Same input. Same prompt. Different output. That's the reality of testing AI agents that write code, and most teams are shipping without solving it.Nick Nisi from WorkOS tackled this by building eval systems for two AI tools:- npx workos@latest, a CLI agent that installs AuthKit into your project- WorkOS agent skills that power LLM responses about SSO, directory sync, and RBAC.The post covers how to test against real project structures, score output that's different every time, and catch when your agent makes up methods that don't exist.Learn more about evals →

Meta tracks its employees' keyboard inputs and mouse movements to train its AI models. Many workers are uncomfortable with the scheme, which they are unable to opt out of on their corporate laptops. The company says there are safeguards in place to protect sensitive content and that the data is not used for any other purpose, but many workers have expressed anger at the policy. Many employees no longer see Meta as a place for a long career.

OpenAI recently allowed employees to sell up to $30 million worth of shares each. The sale offers a peek into the flood of money that will soon hit America's tech capitals. OpenAI and Anthropic are gearing up for what will likely be some of the largest IPOs in history. The newfound wealth is driving up rental prices in San Francisco and sparking concerns about a growing class divide within the city.

Japanese astronomers have discovered a small world about 3.5 billion miles from the Sun that appears to be swaddled in a layer of air. The discovery of an atmosphere is surprising, as the 300-mile-wide celestial body shouldn't have a strong enough gravitational pull to keep it in place. Also, temperatures in that quadrant are so cold that most of the molecules that exist as gases in Earth's atmosphere freeze solid. The atmosphere is very thin, with only a few gas molecules to exert any pressure.

Sperm may be more than just vessels for DNA. Scientists have detected microRNAs and other types of RNA fragments in sperm cells that respond to various environmental factors, such as stress, exercise, and mental health. There is mounting evidence that sperm may be transmitting environmental information to offspring. There are many unanswered questions that scientists are still exploring, but there is enough evidence to rebalance parental responsibility.

Markdown is the dominant file format used by agents, but it is a bit restrictive. HTML can convey much richer information compared to Markdown, making it a better language for model communication. It also allows for interaction and data ingestion and makes files easier to read and share. This post provides some tips on how to get started with getting Claude to generate HTML files.

You have to manage maintenance costs if you want a productive team. AI may help teams produce code faster, but maintenance costs can build up exponentially if teams don't manage the code properly. Even if AI produces code that's just as easy to maintain as human-written code, the productivity gains don't last. Chase improvements to coding speed, but spend just as much time chasing improvements to maintenance costs as well.

Intel will start manufacturing some of the chips that power Apple devices again. It is unclear which products Intel will make chips for. Apple has long been TSMC's top customer, but soaring demand means that Apple no longer has the leverage to secure the supplies it needs. CEO Tim Cook has blamed the lack of availability of advanced chips for the company's inability to meet customer demand for iPhones.

River is an AI agent that lives in Shopify's company Slack. It can read code, run tests, write code, open pull requests, query data, look up production traces, and a lot more. River only works in the open - it doesn't respond to direct messages. This helps employees learn from each other faster and more effectively than if they were using the agent privately. The agent's success comes from Shopify's employees watching each other interact with it.

Instant SSH VMs, persistent storage, and auto-HTTPS. Skip the infra setup; just ship. Spin up your next service at exe.dev. Get started.

There are many jobs where you learn first and then apply, but in research and development, you always begin with an incomplete understanding.

TLDR is hiring a VP of Marketing to scale growth, pipeline, and brand across 8M+ subscribers, 13 newsletters, and advertisers like AWS, Google Cloud, Microsoft, and Anthropic. Apply here.

The tools will be for developers, not for end users, and they will focus on models that run locally to preserve users' privacy.

AI's ability to find and fix bugs is breaking the way computer security is commonly approached.

Some tasks you can leave to agents to work autonomously on, but work that requires taste and originality requires human involvement.

The average computer has become increasingly capable, so work should be done in the browser rather than being offloaded to a remote computer.

Google's next-generation reCAPTCHA system relies on Play Services for verification.