TLDR 2025-05-08
Apple AI search ๐ฑ, Stripe stablecoin accounts ๐ฆ, McKinsey vs startups ๐ผย
Stripe Brings Stablecoin Accounts to More Than 100 Countries (2 minute read)
Stripe is launching USDC and USDB accounts in 101 countries. The new stablecoin-funded accounts will help Stripe facilitate cross-border transactions between businesses. The accounts will also support balances held in British pounds, euros, and US dollars. Stripe has also announced a profiles feature to help businesses find each other and verify their identities and a tool to make it easy for developers to embed artificial intelligence agents into their products.
Apple will add AI search in mobile Safari, challenging Google (2 minute read)
Apple plans to shift the focus of mobile Safari's search experience to AI search engines. Searches in Safari fell for the first time ever last year due to people increasingly using large language model-based solutions. Google and Apple shares dropped after Apple's plans were revealed as investors fear that the deal that makes Google the default search engine may not last. It is unlikely an AI search engine will become the default as the technology still needs to improve, particularly when it comes to indexing.
๐
Science & Futuristic Technology
Google agrees to fund the development of three new nuclear sites (3 minute read)
Elementl Power has signed an agreement with Google to develop three project sites for advanced reactors. Each site will generate at least 600 megawatts of power capacity. Google will have the option to buy the power once the sites are up and running. Elementl has yet to choose the type of reactor it will use. It plans to choose the reactor technology that's furthest along in development when it is ready to begin construction.
Amazon says new warehouse robot can โfeel' items, but won't replace workers (5 minute read)
Amazon's new warehouse robot has a sense of touch, allowing it to handle jobs previously only done by humans. Vulcan has an innovative gripper that can manipulate 75% of the one million unique items in inventory at Amazon's Spokane warehouse. Vulcan will not be used to replace people - Amazon will create new, higher-skilled jobs. The robot can handle items up to 8 pounds and operate for 20 hours a day.
๐ป
Programming, Design & Data Science
The magic of software; or, what makes a good engineer also makes a good engineering organization (10 minute read)
Many of today's best practices were drawn from long-established internet companies like Google. However, the problem with copying its current practices on the basis of its success is that most of those companies found near-invincible business models, and so almost any practice developed or selected at random likely would continue to be successful to some degree. The practices and cultures that have developed at these companies more often reflect optimizations for the benefit of various decision makers rather than optimizations for actually creating something great. The qualities that make a good engineer are often the same ones that create good engineering organizations - both start with deep understanding as the basis for innovation and recognize that vision and engineering are often deeply intertwined and mutually informative rather than linear.
Figma introduces โvibe-coding' AI software design feature (3 minute read)
Figma Make is an artificial intelligence feature that automates the process of building websites and applications. It will be available starting at $16 per person per month when purchased annually - there is no free tier. The tool uses Anthropic's Claude 3.7 Sonnet model. It is currently being tested. There is no information on when it will be released.
Know Your Enemy: How Three Years at McKinsey Shaped My Second Startup (7 minute read)
Zac Townsend was an Associate Partner at McKinsey. This post talks about the reasons he joined the company, his time there, the projects he worked on, and the lessons he learned from the experience. After working there for three years, Townsend learned that it is impossible to fix institutions that have gotten too big - there are too many people, too many processes, and too much regulatory overhang. This led him to create his own company.
Working with LLMs: A Few Lessons (7 minute read)
Large Language Models (LLMs) are inherently probabilistic, which means you have to add evaluation frameworks, human-in-the-loop processes, and design for graceful failure, using LLMs for probabilistic guidance rather than deterministic answers. Unlike with traditional software, there is no way to get better at using AI than using AI - there is no substitute for trial and error. This makes it hard to predict development and makes building future-proof technology almost impossible. While software is inexpensive to distribute once written, LLMs have continued costs.
Get the most interesting stories in startups, tech, and programming delivered in a free daily email.
Join 1,600,000 readers for
one daily email