TLDR 2025-12-30

Meta has agreed to acquire Manus, a Singapore-based company that conducts deep research and performs tasks for paying users. Meta will continue to operate and sell Manus' services while integrating it into its suite of social media products. The deal will help Meta cement its position in the product segment of AI agents. It is one of the first times a major US tech company has bought a startup with Chinese roots.

Nvidia locked in a purchase price of $23.28 per share for Intel when the companies struck a deal in September. The deal had been under scrutiny by the US Federal Trade Commission, but was then greenlighted on December 18. The purchase of 214 million shares closed on December 26, and Intel shares closed Monday at $36.68, making Nvidia's $5 billion purchase worth $7.58 billion. The deal will involve the companies jointly developing multiple generations of chips for datacenters and PCs.

Chinese researchers have created an artificial robotic skin that can sense pressure and locate input and injuries. The neuromorphic robotic e-skin (NRE-skin) is assembled from a collection of segments that snap together using magnetic interlocks, automatically linking up any necessary wiring. Each segment broadcasts a unique identity code, so it is relatively easy to pop out the damaged segment and replace it with fresh hardware if the system identifies damage.

Biomedical progress has become less productive for the last several years despite staggering advances in basic science. One of the reasons for this is that institutional bureaucracy has become harder to overcome. Increasing the number and efficiency of clinical trials would help create a faster feedback loop and result in better data to inform models and ideas. This could help decrease the cost of bringing new drugs to market and break a trend that has held since the 1950s.

Pressure. Extreme diets. All-or-nothing mindset. This doesn't need to be you on 01/01. There is a clinically proven weight loss tool that's helped users lose over 18 million pounds. Take a quick quiz and get a science-backed plan built just for you—plus Blinky, a Tamagotchi-style accountability buddy that keeps you on track. It's crazy effective. Get your personalized plan for at least 50% off.

The primary users of infrastructure software are rapidly shifting from developers to AI agents. AI uses systems very differently from how developers do, and it changes many long-held assumptions about how databases should be used. Many things developers took for granted need rethinking. The focus of engineers is shifting from perfected individual systems to designing foundational capabilities that AI can use at scale, iterate on, and run cheaply.

AI will likely make 24/7 on-call a thing of the past. 24/7 monitoring is a lot simpler than the development process. There are often reference documents that engineers can follow to bring systems back up, and if they fail, there's always a backup and recovery plan in place. On-call jobs have always been more systematic. This post introduces an autonomous monitoring agent that does what an on-call engineer would do, but autonomously, forever.

Labor and capital have traditionally complemented each other. While wealthy people can keep accumulating capital, it becomes less valuable when there aren't enough hands to use all of it, and hands grow more valuable when capital is plentiful. However, this correction mechanism breaks in the world of advanced robotics and AI. A global and highly progressive tax on capital (or at least capital income) may be the only way to prevent inequality from growing extreme.

The immediate AI risk comes from people overestimating AI's capabilities. A lack of reliability and trust is preventing wide adoption. There is a growing AI perception gap between quantitative users and qualitative users. AI leaders aren't even attempting to explain how AI works because it's complicated, and they're also incentivized to oversimplify and overpromise.

Both human and computer expertise grow quickly in the beginning, for a short while, then more slowly, until growth is imperceptible.

2026 will see breakthroughs in continual learning and a blockbuster IPO.

Generic design advice is useless for most practical software design problems.

The LG Gallery TV is a mini-LED TV with a special screen that reduces glare and minimizes reflections for an art-like viewing experience.

AI models excel at isolated React tasks, but don't do as well on multi-step integrations.

The software engineering field will broadly become more mechanized and more productive.