments. The new interface boosts productivity with parallel agent execution, seamless environment switching, and integrated tools like browser testing and plugin marketplaces.

This marks (another) shift toward autonomous software development, where engineers supervise fleets of agents instead of writing every line, hinting at faster workflows and a future of largely self-operating codebases.

Claude can now directly control keyboard and mouse on Windows, enabling it to operate legacy apps like a human and dramatically expand real-world automation.

The Power and Responsibility of Sam Altman at Mostly Human

‘We have a few times in our history realized something really important is working, or about to work so well, that we have to stop a bunch of other projects. In fact, this was the original thing that happened with GPT3. We had a whole portfolio of bets at the time. A lot of them were working well. We shut down many projects that were working well, like robotics which we mentioned, so that we could concentrate our compute, our researchers, our effort into this thing that we said "okay there's a very important thing happening." I did not expect 3 or 6 months ago to be at this point we're at now; where something very big and important is about to happen again with this next generation of models and the agents they can power.'

He goes on to imply there may be a possible future relationship with Disney, then finishes up with:

'we need to concentrate our compute and our product capacity into these next generation of automated researchers and companies.' (Sam Altman in his recent interview with ‘Mostly Human’)

Google just released Gemma 4, and the most interesting part isn't the big models. It's the small ones. The E2B and E4B variants are purpose-built to run powerful AI directly on phones, Raspberry Pis, and IoT devices, completely offline and with near-zero latency. The E4B packs 4.5 billion effective parameters into a footprint that handles text, images, video, and native audio input, all within a 128K context window. The new models are up to 4x faster than previous versions and use up to 60% less battery.

It means developers can build real agentic workflows, think multi-step planning, function calling, speech recognition, on hardware people already carry in their pockets. No cloud dependency, no API costs, no data leaving the device. Google worked closely with Qualcomm, MediaTek, and its own Pixel team to optimize these models for on-device deployment.

And for the first time, the entire Gemma family ships under an Apache 2.0 license, giving developers full commercial freedom to fine-tune, modify, and deploy without restrictions. The weights are already live on Hugging Face. When capable AI runs locally on a $200 phone, the implications for privacy, accessibility, and global reach become hard to ignore.

Why it matters: On-device AI at this level of capability removes the cloud as a bottleneck for billions of users worldwide. It hands developers, startups, and enterprises real sovereignty over their AI stack, their data, their infrastructure, their rules.

Sources:
🔗 https://huggingface.co/google/gemma-4-E4B

🔗 https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/

Viktor is an AI coworker that lives in Slack, right where your team already works.

Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."

Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.

There’s no new software to adopt and no one to train.

Most teams start with one task. Within a week, Viktor is handling half of their ops.

Add Viktor to Slack for free.

“Built from the same world-class research and technology as Gemini 3, Gemma 4 is the most capable model family you can run on your hardware. They complement our Gemini models, giving developers the industry's most powerful combination of both open and proprietary tools.”

This article does not intend to take sides in the ongoing conflict, our goal is to provide analytical, fact-based reporting on how these geopolitical events impact the AI and technology industry.

The AI revolution runs on electricity, silicon, and helium, and right now, all three are under fire. Literally. On April 2, Iran's IRGC claimed strikes on an Oracle data center (see #1 news) in Dubai and an Amazon cloud facility in Bahrain, marking the first time in history that commercial data centers have been deliberately targeted in a military conflict. Before dawn on March 1, Iranian Shahed drones had already struck two AWS data centers in the UAE. Now, the IRGC has explicitly named Nvidia, Microsoft, Google, Apple, Meta, and Oracle as potential targets, calling them complicit in strikes on Iran.

But the physical attacks are only half the story. Over 1,000 miles south of those burning server racks, Qatar's Ras Laffan complex, the world's largest helium production hub, remains offline since March 2. Attacked by Iranian drones on March 2 and again by ballistic missiles on March 18, the facility suffered extensive structural damage and currently has no plans to restart. This is important because helium is irreplaceable in semiconductor manufacturing: it cools EUV lithography machines, detects leaks in vacuum systems, and maintains the ultra-precise conditions needed to print chips at 3 nanometers. The shutdown removed roughly a third of global helium supply from the market in a matter of hours. South Korea, home to Samsung and SK Hynix, sources nearly 65% of its helium from Qatar. Helium spot prices have surged up to 100% since the strikes.

Meanwhile, the Strait of Hormuz, through which roughly 20% of the world's oil and LNG normally flows, remains effectively shut. In total, 27 commercial ships have been attacked or reported incidents since March 1. Oil prices hover near $116 per barrel, with analysts warning of $150 or even $200 if the blockade continues. Executives and analysts warn the strait needs to reopen by mid-April, or oil supply disruptions will get significantly worse. Data centers, which consume enormous amounts of energy, face a double hit: rising electricity costs and shrinking chip supply.

The AI industry has long debated bottlenecks around GPU shortages and power grid capacity. Now, the Iran war is proving that the real chokepoint might be geography itself: a narrow strait, a single gas complex, and a cluster of server farms in the Gulf that nobody thought would become military targets. If this conflict drags on, the ripple effects on AI hardware timelines, cloud costs, and chip production could be felt for quarters and perhaps, years to come.

Sources:

🔗 https://www.cnbc.com/2026/04/01/iran-irgc-nvidia-appple-attack-threat.html

🔗 https://theconversation.com/why-iran-targeted-amazon-data-centers-and-what-that-does-and-doesnt-change-about-warfare-278642

🔗 https://www.taiwannews.com.tw/news/6332336

🔗 https://carraglobe.com/semiconductor-supply-chain-disruption-2026/

Knowing when to retire means knowing what it costs, how long your money needs to last, and where the income comes from. When to Retire: A Quick and Easy Planning Guide helps investors with $1,000,000 or more work through all of it.

Download your free guide.