TLDR 2026-09-21

Anthropic has confirmed that it has a wet biology lab in the Bay Area where it uses AI models to run physical experiments. The lab's main focus is on fundamental biology rather than drug discovery. Anthropic is working on some research in the lab with external partners. The company has launched a Life Sciences Verification program to give vetted bio researchers access to its most powerful models.

Apple is set to introduce a smart home screen as early as next month. The display will serve as a hub for the home, bringing together connected-device controls, video calls, intercom capabilities, media, and access to Apple services and apps. It will feature a new operating system built around Siri AI. The device is currently being tested widely in employees' homes.

China's HL-4 tokamak is designed to achieve steady-state burning operation. It is aiming to become the world's first high-temperature superconducting, high-field steady-state burning experimental platform. The project's goal is to verify the reliability of high-temperature superconducting magnets under complex fusion conditions. Once the core scientific and engineering questions are validated, the program will construct a demonstration reactor to prove the technology commercially.

Elon Musk's Boring Co. venture is working on a tunnel project in Texas to connect Austin and San Antonio. It has also raised $3 billion to support the development of underground transportation infrastructure across the United Arab Emirates. The company is building tunnels in Bastrop, Texas, to link Musk-owned properties and allow the employees of his companies to travel between them.

Join Mike Cannon-Brookes (CEO & Co-Founder, Atlassian), Guillermo Rauch (CEO, Vercel), and leaders from Lovable, Dropbox, and more at State of AI SDLC, a digital summit for engineering and product leaders. Hear how they're adapting development practices, measuring AI's real impact, and working effectively with agents across the software development lifecycle. Save your spot

You never want to be in a situation where people don't know what you are working on. However, instead of talking too much about work, assume everyone around you is operating in good faith and try to be a good teammate. Move away from an outcome-based mindset to a momentum-based mindset - by getting momentum going and fixing your relationships with people, you get stamina. Go for reliability and take the burden off your teammates and manager, because once you build that trust, they will give you bigger work.

AI will make many parts of the current software development environment obsolete. However, the craft of building software will become more important than ever. Models will become so fast that UI will be generated on the fly. It will take a while for this to happen. However, by that time, while people will still be employed to write code, it might not be such a desirable career anymore.

People turn off their brains when they use LLMs. People assume the code they write will just work, and then if it doesn't, they ask the LLM to figure out the problem and solve it. Being a meat proxy is surprisingly effective now. If models keep improving, companies will be able to just run LLMs in a loop and lay off the employee.

Palantir vertically integrated deployment to build the capabilities and process knowledge required to make building institution-specific economic computers as easy to build as possible. Vertical integration is fine if it's the only way to get permissioned access to the ground truth knowledge needed to help customers and society uncover new, valuable secrets. However, used to rent-seek, concentrate power, or build your customers' or partners' replacement, it jeopardizes the whole promise of economic computing: making progress programmable in a way everyone can contribute to and own. It also reduces institutional fitness and creates a broader neglective selection pressure in society.

Tool calling, a 262K-token context window, and complex reasoning across coding, mathematics, and structured outputs. Meet GigaChat 3.5 Reasoning. Explore the model

Society should refuse to outsource moral responsibility to any class of people who claim a special competence and authority to calculate, accelerate, democratize, liberate, harmonize, or otherwise determine the 'correct' future.

If your arguments can justify anything, then your arguments imply nothing.

AI labs are simultaneously competing for mathematical results, selling mathematical assistance, and controlling most of the information required to evaluate either.

Every Firefox fork needs to commit to working together to maintain a hard fork isolated from Mozilla if the independent, non-slop web has any future at all.

China's views on American AI have grown more cynical, but that doesn't mean we should be pessimistic about safety coordination.

GitHub continues to suffer collateral damage from the AI boom.