October 14, 2026
At TechCrunch Disrupt, Cerebras CEO Andrew Feldman said we're nearing a future where every household has a personal AI assistant, split between local hardware and the cloud, as ordinary as a microwave. It's a compelling image, and Feldman has a real stake in it: Cerebras builds the chips that would power exactly this kind of split-compute future. That doesn't make him wrong, but it does mean the prediction is also a sales pitch.
The business version of this question isn't 'will my house have an assistant' -- it's 'when does my team get one that actually works inside the tools we already use.' That's a very different bar than a chatbot that answers trivia on a kitchen counter. It means an assistant wired into your CRM, your support queue, your scheduling, not a standalone gadget. Companies evaluating this now are better served looking at how workflow automation and an AI app builder already let them build that assistant into daily operations today, rather than waiting for a consumer hardware wave that may take years to arrive and longer to prove useful at work.
In a second session at the same event, Feldman took a swing at AI leaders who keep warning about existential risk, arguing their doom talk is crowding out a fair accounting of AI's benefits. There's something to this -- a steady diet of extinction-level warnings from the people building the technology is genuinely confusing for anyone trying to make a sober purchasing decision. If the people closest to the tech can't agree whether it's dangerous, why should a mid-size business trust it with customer data?
But flip it around and Feldman's complaint is just as self-interested as the doom talk he's criticizing. He runs a chip company that benefits directly from AI adoption accelerating, unslowed by public anxiety. Dismissing risk warnings as noise is as convenient for his business as amplifying them is for a lab trying to justify safety-driven caution or regulatory capture. Neither posture is neutral, and a business reader shouldn't take either side's word for it. The honest position is that AI vendors, including the ones ViibeStack covers daily, have incentives on both sides of this argument, which is exactly why buyers need their own criteria -- things like where data lives, who can see it, and what happens when an agent gets something wrong -- rather than vibes from a conference stage. That's the kind of groundwork we've laid out in our own security and trust center pages, and it matters more than whichever CEO sounds most confident this week.
What's notable about pairing these two stories is less either claim and more what they signal together: chipmakers are now positioning themselves as AI's spokespeople, not just its suppliers. Disrupt's own framing this year -- 'physical and software AI' -- suggests the industry increasingly wants buyers to think of compute, robotics, and assistants as one continuous story. For a business owner, that's a reminder to separate the infrastructure pitch from the product reality. A faster chip doesn't automatically mean a more useful assistant in your support queue or your HR workflow, any more than a bigger engine automatically means a better car. The gap between 'this is coming' and 'this works for my five-person team today' is still wide, and it's worth asking vendors to show the latter before buying into the former.
Where do you land: is the AI-safety conversation genuinely drowning out useful progress, or is that framing just another way to wave off legitimate concerns?
Sources