Search the archive

leaders.

SILICON VALLEY CAN'T IMPORT TALENT LIKE BEFORE. SO IT'S EXPORTING JOBS

Hiring by Big Tech has risen sharply in India in recent months.

HOW STAYING SMALL BECAME AI STARTUPS' BIGGEST FLEX

AI tools allow startups to run leaner than ever, spawning new benchmarks like 'revenue-per-employee'.

INSIDER AMNESIA

Experts correctly disregard bad sources in their fields of expertise, but trust those same sources on other topics.

FULLY REMOTE)

As the first engineer on TLDR's new Applied AI team, you'll build AI agents and composable Claude Skills to let non-technical teammates create their own AI workflows. Learn more [23].

MICROSOFT EXECS WORRY AI WILL EAT ENTRY LEVEL CODING JOBS

Two Microsoft executives have written a paper arguing that senior software engineers must mentor junior developers to prevent AI from hollowing out the profession's future skills base.

RED/GREEN TDD

Every good model understands that 'use red/green TDD' means to write automated tests, confirm they fail, then iterate on the implementation until the tests pass.

DICTIONARY COMPRESSION IS FINALLY HERE, AND IT'S RIDICULOUSLY GOOD

Dictionary compression can shrink YouTube JavaScript download size for returning desktop users up to 90% compared to existing best-practice compression.

Optimize your workload now

CAN OPENAI BUILD ALEXA BEFORE AMAZON CAN BUILD CHATGPT?

OpenAI and Amazon are currently racing to execute an AI device strategy. Interestingly, Amazon reportedly plans to invest $50 billion into OpenAI's newest funding round. Amazon is also the largest shareholder in OpenAI's chief startup rival, Anthropic. New reports suggest that OpenAI may be building custom models for Amazon's products, even while it is hard at work on its first devices.

APPLE'S NEXT BIG THING IS A PUSH INTO VISUAL ARTIFICIAL INTELLIGENCE

The defining feature of Apple's push into wearable AI devices will be Visual Intelligence. Apple CEO Tim Cook has been dropping hints about AI wearables built around AI technology that can see the surrounding environment and use that context to take action. The company is working on its own visual models with the intention of making them central to its upcoming set of AI devices, which include advanced AirPods, smart glasses, and a pendant. Apple plans to hold its first product launches of the year in March.

HAVE WE LEAPT INTO COMMERCIAL GENETIC TESTING WITHOUT UNDERSTANDING IT?

Polygenic scores tally the contributions of many genes to particular traits to predict certain outcomes. They are probabilistic, not deterministic, as many traits are determined by environmental factors as well as genes. Potential parents using IVF are now able to choose embryos based on their polygenic scores. While it is understandable that parents want to give their kids the best chance of success, the practice may have unintended consequences on the human gene pool, not to mention the ethical implications.

AI DATA CENTERS TURN TO HIGH-TEMPERATURE SUPERCONDUCTORS

Microsoft is exploring the potential use of high-temperature superconductors (HTS) as a replacement for copper wiring. HTS could improve energy efficiency by reducing transmission loss. It could also increase the resiliency of electrical grids and limit the impact of data centers on communities by reducing the amount of space required to move power. Microsoft has invested $75 million into Veir, a superconducting power technology developer, to advance the technology.

CODE MODE: GIVE AGENTS AN ENTIRE API IN 1,000 TOKENS

Cloudflare has introduced a new MCP for the entire Cloudflare API that uses Code Mode. The server can provide access to the entire Cloudflare API over MCP while consuming only around 1,000 tokens. Code Mode is a technique for reducing context window usage during agent tool use. It involves letting the model write code against a typed SDK and executing the code safely in a Dynamic Worker Loader instead of describing every operation as a separate tool. The code acts as a compact plan, and the model can explore tool operations, compose multiple calls, and return just the data it needs.

HOW OPENAI'S CODEX TEAM WORKS AND LEVERAGES AI

OpenAI's Codex team has reimagined how an engineering organization can function in an AI-native world. The team operates with a high level of trust, autonomy, and speed. Decision-making is fast, ownership is clear, and individuals are empowered to build end-to-end. The team has truly embraced AI as a way to be more productive. AI is treated as a first-class teammate that shapes planning, execution, review, onboarding, and iteration. This allows the team to move with startup-level velocity while working on some of the most complex systems in the world.

ENTIRELY FORESEEABLE AWS OUTAGES

Amazon Web Services experienced at least two minor outages at the end of last year caused by malfunctions in its internal AI tooling. These outages were entirely foreseeable, even by employees. Large language models are extremely complex systems, and this complexity creates real risk. Amazon never misses a chance to point to 'AI' when it is useful, but when there's an outage, suddenly it's just a 'coincidence' that AI tools were involved.

ATOMS ARE CHEAP, PROCESS IS PRICEY

Fast feedback loops allow you to get ideas to reality more quickly. This pattern is seen in every successful frontier tech effort. The common thread is treating reality as the teacher and getting to class as often as possible. Iterate fast, and validate using reality instead of thinking your way to perfect solutions.

MINIONS: STRIPE'S ONE-SHOT, END-TO-END CODING AGENTS—PART 2

Stripe's Minions are a great example of how to blend industry-standard concepts with internal tooling and infrastructure to maximize developer productivity.

MICROSOFT GAMING CHIEF PHIL SPENCER STEPS DOWN AFTER 38 YEARS WITH COMPANY

Asha Sharma, the executive currently in charge of Microsoft's CoreAI division, will take his place.

HALF THE AI AGENT MARKET IS ONE CATEGORY. THE REST IS WIDE OPEN

Software engineering accounts for nearly half of all AI agent tool calls, while fields like healthcare, legal, and finance are barely touched.

HOW I THINK ABOUT CODEX

Codex is comprised of a model that provides intelligence, a harness that turns that intelligence into something that can safely operate in a real development environment, and runtimes and interfaces where agents can be put to work.

DESIGN DOCS CONSIDERED HARMFUL

You have the least amount of information at the beginning of a project, which is when design docs ask you to make the most decisions.

PEOPLE LOVED THE DOT-COM BOOM. THE AI BOOM, NOT SO MUCH

The public is showing underwhelming enthusiasm for the tech industry's plans to remake the world with artificial intelligence.

Build faster with WorkOS →

GOOGLE ANNOUNCES GEMINI 3.1 PRO, SAYS IT'S BETTER AT COMPLEX PROBLEM-SOLVING

Google's new Gemini 3.1 Pro is now rolling out in preview. The model's benchmark results show mostly modest improvements. The updated model is now available in AI Studio and the Antigravity IDE. Enterprise users will see the model in Vertex AI and Gemini Enterprise, and regular users can use it in both the Gemini app and NotebookLM. The API costs for developers have not changed, nor has the context window.

AMAZON DETHRONES WALMART AS WORLD'S BIGGEST COMPANY BY SALES

Amazon is now the biggest global company by revenue. Walmart was previously the leader for more than a decade. Jeff Bezos had carefully studied Walmart's founder, Sam Walton, and embedded many of his business strategies while building Amazon. Amazon's revenue increased at almost 10 times the pace of Walmart's over the past decade. Amazon is the biggest online retailer, attracting 2.7 billion visits to its websites and mobile apps each month, while Walmart is the biggest physical retailer in the world, with more than 10,000 stores and shopping clubs globally.

THE FICKLENESS OF SCALING LAWS

Scaling laws have defined tech in the last five years. Large language models have demonstrated consistent returns to scaling. However, these 'laws' apply to very few other fields. Little work has been done to systematically understand, in theory or in general, what causes scaling laws or determines their slope and robustness. This article looks at what causes scaling laws, what their preconditions are, when they stop, why they have worked in some fields but not others, and more.

RESEARCH NOTE: FIVE LESSONS FROM HAVING HELPED RUN AN AI-BIOLOGY RCT

Almost everyone agrees that AI policy should be evidence-based. However, science is often messy and full of caveats, which doesn't fit well with policymakers' demands for clean answers. Many AI benchmarks have saturated and can no longer give a clear 'no' on risks. Experiments show how much we still have to learn. We need to spend less time proving that today's AIs are safe and more time figuring out how to tell if future AIs are dangerous.

went down during the big launch. Build without limits with Azure

BUILDING AN ELITE AI ENGINEERING CULTURE IN 2026

The winning formula in early 2026 is structured context, tiered rigor, smaller teams with higher leverage, and relentless measurement of downstream effects. The teams that adapt AI tools without the underlying taste and discipline are discovering that AI just makes their existing problems bigger. Teams with strong engineering cultures, robust CI/CD, clear architectural standards, and effective review processes see AI compound their advantages. This is widening the gap between elite and average teams.

STOP THINKING OF AI AS A COWORKER. IT'S AN EXOSKELETON

AI should be treated as an amplifier of human capacity rather than a replacement. AI can still act autonomously with specific tasks, but in a way that is an extension of human decision-making and context. This allows humans to do dramatically more work, more sustainably, with less injury and fatigue. The AI industry's dream of having a fully autonomous AI employee is seductive, but it will likely set us up for disappointment.

HOW WILL OPENAI COMPETE?

OpenAI's technology isn't unique. While it has a big userbase, there is limited engagement and stickiness, and no network effects. Its competitors have matched its technology and are leveraging their product distribution. Executing better than everyone else is certainly an aspiration, but it isn't a strategy.

BYTEDANCE BUILDING OUT ARTIFICIAL INTELLIGENCE TEAM IN US

ByteDance is hiring nearly 100 employees for its AI division in the US. The company's AI team, called Seed, was established in 2023 and has labs across the US, Singapore, and China. The open roles include various job responsibilities such as producing international data, advancing ByteDance's text, image, and video generation tools, AI research, and building models for drug discovery and design. The company's recent AI launches have thrust it into the spotlight in the US. There has been some controversy over the data used to train its models.

with less open tabs. Simplify with Azure

META'S METAVERSE IS GOING MOBILE-FIRST

Meta is separating its Quest VR platform from its Worlds platform, shifting the focus of Worlds to be almost exclusively mobile.

THEFT OF TRADE SECRETS IS ON THE RISE—AND AI IS MAKING IT WORSE

There were around 1,500 federal trade secrets cases filed in the US last year, up 20% from the previous year.

THERE IS NO PRODUCT

Software that can be generated in a weekend is disposable inventory, not something you can build a startup around.

WHITE-COLLAR APOCALYPSE ISN'T AROUND THE CORNER—BUT AI HAS ALREADY FUNDAMENTALLY CHANGED THE ECONOMY

The productivity gains from AI are real, but it won't be catastrophic, as it will just devalue mechanical skills while elevating judgment skills.

BUILDING WITH AI AGENTS

Nine observations from building with AI agents over the last year.

THE SOFTWARE SHAKEOUT: WHAT IS DURABLE AND WHAT IS NOT IN THE AGE OF AI?

The most value has always historically accrued to the companies that did the hardest things.

Schedule a demo to learn more

META TO CHALLENGE APPLE WITH ITS FIRST SMARTWATCH — AND IT'S REPORTEDLY LAUNCHING THIS YEAR

Meta plans to release its first smartwatch sometime this year. Code-named Malibu 2, the smartwatch will offer health-tracking tools along with a built-in Meta AI assistant. A Meta smartwatch could theoretically play nicely with both Android and iOS smartphones, unlike the Apple Watch. A smartwatch has the potential to unlock a lot more gesture-based functionality for Meta's glasses.

MARK ZUCKERBERG SAID HE REACHED OUT TO APPLE CEO TIM COOK TO DISCUSS ‘WELLBEING OF TEENS AND KIDS'

Mark Zuckerberg reached out to Tim Cook in 2018 to discuss the wellbeing of teens and kids. He thought there were opportunities that Meta and Apple could be doing and wanted to talk about them. The email exchange was revealed as part of a trial related to alleged harms of certain digital filters promoting cosmetic surgery. It was used to support the claim that Zuckerberg cares about the wellbeing of teens and kids using Meta's services.

MICROSOFT'S NEW 10,000-YEAR DATA STORAGE MEDIUM: GLASS

Microsoft's Project Silica is a working demonstration of a system that can read and write data into small slabs of glass with a density of over a Gigabit per cubic millimeter. The system can write up to 4.84TB in a single slab of glass of 12cm x 12cm x 0.2cm in about 150 hours. The main draws of the technology are that it takes no energy to preserve data and the data can be retrieved rapidly if needed. The glass also offers extreme stability, with experiments suggesting the data would be stable for over 10,000 years at room temperature.