Jamin Ball: “What’s happened over the last 12 months is the cost per API call (ie the cost for inference) for these models has plummeted. Open source models like Llama, R1 from DeepSeek, etc have all contributed to this. It’s become even more clear the model calls themselves are commoditizing quickly. And this is great! If the variable “COGS” component of marginal API call approaches zero, many of the questions listed above start to go away. You don’t worry about your margins shrinking, changing the pricing doesn’t become a must, etc. This in turn leads to a LOT more experimenting with AI features / functionalities. The radius of complexity shrinks. This is all great for AI application builders, and one reason I expect we will start to see an explosion of AI apps this year. As the costs go down, the experimenting will go up, and what we discovery will really start to impress.”
Ashu Garg: “AI’s trajectory is following a familiar pattern in computing history. First comes the scale-up phase, where early leaders extend their dominance through massive investments in infrastructure and talent. Then follows the scale-out phase, where efficiency and architectural innovation redistribute power from centralized incumbents to more decentralized, cost-effective alternatives. “Bigger is better” yields to “smarter is better.” This story has played out repeatedly: mainframes gave way to personal computers, AT&T’s centralized internet lost to TCP/IP’s distributed model, and Oracle’s database dominance ceded to MySQL and Postgres. Each time, centralization drove early progress before architectural breakthroughs enabled broader participation and new forms of value creation. Now, this same dynamic is reshaping AI.”
WSJ: “Tech and power companies are raising cash every which way: issuing shares, loans and bonds on publicly traded markets and in private deals. Large firms, or hyperscalers, such as Amazon.com and Microsoft could easily spend about $3 trillion by 2030 to build and operate data centers for their businesses, according to BlackRock Investment Institute. The trick for financiers is avoiding the losers in a new industry that can turn treacherous on a dime, as it did last week when Chinese upstart DeepSeek triggered a selloff in AI stocks. The recent frenzy echoes previous bonanzas such as fiber-optic cable, which boomed in the 1990s, busted in the 2000s and ultimately paid out. Fracking for oil and natural gas went through a similar cycle over the past 15 years.”
Ezra Klein: “Muzzle velocity. Bannon’s insight here is real. Focus is the fundamental substance of democracy. It is particularly the substance of opposition. People largely learn of what the government is doing through the media — be it mainstream media or social media. If you overwhelm the media — if you give it too many places it needs to look, all at once, if you keep it moving from one thing to the next — no coherent opposition can emerge. It is hard to even think coherently.”
NYTimes on how doctors can integrate AI into medical care: “Research points to three distinct approaches. In the first model, physicians start by interviewing patients and conducting physical examinations to gather medical information. A Harvard-Stanford study that Dr. Rajpurkar helped write demonstrates why this sequence matters — when A.I. systems attempted to gather patient information through direct interviews, their diagnostic accuracy plummeted — in one case from 82 percent to 63 percent. The study revealed that A.I. still struggles with guiding natural conversations and knowing which follow-up questions will yield crucial diagnostic information. By having doctors gather this clinical data first, A.I. can then apply pattern recognition to analyze that information and suggest potential diagnoses. In another approach, A.I. begins with analyzing medical data and suggesting possible diagnoses and treatment plans. A.I. seems to have a natural penchant for such tasks: A 2024 study showed that OpenAI’s latest models perform well at complex critical thinking tasks like generating diagnoses and managing health conditions when tested on case studies, medical literature and patient scenarios. The physician’s role is to then apply his clinical judgment to turn A.I.’s suggestions into a treatment plan, adjusting the recommendations based on a patient’s physical limitations, insurance coverage and health care resources. The most radical model might be complete separation: having A.I. handle certain routine cases independently (like normal chest X-rays or low-risk mammograms), while doctors focus on more complex disorders or rare conditions with atypical features.”
