TLDR 2025-05-26

Authentication isn't just login. It's building secure flows, managing sessions, handling tokens, supporting SSO, and protecting user data. AuthKit provides a fully hosted, customizable UI for sign-up, password resets, MFA, and seamless SSO — production-ready from day one.

But authentication alone doesn't stop abuse. WorkOS Radar adds real-time threat prevention, detecting brute force, credential stuffing, and more. Built-in protections can be enabled instantly — no scripts or custom logic required.

Build secure, scalable authentication from day one with AuthKit and WorkSO Radar.

Nvidia's DGX Cloud Lepton is a new service designed to link artificial intelligence developers with Nvidia's network of cloud providers, including CoreWeave, Lambda, and Crusoe. It provides access to Nvidia's graphics processing units, making those chips more widely accessible to all developers. The platform will have a marketplace of GPU cloud vendors that developers can pick from to train and use their AI models. Developers can choose which cloud provider to work with and even work with multiple cloud providers.

Demis Hassabis, the chief executive of Google DeepMind, has been dreaming of artificial general intelligence (AGI) for years. He was already taking the possibility of AGI seriously when Google acquired DeepMind in 2014. Today, he is among a growing number of tech leaders racing to build AGI. Hassabis and a colleague received the Nobel Prize in Chemistry last year for their work on AlphaFold, an AI system capable of predicting the three-dimensional structures of proteins. This article contains a transcript of an interview with Hassabis about his views on AGI and the future that follows its arrival.

Autonomous robots aren't being used to operate on people yet, but the technology is almost capable of such a feat and more autonomy is on the way. The technology could help hospitals meet the high demand for safety and consistency during surgeries, manage routine tasks, and prevent mistakes. Future robots will use AI to autonomously manage movements and decision-making during surgical tasks, reducing the need for constant human input. With the number of surgeons around the world declining while the number of people who need surgery continues to increase, the need for the technology is clear.

Starfish Neuroscience, a brain-computer interface startup incorporated by Valve co-founder and CEO Gabe Newell, plans to produce its very first brain chip later this year. The company is currently seeking collaborators to open up avenues for the technology. Starfish's goal is to create a smaller and less invasive implant than the competition that can enable simultaneous access to multiple brain regions. Details from the chip's current spec sheet are available in the article. The article also links to a video of a talk from Valve that explains how the technology could enhance gaming.

Will Larson worked at Carta as its CTO for the past two years. This post contains the most important lessons Larson learned in his time at the company, covering topics such as engineering strategy, large language model adoption, multi-dimensional tradeoffs, creating software quality, and more.

Bohan Zhang, a member of OpenAI's Infrastructure team, recently shared the company's best practices with PostgreSQL at the PGConf.dev 2025 Global Developer Conference. The talk offers a glimpse into the database usage of one of the most prominent unicorn companies and demonstrates how PostgreSQL can scale gracefully under massive read loads. OpenAI utilizes managed databases on Azure, employing a classic PostgreSQL primary-replication architecture without sharding. This article covers the content of Zhang's talk along with comments and includes some more details about OpenAI's infrastructure.

As AI spreads through the labor force, many white-collar workers say their work is becoming more routine, less thoughtful, and much faster-paced. Companies seem to be persuaded that AI can increase productivity. Teams are getting smaller, but they're being expected to output the same amount of work by using AI. Managers say that AI can relieve employees of tedious tasks and enable them to perform more interesting work, but for inexperienced workers, the change resembles the shift from artisanal work to factory work in the 19th and 20th centuries.

Large language models are becoming commodities. Many people use them, but few people are developing them from scratch. Developing large language models involves a lot of hardcore computer science and math - people can use them without having to know any of the internals. While it might be helpful to understand some of the internals and math, it's probably best to stick to learning how to get good at using these models and integrating what they do into agentic systems.

We also write TLDR AI. Same format, all about LLMs, GenAI, and deep learning. Delivered every weekday. Subscribe for free.

F2 is a cross-platform command-line tool for batch renaming files and directories quickly and safely.

MCP could inspire a new generation of developers to push for the web to go back towards its natural architecture, programmable through janky specs that everybody adopted for the sheer joy of something fun to hack on.

TrailBase is an application server with sub-millisecond latencies, eliminating the need for dedicated caches.

Rooting for your friends is the best way to improve your career, happiness, and the people you care about.

You should learn from others, but you also need to put your knowledge to the test so you learn enough about your field to advance it.

This post presents things to consider if you are thinking about creating a new file format.