TLDR 2025-02-24

The announcement around the Stargate project made it clear that Microsoft's relationship with OpenAI was to evolve, allowing OpenAI to use other cloud providers. Microsoft is not fully breaking up with OpenAI - it will remain a key partner in many regards. However, what was a tight coupling of OpenAI and Microsoft will now be a tight coupling of OpenAI and SoftBank.

The iOS 18.4 beta adds a new 'Priority Notifications' feature. Powered by Apple Intelligence, the feature is designed to help users manage notifications by prioritizing important alerts and minimizing distractions from less important ones. Priority Notifications will be displayed in a separate section on the iPhone's Lock Screen. Users will still be able to swipe to view all of their notifications. The feature will be off by default but can be enabled through the Notifications section in Settings.

Genetic technology could create humans that live longer, have the raw intellectual power to do Nobel Prize-worthy scientific research, and very rarely suffer from health disorders. Scientists don't do this though due to ethical issues. As a result, there is a lot of low-hanging fruit in this field of research. This post looks into how to do large-scale, heritable editing of the human genome to improve everything from diabetes risk to intelligence. It summarizes the state of knowledge in the field and lays out a technical roadmap that examines how remaining barriers may be overcome.

AstroForge is planning to heave a microwave oven-sized spacecraft toward an asteroid later this week as part of the company's plans to create a future where precious metals are mined around the solar system. Its Odin spacecraft will launch on a SpaceX Falcon 9 rocket as soon as Wednesday. Odin's target is an object called 2022 OB5, which is about the size of a football field and is likely to contain a high amount of metal. Odin will arrive in late 2025 after a journey of about 300 days.

Google's Director of AI Dr. Ali Arsanjani, and Datadog's VP of Engineering Sajid Mehmood, sat down to discuss the current and future states of AI, ML, and LLMs on Google Cloud. This guide distills the top 7 insights and actions for technical leaders, covering everything from upskilling teams to observability best practices. Read it here

Building an LLM (Large Language Model) chat app is way more complicated than just making an API call. There are many considerations required - building an LLM chat app is the gauntlet that separates the juniors from the devs who actually know what they are doing. This article discusses how to start building a scalable LLM chat app. Learning how to build the backend architecture allows developers to build better apps that can handle more users and dodge the shame of a launch-day crash.

CSS Zero is an opinionated CSS starter kit for Ruby on Rails applications. It is akin to a 'no build' Tailwind CSS. CSS Zero includes custom templates for scaffolds and authentication.

A large part of the economy just won't adopt AI or use it effectively all that quickly. As AI becomes more productive, human bottlenecks will become more important. Not all bottlenecks will be regulatory: there will be slow adopters, people in the workplace who hate AI, energy constraints, and other factors. It is the worst performer who sets the overall level of productivity - soon, the worst performer will be humans. While AI could boost economic rates, that reality will still be a few decades out, if it happens.

The thinking that startups should prioritize growth over everything else, even profitability, was always flawed - profitability isn't unambitious. It means not relying on investors for survival and being able to focus on an unaltered vision and mission. It's now easy to be profitable while also growing fast. The factor holding founders back is rarely team size - it's the clarity of their focus, skill, and ability to execute. Measure what matters, understand your risk profile, hire intentionally and slowly, and raise on your own terms.

Google's new video-generating AI model can create clips that are two minutes or more.

The methods that AI labs use to present AI benchmark results can be questionable.

The transition to IPv6 has progressed very slowly because there's no real lasting benefit in trying to leap across to another 1980s address-based architecture with only a few differences.

Apple's C1 modem, which lacks mmWave support but features custom GPS and satellite connectivity, was designed to cater specifically to the needs of the iPhone 16e and future devices.

Uber, Waymo, and Tesla will be competing against each other when Tesla launches its robotaxis in Austin.

NASA has awarded SpaceX a contract to deorbit the International Space Station no earlier than 2030.