Zuckerberg Calls His New Model "Frontier Performance Almost Too Cheap to Meter."

Ready to transform artificial intelligence from a buzzword into your personal revenue generator?

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get The Free Guide

1  Meta released Muse Spark 1.3, a frontier-level model priced significantly below comparable offerings from Anthropic and OpenAI.

Zuckerberg and Meta are back in the frontier race with Muse Spark 1.3, a new model that puts Meta on par with the top offerings from Anthropic and OpenAI. It is currently free on OpenCode, with API pricing well below many comparable models. Zuckerberg described it as "frontier performance almost too cheap to meter." A company that has spent the past year trailing on model quality shipping something genuinely competitive, and pricing it aggressively on top, is a real change in posture rather than an incremental update.

2  Google shipped a new image tool and its third Flash model in six weeks, on the same day.

Google Pics generates Nano-Banana-quality images with precise editing, letting you adjust text or individual elements directly, and integrates across Google Workspace's tools. It is generally available now to Workspace, AI Pro, and Ultra customers. The company also debuted Gemini 3.8 Flash, a cost-efficient reasoning model and its third Flash release in six weeks. That cadence on the cheap tier says as much about where Google sees the real competitive fight as any single flagship launch would.

3  Claude Cowork and Claude Code can now run tasks in the background on your computer while you do something else.

Both tools can now run in the background to complete tasks on your computer, freeing you up to work on something else in the meantime. The feature is in beta on Pro and Max plans on macOS, and only works for apps you've explicitly given Claude access to. Anthropic also published a prompting guide for Fable 5.1 that includes instructions for getting the model to write more clearly.

You don't have to scroll every AI thread, track every new tool, or watch every demo. 

The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning. 

Trusted by 2M+ professionals at Apple, Google, and NASA.

Sign up to start learning.

The scale of the moat. Just three years ago, Nvidia was doing roughly $7B in quarterly revenue. Now it's up to $96B and on track to command 75% of the AI accelerator market in 2026. Nvidia's GPUs still train and run almost every large model in existence. That dominance is exactly why every other serious player in the industry now has a reason to try to chip away at it.

Four fronts, all at once. Chip startups have raised $8.3B as of April, arguing GPUs were never purpose-built for AI and that leaner designs can cut both cost and energy use. AI labs are moving into hardware directly: OpenAI claims its first chip Jalapeño offers industry-leading performance, Anthropic is hiring hardware executives, and rival chipmaker Cerebras went public in May to fund its own push. Big Tech is catching up too: Google launched two new chips in April, Amazon keeps iterating on its Trainium line, and Microsoft is expected to unveil its Maia 300 sometime this month, on top of the existing competition from AMD and Broadcom.

China is quietly closing the gap fastest. Nvidia controlled 66% of China's accelerator market in 2024, but is on track to own just 8% by 2026 as the country approaches self-sufficiency. Of the four fronts described here, this is the one moving with the most government backing behind it, which makes it the hardest for a single company to compete against on pricing or performance alone.

Why Nvidia is still far out in front. The chipmaker's real moat isn't the silicon itself. It's CUDA, the software layer millions of developers have already built their tools and workflows on top of. Even as viable hardware alternatives ship from every direction, the broader AI industry is expanding faster than any single challenger can absorb, which means Nvidia doesn't need to win every front to stay ahead. It just needs the overall market to keep growing faster than competitors can eat into its share.

🤑 Speedier gadgets. Hobbyists are using AI to enhance tech products they already own, instead of buying new ones. Two viral posts claim Claude improved their Wi-Fi speeds and de-bloated a sluggish TV (1M views).

🧓 AI for seniors. A new smart speaker acts as a subtle caretaker for elderly relatives. No screen, no setup, just plug it in: a voice agent offers daily companionship while keeping tabs on their health (1M views).

🤔 Did you know. Despite its recent popularity, AI actually dates back 70 years and already survived two industry "winters" in the 1970s and 80s. Learn 18 other facts most people don't know about the industry.

📝 Prompt audit. Consider having Fable 5.1 run a prompt audit to remove outdated rules or redundancies from your workflows. AI educator Peter Yang recommends doing this whenever a strong new model drops.

🔍 Going small. Shopify CEO Tobi Lütke says the company has a strategy for beating frontier model performance: training small models to complete niche tasks with a self-improving flywheel. He also posted the internal platform Shopify built to do this.