Google unveils TurboQuant!

Also: Claude and Cowork can now control your computer, while Granola raises $125M as it expands beyond meeting notes 📝.

Author

The Neural Frontier
March 27, 2026

In partnership with

88% resolved. 22% stayed loyal. What went wrong?

That's the AI paradox hiding in your CX stack. Tickets close. Customers leave. And most teams don't see it coming because they're measuring the wrong things.

Efficiency metrics look great on paper. Handle time down. Containment rate up. But customer loyalty? That's a different story — and it's one your current dashboards probably aren't telling you.

Gladly's 2026 Customer Expectations Report surveyed thousands of real consumers to find out exactly where AI-powered service breaks trust, and what separates the platforms that drive retention from the ones that quietly erode it.

If you're architecting the CX stack, this is the data you need to build it right. Not just fast. Not just cheap. Built to last.

If you’ve ever watched the HBO hit show, Silicon Valley, Google’s TurboQuant would probably look a lot like Pied Piper 😼. And if you haven’t, that’s a free recc to kick off your weekend. 

Forward thinkers, welcome to issue #151 of the Neural Frontier.

The headlines say enough; so let’s unpack, shall we? 

In a rush? Here's your quick byte: 

🤖 Google unveils TurboQuant!

💻  Claude and Cowork can now control your computer. 

📝 Granola raises $125M as it expands beyond meeting notes!

🎯 Everything else you missed this week.

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week.

Source: Google

Google researchers have introduced TurboQuant, a new AI memory compression technique that could significantly reduce how much memory AI systems need — without sacrificing performance.

And yes, the internet immediately compared it to “Pied Piper” from Silicon Valley — the fictional startup known for its near-perfect compression algorithm.

⚡ What TurboQuant actually does

At a high level, TurboQuant tackles one of AI’s biggest bottlenecks: working memory during inference (often called the KV cache).

The breakthrough:

  • Compresses AI memory without degrading output quality

  • Uses vector quantization techniques to shrink memory usage

  • Clears cache bottlenecks during model processing

The result? Models can “remember” more while using less memory — a big deal for performance and cost.

🔬 Under the hood (without the headache)

TurboQuant builds on two core methods:

  • PolarQuant → a new quantization technique

  • QJL → a training and optimization method

You don’t need to understand the math to get the impact: this is about making AI systems more efficient at runtime.

💸 Why this matters

If TurboQuant works at scale, it could:

  • Reduce AI memory usage by 6× or more

  • Lower the cost of running AI systems

  • Improve performance in memory-constrained environments

That’s why some industry voices are calling it a potential “efficiency breakthrough” moment, similar to how smaller, cheaper models disrupted expectations.

⚠️ Important limitation

There’s a catch: TurboQuant focuses on inference, not training. That means:

  • It helps when AI is running tasks

  • It does not reduce the massive compute and memory needed to train models

So while it can make AI cheaper to use, it doesn’t fully solve the broader infrastructure demands of AI development.

TurboQuant is currently a research-stage breakthrough, with plans to be presented at ICLR 2026.

It hasn’t yet been widely deployed in production systems, so real-world impact is still to be proven.

Source: Anthropic

Anthropic is pushing deeper into agentic AI with a new update that lets Claude Code and Cowork directly use your computer to complete tasks.

The feature, now in research preview, allows Claude to go beyond suggestions — and actually open apps, browse the web, and execute workflows on your behalf.

🤖 From assistant to operator

With this update, Claude can open files and applications, use your web browser, run developer tools, and interact with your system (mouse, keyboard, display). 

Instead of relying only on integrations, Claude can operate your computer like a human would — clicking, scrolling, and navigating interfaces to get work done.

🔗 Works with (and beyond) integrations

Claude prioritizes existing connectors where possible, including Slack and Google Workspace apps.  But if no integration exists, it can fallback to direct control of your system, effectively removing the need for native integrations in many cases.

📱 Designed for async workflows

The feature pairs with Anthropic’s Dispatch system, allowing users to:

  • Assign tasks from their phone

  • Let Claude execute them on their desktop

  • Return later to completed work

This makes it possible to delegate tasks and walk away, similar to how other agent systems are evolving.

🔐 Human-in-the-loop (for now)

Anthropic emphasizes that Claude always asks for explicit permission before taking actions, shows what it plans to do, and requires approval before executing certain steps. This keeps users in control while still enabling automation.

⚠️ Early-stage limitations

The feature is still experimental, with a few constraints:

  • macOS only (for now)

  • Available to Claude Pro and Max users

  • Complex tasks may require retries

  • Direct system control can be slower than native integrations

Anthropic is releasing it early to gather feedback and understand where it works best.

📝 Granola raises $125M as it expands beyond meeting notes!

Source: Granola

Granola — the AI meeting notetaker quietly gaining traction — has raised $125 million in Series C funding, pushing its valuation to $1.5 billion as it evolves into a broader enterprise AI tool.

📈 From simple notes to enterprise workflows

Granola started as a lightweight desktop app that transcribes meetings and generates notes — with a key insight: Users don’t like bots in meetings, but they’re fine with silent note-taking in the background.

Now, it’s expanding into a more collaborative and enterprise-ready system, adding:

  • Shared notes across teams

  • Collaborative editing

  • Structured workspaces called Spaces

  • Folder-level permissions and access controls

🧠 Turning notes into actionable context

Granola is betting that note-taking alone is becoming a commodity. So the next phase is about making notes useful in workflows, not just stored. New additions include:

  • Personal API → access your notes + shared notes

  • Enterprise API → org-wide context for teams and admins

  • MCP server updates → better integration with AI agents

This allows notes to plug into workflows like drafting follow-up emails, scheduling meetings, pulling insights from company knowledge, and supporting sales and CRM workflows. 

🔗 Built for the AI ecosystem

Granola is positioning itself as a context layer for AI tools, already integrating with:

  • ChatGPT

  • Claude

  • Figma Make

  • Replit

  • v0, Bolt, and other dev/design tools

The idea: your meeting data becomes fuel for downstream AI workflows.

🎯 Everything else you missed this week.

Source: Bytedance

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week. 

1. 🚀 Fastlane is a viral marketing platform that analyzes your website to generate trending TikTok, Instagram Reels, and YouTube Shorts content with AI UGC avatars.

2. ⚡ Zeus AI is a local-first AI platform that provides autonomous AI workers with dedicated identities (name, email, personality) who complete hour-long tasks 24/7 across sales, recruiting, and customer success.

3. 🎬 Image to Video AI is an AI video generation platform that turns images into videos using 10+ leading models, letting users control motion, style, camera angles, and mood with text prompts.

Wrapping up…

If this year has an AI theme, it’s definitely something along the lines of “Agentic AI, the 9th wonder of the world.”

This week had it all: Compression tech that we could only dream of years ago, agentic AI at the help of affairs, and product evolutions.

We have no idea what the tide will bring in next week, but we can tell you this: we’ll be right here with the scoop, as always 😏.

Till then, this has been the Neural Frontier, and we’ll catch you on the flip side! 👋

PS: If you’re the “friend” this mail was forwarded to, and you enjoyed it, hit the Subscribe button to see more content like this every week 🙂.