A Startup Just Raised $9M to Fix the Biggest Problem With AI
Today’s AI news has a common thread: trust. Can you trust your AI assistant to get the facts right? Can you trust your phone’s AI to actually help you multitask? Can you trust the glasses on your face to keep up? Three stories today all push toward the same answer — and it’s more complicated than the industry usually admits.
A Startup Is Betting $9M That AI Needs to Be More Like a Calculator
Most AI assistants are confident even when they’re wrong. A chatbot might tell you a drug has no side effects, cite a court case that never existed, or invent a statistic with complete conviction. This problem has a name: hallucination, meaning the AI generates plausible-sounding text that isn’t grounded in fact.
A startup called Probably wants to fix that. The company just closed a $9 million seed round led by Andreessen Horowitz — one of Silicon Valley’s most prominent venture firms — to build AI systems that behave more like traditional software. Traditional software is deterministic, meaning it gives you the same correct output every time you run it. Two plus two always equals four. Current AI language models don’t work that way.
Probably raised $9 million in seed funding to prevent AI hallucinations and factual errors.
Think of it this way: a calculator never guesses. Probably is trying to bring that reliability to the kind of AI that reads documents, answers questions, and summarizes reports. If they succeed, AI tools used in medicine, law, and finance could become genuinely trustworthy rather than just impressively articulate. For regular people, that’s the difference between using AI to draft a contract and actually relying on what it says. TechCrunch has the full story.
Why this matters: The usefulness of AI in high-stakes situations depends entirely on accuracy. Funding a fix for hallucinations is one of the most practical bets anyone can make right now.
Qualcomm Built a Chip That Could Make Smart Glasses Worth Wearing
Smart glasses have existed for years, but most of them have been underwhelming — either too bulky, too dumb, or too dependent on your phone to do the real work. Qualcomm’s new Snapdragon Reality Elite processor is designed to change that by putting serious computing power directly inside the glasses frame.
The chip is built for mixed-reality devices, meaning glasses that can overlay digital information onto the physical world you’re already looking at. The first product to use it will be Xreal’s Aura, a pair of Android XR glasses scheduled to launch this fall. Android XR is Google’s operating system built specifically for headsets and glasses.
Snapdragon Reality Elite will first arrive in Xreal’s Aura Android XR glasses this fall.
For everyday people, this matters because more powerful glasses means more useful glasses. Imagine looking at a restaurant and seeing reviews floating next to it, or following turn-by-turn directions without staring at your phone, or reading translated text overlaid on a foreign menu. None of that works smoothly without enough processing power. Qualcomm is essentially making the case that glasses could eventually replace some of what you currently reach for your phone to do. The Verge broke down the technical specs and what they mean for the hardware.
Why this matters: The glasses form factor has always had a power problem. A chip designed specifically for it could finally close the gap between concept and daily usefulness.
Android 17 Is Here, and Gemini Is Baked in Much More Deeply
Google released Android 17 this week, along with Wear OS 7 for smartwatches. The update brings cleaner multitasking tools — making it easier to run two apps side by side on larger screens — and deeper integration with Gemini, Google’s family of AI models.
Gemini integration means the AI assistant is no longer just a separate app you open when you need it. On Pixel devices, it can now operate more fluidly across the system, helping with tasks in other apps rather than requiring you to switch contexts. There are also new parental controls and security improvements baked into the release.
For smartwatch users, Wear OS 7 adds software features that make the watch smarter without requiring a constant connection to your phone. This matters to anyone who exercises, commutes, or simply wants their wrist to do more on its own. TechCrunch has the full breakdown of what changed.
Why this matters: Every Android update that makes Gemini more embedded is a quiet expansion of AI’s role in daily life — often before users even realize it’s happening.
Also Happening in AI
Robinhood announced it’s cutting 10% of its workforce, but notably avoided blaming AI — a contrast to companies like Cloudflare and Coinbase, which have been more explicit about AI reducing their headcount, according to TechCrunch. Anthropic temporarily paused a planned shift to token-based billing for its Claude Agent SDK — tokens being the small chunks of text AI models process — after pushback from developers, Ars Technica reported. OpenAI quietly released version 2.42.0 of its Python library, adding spending alerts for developers managing API costs. On a more contentious note, TechCrunch reports that the Department of Justice argued xAI should be allowed to keep operating unpermitted gas turbines at its data center, framing the situation as a matter of national and energy security. And Wired published a personal account of AI-powered home monitoring devices designed to help aging parents stay safe independently — a quieter, more human side of where this technology is heading.
What to Watch
The Probably funding round is worth following closely, because it signals that serious money is now moving toward AI reliability rather than just AI capability. Watch for more startups and established players to announce similar efforts in the next few months. Meanwhile, the Qualcomm chip announcement and Android 17’s Gemini expansion together suggest the second half of 2026 will see AI move further off the screen and into the physical environment around you — which raises real questions about how much of your daily life you’re comfortable having monitored, assisted, and interpreted.