Daily Tech Brief · October 22nd

IBM's CEO, Arvind Krishna, shared some intriguing insights during the company’s Q3 earnings call, particularly about open weight models for on-premise enterprise use. He noted that while closed frontier models have been superior in the past, the gap is narrowing. By late 2025, he expects the difference in effectiveness to shrink from a score of 70 to 50. Currently, open models are significantly cheaper—costing only 10 to 25 percent of their closed counterparts—while still being about 90 percent as effective. This makes them appealing for tasks like summarizing emails or answering customer calls.

Krishna highlighted that many businesses may prefer these cost-effective options, but he acknowledged a challenge: implementing the necessary AI infrastructure. He pointed out that enterprises often struggle to find the right talent to manage the underlying technology, like kernel engineers and network specialists. IBM plans to develop software to help ease these deployment hurdles.

On the financial side, IBM's infrastructure revenue grew modestly by five percent, while software revenue saw a more robust nine percent increase, largely driven by Red Hat's success. Total quarterly revenue reached $17.1 billion, though the company's share price dipped slightly in after-hours trading. Additionally, IBM faced a significant cloud incident that disrupted resource provisioning for users, which they are working to resolve.

On the markets — Kalshi traders have been actively repricing this story in the last day.

Today’s tech scene is buzzing, especially with NVIDIA’s GTC Berlin event. CEO Jensen Huang showcased some quirky robotic ducks to illustrate Europe’s potential in physical AI, though one duck had a bit of a meltdown during the demo. It’s a reminder of how even the best tech can have its hiccups. Meanwhile, a startup called Flapping Airplanes has just raised a whopping $1 billion in Series A funding, hitting a $6 billion valuation. They’re tackling the challenge of making AI more efficient, which is crucial as demand continues to surge.

OpenAI is making waves too, having used its GPT-5.6 Sol model to identify significant vulnerabilities in PHP, potentially safeguarding 70% of the world’s websites. It’s impressive how AI can help patch security holes before they’re exploited. Dario Amodei from Anthropic is defending their strategy to stay at the forefront of AI development, emphasizing that leading the charge can enhance safety measures.

In other news, DeepMind’s Demis Hassabis has shifted his timeline for AGI’s arrival to just two to three years, citing rapid progress. Anthropic is rolling out Claude Duet, an AI assistant designed to manage tasks across various applications, while NVIDIA has introduced OpenShell, a new security framework for autonomous agents. Lastly, a cybercrime group has been using AI models to orchestrate attacks on banks in Latin America, highlighting the darker side of AI’s capabilities. It’s a fascinating and complex landscape we’re navigating.

Apple's new HomeView, set to launch on November 13, is designed to enhance smart home interactions through Siri AI's evolved capabilities. In a recent interview, Apple’s product marketing team emphasized how Siri can now handle requests more naturally, making it accessible for a wider range of users, from tech-savvy Gen Z to those less familiar with smart tech, like your mother-in-law.

The focus is on simplicity; Siri aims to complete tasks with minimal back-and-forth, letting users move on quickly instead of getting bogged down with questions. HomeView can recognize up to six users, tailoring information for each, while still allowing general queries without needing a compatible device. Priced at $299, it seems Apple is really honing in on making smart home tech feel more intuitive and personal.

An anonymous reader quotes a report from The New York Times: For the first time, scientists say they have used viruses designed by artificial intelligence to treat a human patient. Researchers announced on Tuesday that they had administered the virus, known as a phage, to a patient suffering from a deadly bacterial infection. Phages attack bacteria, and the hope is that this one will kill the microbes causing the infection. Radical Numerics, the start-up that created the viruses, posted a technical report about them on its website.

With the newest AI trend of people vibe-coding ways to debloat their smart TV , one person was inspired and decided to try it on their own device. Unfortunately, what they thought was an Android TV ended up not being Android at all, but after almost half a day of work, Claude managed to install a new Apple TV-like home screen, and it needed the user's phone to get the job done.

- GridCraft is almost the MS Excel clone it intends to be. - GridCraft is both free and open-source. - The only thing missing is a strong template feature. ArtCraft, the development team behind PhotoCraft (and a number of other similar apps) reverse-engineered yet another de facto standard app: GridCraft, meant to mimic Microsoft Excel. ArtCraft’s collection is an important step forward for open source and the Linux operating system because some of the apps they’ve built have long been considered roadblocks to user adoption of Linux.

Watchtower Labs, a Bellevue, Wash.-based defense tech startup, emerged from stealth Wednesday with $35 million to build AI software that helps military planners make faster decisions about logistics and operations. Shield Capital and Conviction co-led the seed round, with participation from Founders Fund and In-Q-Tel, the investment arm that supports the U.S. intelligence community. It’s the first outside funding for Watchtower, which was incubated inside Shield.

That’s a rather unusual timeline, but one in keeping with Musk’s penchant for outlandish plans and overpromising on delivery date projections. Musk announced the giant chip factory project in Austin, Texas back in March with claims that it will produce chips for both terrestrial and space-based use cases. But chip factories are notoriously tough and costly to build, especially in the United States instead of somewhere like Taiwan that has established expertise. Making the factory operational alone can often take several years. But Musk promises a more streamlined timeline for his fab.

Meta CEO Mark Zuckerberg wants everyone chatting with its Muse agents. And then he wants those agents chatting with each other. “We expect to grow Muse to serve billions of people. This is both our mission and our strategy,” Zuckerberg said.

Nvidia has decided that its future prospects rest on an improved silicon substrate and packing supply chain, so has dipped into its giant pile of cash to ensure it can get the parts it needs by ensuring its suppliers have the resources needed to play their small part in keeping the flow of GPUs and AI processors strong. It’s not unusual for manufacturers to invest in their suppliers, because it helps to create certainty and spread financial risk.