When we talk about the cost of AI infrastructure, the focus is usually on Nvidia and GPUs ' but memory is an increasingly important part of the picture. As hyperscalers prepare to build out billions of dollars' worth of new data centers, the price for DRAM chips has jumped roughly 7x in the last year. At the same time, there's a growing discipline in orchestrating all that memory to make sure the right data gets to the right agent at the right time. The companies that master it will be able to make the same queries with fewer tokens, which can be the difference between folding and staying in business. Semiconductor analyst Doug O'Laughlin has an interesting look at the importance of memory chips on his Substack, where he talks with Val Bercovici, chief AI officer at Weka. They're both semiconductor guys, so the focus is more on the chips than the broader architecture; the implications for AI software are pretty significant too. The tell is if we go to Anthropic's prompt caching pricing page. It started off as a very simple page six or seven months ago, especially as Claude Code was launching ' just 'use caching, it's cheaper.' Now it's an encyclopedia of advice on exactly how many cache writes to pre-buy. You've got 5-minute tiers, which are very common across the industry, or 1-hour tiers ' and nothing above. That's a really important tell. Then of course you've got all sorts of arbitrage opportunities around the pricing for cache reads based on how many cache writes you've pre-purchased....

OpenClaw is an open-source artificial-intelligence agent designed to assist users with everyday tasks, such as sending e-mails and managing their calendars.Credit: Thomas Fuller/SOPA Images/LightRocket via Getty The sudden rise of a huge network of artificial-intelligence bots talking to each other about religion and their human 'handlers' has captivated a corner of the Internet. The phenomenon has also given scientists a glimpse into how AI agents interact with each other ' and how humans respond to those discussions. OpenClaw is an AI agent capable of performing tasks on personal devices, such as scheduling calendar events, reading e-mails, sending messages through apps and using the Internet to make purchases. Most of the popular AI tools, such as OpenAI's ChatGPT chatbot, work by interacting directly with user prompts, whereas agentic AI models such as OpenClaw can carry out actions autonomously in response to instructions. Agentic AI tools have been used in some industries for years, such as for automated trading and optimizing logistics, but their adoption by the general public has been minimal. Improvements in the capabilities of large language models have made it possible to create more versatile AI tools, researchers say. 'OpenClaw promises something especially appealing: a capable assistant embedded in the everyday apps people already rely on,' says Barbara Barbosa Neves, a sociologist who focuses on technology at the University of Sydney in Australia....

The political takeover of Venezuela does not appear to be going according to plan, if there ever was a plan. A day after the raid that captured Venezuelan President Nicolas Maduro, President Donald Trump used the phrase 'run the country.' A day later, Secretary of State Marco Rubio watered it down to 'running policy' and made the next phase sound more like a series of strong suggestions. Trump also publicly suggested that Venezuelan Vice President Delcy Rodriguez might cooperate with the United States. But then Rodriguez gave a televised address saying Venezuela 'will never return to being the colony of another empire.' Trump followed that up with a threat, delivered through an interview with The Atlantic: 'If she doesn't do what's right, she is going to pay a very big price, probably bigger than Maduro.' Unlike the precise military operation that whisked Maduro out of Venezuela, the political follow-up seems like improvisation and chaos, but maybe chaos is the point. In this episode of Radio Atlantic we talk with staff writers Michael Scherer, who talked with President Trump on Sunday, and Vivian Salama, who writes about the administration and Venezuela. They talk about Trump's evolving vision of his role in the world and how it might be unsettling other world leaders, including Russian President Vladimir Putin. Salama notes that world leaders now may 'think twice about crossing the U.S. when President Trump is so unpredictable and uses the U.S. military in very unconventional ways, and shuns the guardrails that keep a U.S. president in check.'...

The National Highway Traffic Safety Administration (NHTSA) has identified at least 80 instances in which Tesla's Full Self-Driving (Supervised) software violated road rules by running red lights or crossing into the wrong lane, according to a new letter sent to the automaker this week. NHTSA said in the letter it has received 62 complaints from Tesla drivers, 14 reports submitted by Tesla, and four media reports that describe potential violations. That's up from around 50 violations NHTSA cited when it opened an investigation into the behavior in October. The federal safety agency's Office of Defects Investigation (ODI) is probing whether Tesla's driver assistance software can 'accurately detect and appropriately respond to traffic signals, signs and lane markings,' according to the letter. ODI is also evaluating whether Tesla's software is providing sufficient warnings to drivers in these situations. Tesla's responses are due January 19, 2026. The increase in complaints is notable in part because the original batch reported by ODI in October included multiple reports from one particular intersection in Joppa, Maryland. Tesla told the agency at the time that it had already 'taken action to address the issue at this intersection.' The agency didn't say where, geographically, these newly reported incidents took place. Tesla heavily redacts its own submissions to the agency....