Anthropic: “Multi-agent systems work mainly because they help spend enough tokens to solve the problem. In our analysis, three factors explained 95% of the performance variance in the BrowseComp evaluation (which tests the ability of browsing agents to locate hard-to-find information). We found that token usage by itself explains 80% of the variance, with the number of tool calls and the model choice as the two other explanatory factors. This finding validates our architecture that distributes work across agents with separate context windows to add more capacity for parallel reasoning. The latest Claude models act as large efficiency multipliers on token use, as upgrading to Claude Sonnet 4 is a larger performance gain than doubling the token budget on Claude Sonnet 3.7. Multi-agent architectures effectively scale token usage for tasks that exceed the limits of single agents.”
Bloomberg: “Nearly one in four Indians said in a survey they use only mobile phones to consume media content, as TVs stay unaffordable for many, prompting companies from Meta Platforms Inc. to Netflix Inc. to boost digital strategies in the world’s largest consumer market. The number of users who only use digital channels ballooned to 23% in the March quarter of 2025, according to market research firm Kantar’s Media Compass report this week, which surveyed 87,000 Indians across the country. That compares with 15% in the same period in 2023. The trend is skewed toward lower-income groups and rural users, especially men, said Puneet Avasthi, director of specialist businesses at Kantar’s Insights Division for South Asia. Many people who are now consuming digital content on mobile phones were “media dark,” he said, as television sets and multiple OTT subscriptions are costly.”
Vox: “Scientist AI won’t be like an AI agent — it’ll have no autonomy and no goals of its own. Instead, its main job will be to calculate the probability that some other AI’s action would cause harm — and, if the action is too risky, block it. AI companies could overlay Scientist AI onto their models to stop them from doing something dangerous, akin to how we put guardrails along highways to stop cars from veering off course.”
Ryan Roslansky: “The basic task of, “Hey, there’s a billion people on LinkedIn. I need to find someone like [them]” — it’s really good at that. Step two: “Reach out to this person to make sure that she’s actually in the market for a job or wants to have a conversation.” Historically, it’s not great at that. But that’s a pretty low-level task we’ve gotten really, really good at. “Talk to this person and convince them to come work at LinkedIn?” No way. A human being is still way better at that sales process. That’s where I think that the tools stop. I don’t know if they’ll ever really catch up.”
100 best movies of the 21st century, from NYTimes.
