October 8, 2026
Google just shipped AI Edge Foresight, an offline meeting note-taker aimed squarely at Granola, according to TechCrunch. It transcribes conversations, writes up notes, and answers follow-up questions entirely on-device. That last part is the headline, not the feature list. For a year now, the pitch for AI meeting tools has been convenience; the pitch here is that your conversations never have to leave your laptop.
I think this is Google reading the room correctly. Business buyers evaluating AI note-takers have quietly moved past 'does it transcribe well' to 'where does the transcript live.' Legal, HR, and board-level conversations are exactly the meetings people most want notes from and least want sitting on a third-party server. If Google can make on-device AI genuinely competitive with cloud models on quality, it changes the calculus for every vendor in this category, not just Granola. The tradeoff is real, though: local models are usually smaller and less capable than their cloud counterparts, so there's a quality tax for the privacy gain. Companies weighing this kind of tool alongside their broader internal tools stack should ask vendors directly where processing happens, not just what the output looks like.
Manus, the Chinese AI agent startup that split from Meta earlier this year, just closed a funding round north of $500 million, TechCrunch reports, led by Boyu Capital and IDG Capital with Tencent, HSG, and ZhenFund returning as backers. This is the company's first raise since going independent, and the size of it is notable at a time when plenty of agentic-AI startups are struggling to show they're worth the hype.
What stands out to me is who's writing the checks: deep-pocketed, patient Chinese capital that's betting on autonomous agents as infrastructure, not a feature. That's a different bet than the US market has mostly been making lately, where investors have grown more skeptical of agent demos that don't survive contact with messy real-world workflows. We've written before about how agents still struggle to earn real accountability once they're deployed at scale, and a $500 million round doesn't resolve that -- it just buys more runway to try. Businesses watching this space should separate the funding signal from the product signal: money flowing into agentic AI tells you where investors think the puck is going, not whether any given agent is ready to run your operations today.
OpenAI is rolling out a new interface that brings interactive visuals into ChatGPT, per TechCrunch, a shift from the text-box-and-scroll format that's defined the product since launch. Combined with the recent launch of Dots and OpenAI's Alexander Embiricos confirmed to appear at TechCrunch Disrupt 2026, it's clear OpenAI is pushing hard to make ChatGPT feel less like a chat log and more like a workspace.
This matters because interface is strategy. A more visual ChatGPT is stickier, easier to demo, and harder to churn away from -- exactly what you'd expect from a company fighting Google and Meta for daily attention, a race we've covered before. But stickiness isn't the same as utility for a business user who just needs a reliable answer or a document drafted. The risk is that chasing engagement metrics pulls product decisions away from the workflows that actually save a team time. Anyone using ChatGPT as a layer on top of their existing CRM or support tools should watch whether this redesign adds real functionality or just a prettier wrapper, especially if they're already weighing it against purpose-built options in their CRM or helpdesk stack.
Which of these three bets would you put your money on: local-first privacy, agentic infrastructure, or interface-driven stickiness? I'd love to hear which one you think actually changes how your team works day to day.
Sources