On Lex Fridman's podcast, Jensen Huang defined AGI as AI agents that can create autonomous economic value and declared that threshold has been crossed. He cited OpenClaw as evidence and immediately qualified: the probability current agents can build a company the size of Nvidia is "zero percent."

So what? When the world's most valuable semiconductor CEO defines AGI by economic output rather than human-level intelligence, the industry is converging on a market-declared standard, not a researcher-declared one. For anyone building AI agents today: Huang's definition tells you more about where compute investment is heading than any benchmark score.

Similarweb data cited by Forbes: Claude recorded a 1,487% increase in sessions in March 2026. The surge tracked alongside Anthropic's launch of Computer Use and continued coding improvements. ChatGPT's market share dropped from 86.7% in early 2025 to 64.5% today.

So what? The AI assistant market is not locked in. User behavior can shift by an order of magnitude in a single month when a product delivers meaningfully better capability for a specific workflow. For builders and operators: Claude's March surge is a signal to re-evaluate which model actually fits your stack today, not the benchmark you ran six months ago.

OpenAI is pitching private equity firms a guaranteed minimum return of 17.5% to form joint ventures distributing AI to enterprise clients, targeting roughly $4B in capital. Terms include early model access and significantly exceed standard preferred instruments. OpenAI is competing directly with Anthropic, also courting PE firms, while its latest IPO documents name dependency on Microsoft as a key risk.

So what? This is no longer just a model race. Whoever locks in PE firms gains a fast lane into thousands of portfolio companies that would never adopt AI independently. For operators on either platform: a wave of PE-driven enterprise deployment is coming at a scale that organic go-to-market cannot match.

Anthropic launched Computer Use in research preview for Claude Pro and Max subscribers on macOS. Claude can click, type, and navigate apps on your Mac while you assign tasks from your phone via Dispatch. This is the first product shipped by the Vercept team, acquired by Anthropic in February, four weeks after the acquisition closed. Windows support is in the pipeline.

Sam Altman's stated North Star: an AI intern running autonomously by September 2026, a full multi-agent research system by March 2028. Chief Scientist Jakub Pachocki described a system that sets its own hypotheses, runs experiments, and writes results with minimal supervision. OpenAI has committed $1.4T in infrastructure to support this roadmap.

Apple confirmed WWDC 2026 focused on AI advancements. Per TechCrunch and MacRumors, Siri will be rebuilt with Google Gemini as the foundation, fully white-labeled with no visible Google branding. Apple chose Gemini over OpenAI for its long-term AI strategy across one billion-plus devices.

The largest study yet on AI's cognitive impact: 1,372 participants, 9,593 trials. Users followed AI answers 79.8% of the time even when the AI was wrong, dropping their accuracy to 31.5%, below baseline with no AI at all. Confidence increased 11.7 percentage points regardless of correctness. Researchers call this "cognitive surrender." The same week, Senator Bernie Sanders' interview with Claude drew 4.4M views, showing Claude visibly shifting its answers based on how questions were framed.

Microsoft recruited Ali Farhadi, CEO of Ai2 (the world's largest AI nonprofit, focused on open-source research), to join Mustafa Suleyman's AI research team. The hire signals Microsoft building independent AI research capacity to reduce reliance on OpenAI ahead of OpenAI's IPO.

Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:

Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.

Download the Free Report

Dispatch is not "Claude chat on your phone." Dispatch is an orchestrator. From a single conversation on your phone, you spawn and manage multiple task sessions running in parallel on your desktop. Your phone is the command chair. Your desktop does the heavy lifting.

One-time setup: Connect Cowork on your desktop with the tools in your stack: Gmail, Notion, Slack. Keep your desktop awake and Claude running. Open Claude mobile, go to the Dispatch tab, and start.

How to run it correctly: Do not queue tasks one at a time. In a single Dispatch conversation, start Task 1 ("Summarize my unanswered emails this week") and immediately start Task 2 ("Draft the intro for the article in my workspace"). Both tasks run in separate sandboxed sessions on your desktop. You go do something else. When you return, redirect Task 1, launch Task 3, all from the same thread on your phone.

Three things to know before you start: Each subtask will request folder access separately on your desktop. There is no workaround, so keep your screen accessible. File transfer is not built in yet: sync your Cowork workspace folder with Google Drive and every output Claude creates appears on your phone automatically, and vice versa. Most importantly: tell Dispatch to read your CLAUDE.md before assigning any tasks. The instructions it writes for each subtask will be noticeably more precise.

Real numbers from 2 days of testing: 60 task sessions directed from a phone. Total direction time: roughly 25 minutes. Claude execution running in parallel: over 3 hours. The model is not "work during dead time." It is design your day differently because the work runs without you sitting in front of a computer.

A physics professor assigned Claude full ownership of a theoretical research paper, hypothesis through write-up, without touching a single file himself. The result: a publishable paper in two weeks instead of the typical year. The key finding: AI can do frontier science with guidance, but domain expertise remains essential to evaluate accuracy.

👉 Forward it to someone trying to keep up with AI.

👉 Read online: techzip.beehiiv.com

Techzip Newsletter
| Zipping what truly matters in the AI era.