Search the archive

GOOGLE IS ROLLING OUT GEMINI'S REAL-TIME AI VIDEO FEATURES

Google Gemini's new AI features let it see users' screens or through their cameras and answer questions about either in real time.

ANOTHER OBSERVABILITY 3.0 APPEARS ON THE HORIZON

A NEW, CHALLENGING AGI TEST STUMPS MOST AI MODELS

The ARC-AGI tests consist of puzzle-like problems where an AI has to identify visual patterns from a collection of different-colored squares and generate the correct “answer” grid - reasoning AI models like OpenAI's o1-pro and DeepSeek's R1 score between 1% and 1.3%.

OPENAI EXPANDS COO BRAD LIGHTCAP'S JOB TO INCLUDE BUSINESS OVERSIGHT, AS ALTMAN FOCUSES ON RESEARCH

Brad Lightcap joined OpenAI in 2018, having previously worked with Sam Altman at Y Combinator.

TESLA AIMING TO PRODUCE FIRST “LEGION” OF OPTIMUS ROBOTS THIS 2025

Tesla aims to start producing Optimus this year. The company plans to produce enough parts to make 10,000 to 12,000 Optimus robots this year and 50,000 units next year. The product is totally new, so Tesla's measure of success is to produce at least 5,000 robots. The article includes a video of Tesla's Q1 2025 All-Hands Meeting, during which Elon Musk discusses the project.

APPLE WANTS TO TURN ITS WATCHES INTO WEARABLE AI

Apple plans to turn the Apple Watch into an AI-infused wearable. It is working on a version of the smartwatch equipped with cameras, which will allow the device to do things like identify objects or translate text in a picture. Apple has had issues with implementing AI features and its past attempts at expanding the functionality of the Apple Watch have not gone well. The company is reportedly close to ditching a plastic version of the Apple Watch SE that was supposed to help bring the price of the device down.

1X WILL TEST HUMANOID ROBOTS IN ‘A FEW HUNDRED' HOMES IN 2025

1X plans to start early tests of its Neo Gamma humanoid robot in up to a few thousand homes by the end of 2025. The company will be initially relying on teleoperators - customers will be able to decide when a 1X employee can view Neo Gamma's surroundings. The in-home tests will allow the company to collect data on how the robot operates to create a dataset for training AI models and upgrading the robot's capabilities. A video of the robot vacuuming carpet during a demo at a show is available in the article.

POST 50: GOOD RESEARCH TAKES ARE NOT SUFFICIENT FOR GOOD STRATEGIC TAKES

Having a good research track record is evidence of good big-picture takes about AGI, but it's weak evidence. Strategic thinking requires different skills - people often conflate these skills, leading to excessive deference to researchers without evidence of whether that person is good at strategic thinking specifically. Having good strategic takes is important and researchers should spend a fair amount of time developing the skill. People should be skeptical about people's strategic takes until they have proven they have the skill to make good strategic takes.

Get the Datadog Azure OpenAI Cheatsheet

REVENGE OF THE JUNIOR DEVELOPER

There will still be plenty of jobs in the software industry in the future - just not the kind that involves writing code by hand. The world is changing, so junior developers are adapting. Junior developers have been far more eager to adopt AI than senior developers. Developers who don't adopt the new technology have lost.

TTYD

ttyd is a command-line tool for sharing terminal over the web. It provides a fully-featured terminal with CJK and IME support, ZMODEM/trzsz file transfer support, Sixel image output support, and more. ttyd is cross-platform and works on macOS, Linux, FreeBSD/OpenBSD, OpenWrt, and Windows. Only the latest versions of Chrome, Edge, Firefox, and Safari are supported.

THE STRUGGLE THAT MADE US IN WATERLOO

The power in the software industry is transferring from the employee to the employer, causing many seasoned professionals to question their self-worth and students to worry. Being a student back amid the global financial crisis was a stressful experience, with the pressures causing many students to drop out every year. It didn't get easier for students - they just got better at playing the game. Current students need to stay strong and keep going. School can be challenging - the experience will change you in ways you won't understand until much later, but when you finally do, you'll look back and know it was worth it.

META HAS REVENUE SHARING AGREEMENTS WITH LLAMA AI MODEL HOSTS, FILING REVEALS

While Mark Zuckerberg claims that selling access to Meta's Llama AI models isn't in Meta's business model, the company does at least make some money from the models through revenue-sharing agreements. A court filing has revealed that Meta shares a percentage of the revenue that companies hosting its Llama models generate from users. These companies may include AWS, Nvidia, Databricks, Groq, Dell, Azure, Google Cloud, and Snowflake.

PERPLEXITY AI SAYS IT WOULD REBUILD TIKTOK'S ALGORITHM AND ADD COMMUNITY NOTES FEATURES

Perplexity AI has outlined its vision for TikTok in detail publicly for the first time.

MOST AI VALUE WILL COME FROM BROAD AUTOMATION, NOT FROM R&D

Research and development is generally not as economically valuable as people assume and it's significantly harder to automate R&D jobs than it might naively seem.

I WANT A GOOD PARALLEL COMPUTER

There have been some promising parallel computer designs in the past, but they have failed for various complex reasons, leaving us with overly complex and limited GPUs and AI accelerators.

THE BITTER LESSON: RETHINKING HOW WE BUILD AI SYSTEMS

The AI engineer's role is to design systems that can effectively utilize increasing compute resources, build robust learning environments that scale, and create architectures that can grow without requiring fundamental redesigns.

GOOGLE'S RACE FROM BEHIND AGAINST OPENAI HAS 'BEEN LIKE SPRINTING A MARATHON'

Google's Gemini AI models are now among the top-ranked chatbots on community-driven platform Chatbot Arena.

A CLOSER LOOK AT DYNAMO, NVIDIA'S 'OPERATING SYSTEM' FOR AI INFERENCE

Nvidia's Dynamo is tuned for Nvidia's hardware and software stacks - it will run on any Nvidia GPU going back to Ampere.

AN INTERVIEW WITH OPENAI CEO SAM ALTMAN ABOUT BUILDING A CONSUMER TECH COMPANY

OpenAI is clearly one of the defining companies of this era. Its founder, Sam Altman, recently appeared on a podcast interview that covered his background, the OpenAI origin story, and the genesis of ChatGPT. He also gave some hints about what's coming next, answered some philosophical questions about AI, and provided advice for seniors graduating from high school. A link to the audio for the podcast is available.

APPLE'S NEW SHAKE-UP PROVES SIRI'S NOW BEING TAKEN SERIOUSLY, HERE'S WHY

Former Apple Vision Pro head Mike Rockwell is now in charge of Siri. While Rockwell's experience doesn't make him a natural choice, it shows that Apple has decided that Siri deserves attention from a top-tier Apple leader. Rockwell has developed and launched major hardware and software platforms for Apple. The new leadership represents a big shift in how much the company is valuing and prioritizing Siri.

BOSTON DYNAMICS SHOWS OFF ANOTHER MAJOR LEAP IN HUMANOID MOBILITY

Boston Dynamics has released a new video (available in the article) that shows its Atlas robot running, cartwheeling, and breakdancing. Atlas' walking now resembles walking rather than taking a bunch of steps - its other movements are starting to look less forced too. The current explosion in humanoid robotics is still at a very early stage, but it is looking clearer than ever that humans and androids will soon be interacting regularly in daily life. Chinese company Unitree has also been doing incredible work with its lightweight G1 humanoid, which has a starting price tag of $16,000.

INSIDE ‘ECCENTRIC' RIPPLE FOUNDER'S MULTIBILLION-DOLLAR SPACE STATION PLAN

Jed McCaleb, the serial entrepreneur who founded Mt. Gox and Ripple, has a new company, Vast, that is on track to launch a commercial space station into orbit by May 2026. Vast is building its spacecraft with components developed by SpaceX. If it succeeds, it will be better positioned to win a lucrative contract from NASA to replace the International Space Station - if it fails, McCaleb could see $1 billion wiped from his net worth. Vast is competing with Axiom Space, Voyager Space Holdings, Lockheed Martin, and Blue Origin to win the NASA contract.

ACCELERATING LARGE-SCALE TEST MIGRATION WITH LLMS

Airbnb recently completed its first large-scale large language model-driven code migration. It updated nearly 3,500 React component test files from Enzyme to use React Testing Library instead. This was estimated to take 1.5 years of engineering to do by hand, but Airbnb managed to finish the entire migration in just six weeks using a combination of frontier models and robust automation. This blog post looks at the challenges the Airbnb team faced during the migration, how large language models helped solve this particular type of challenge, and how the team structured its tooling to run an AI-driven migration at scale.

LIFE ALTERING POSTGRESQL PATTERNS

There are several things you can do when working with Postgres databases that can make the experience much more pleasant. These include using UUID primary keys, naming tables singularly, and representing statuses as logs. This post presents a list of tips for working with Postgres. Each tip is backed up by a reason for why it is helpful and code examples.

PERPLEXITY'S CEO HAS AN AMBITIOUS PLAN TO TAKE ON GOOGLE

Perplexity AI is ramping up its rivalry with Google on multiple fronts. It has expanded its research options for businesses, created tools to look up financial data, expanded a partnership with SoftBank for its sales team to resell Perplexity's premium products, and made a long-shot bid to merge with TikTok. Perplexity's CEO, Aravind Srinivas, envisions a world where Perplexity serves as a central hub for more sophisticated and affordable AI tools that can generally act as a personal assistant for users. This article contains an interview with Srinivas where he discusses how he's growing Perplexity's enterprise business, why he's bullish on DeepSeek, and the thinking behind his TikTok bid.

APPLE LOSES $1B A YEAR ON PRESTIGIOUS, MINIMALLY VIEWED APPLE TV+

The Apple TV+ streaming service is losing more than $1 billion annually. The service's subscriber count reached around 45 million in 2024. Apple TV+ typically spends over $5 billion annually on content, but it cut its budget by about $500 million last year. It is reportedly the only part of Apple's services business that is not turning a profit. While its library boasts prestige, its viewership is scant, hindering its ability to make money.

ANUBIS

Anubis is a tool that tests connections using a sha256 proof-of-work challenge to protect upstream resources from scraper bots.

A DEEP DIVE INTO MCP AND THE FUTURE OF AI TOOLING

MCP is an open protocol that allows systems to provide context to AI models in a manner that's generalizable across integrations.

ANTHROPIC ADDS WEB SEARCH TO ITS CLAUDE CHATBOT

Paid Claude users in the US can now access web search in preview - support for free users and additional countries is coming soon.

FETCHER MCP

Fetcher MCP is a model context protocol that fetches web page content using Playwright.

OPENAI UPGRADES ITS TRANSCRIPTION AND VOICE-GENERATING AI MODELS

OpenAI will not be making its new transcription models openly available, unlike Whisper.

AUTOMATING MATH

AI models will soon find utility as assistants for many mathematicians, but the extent to which this will increase the rate of true math breakthroughs is unknown.

Attend in person ($295) [5] or register for the livestream (free)

OPENAI'S O1-PRO IS THE COMPANY'S MOST EXPENSIVE AI MODEL YET

OpenAI has launched a more powerful version of its o1 reasoning model in its developer API. o1-pro uses more computing to provide better responses. It costs $150 per million input tokens and $600 per million output tokens - ten times the price of regular o1. OpenAI's internal benchmarks show that while o1-pro answers coding and math problems more reliably, it only performs slightly better than the standard o1.

GOOGLE'S BUDGET-FRIENDLY PIXEL 9A IS FINALLY HERE, AND IT COMES WITH A BOLD NEW LOOK

Google has finally launched the Pixel 9a. Starting at $499, the device, which comes in four colors (Peony, Iris, Porcelain, and Obsidian), is powered by the Google Tensor G4, Google's most efficient processor so far. The Pixel 9a has a 6.3-inch display with 2,700 nits of peak brightness and a 120 Hz adaptive refresh rate. It has a battery life of 30 hours - over 100 hours with Extreme Battery Saver enabled. The device has a 13 MP ultra-wide camera and a 48 MP main camera. It is the first Pixel A series device that features Macro Focus.

AMAZING NEW TECHNOLOGY CAN 'BEND' SOUNDS INTO YOUR EARS ONLY

Scientists have discovered a way to create localized pockets of sound isolated from their surroundings, creating sound exactly where it needs to be. The technique involves using ultrasound as a carrier for audible sound - ultrasound transports sound through space silently, becoming audible only when desired. The ultrasonic beams can bend as they travel, allowing for the creation of curved sound paths that can navigate around obstacles and meet at a specific target location. Many more details about how this technology works are available in the article.

NOT EVERYONE IS CONVINCED BY MICROSOFT'S TOPOLOGICAL QUBITS

Microsoft's quantum team made waves last month by announcing its first topological quantum chip, the Majorana 1. Quantum computers run on qubits valued at 0, 1, or some superposition of the two - topological quantum computing is a fundamentally different approach to building a qubit that would be theoretically less fragile. Proving the claims that it has created at least a single topological qubit remains difficult for the Microsoft team. Despite difficulties, topological quantum continues to be a very promising approach.

MAKING AI WORK IN LEGACY CODEBASES

AI can help engineers even when code bases are big and complex. This article contains several examples of how to use AI on legacy code bases. The examples contain prompts, processes, tool tips, links to resources, and more for better understanding.

NOT ALL AI-ASSISTED PROGRAMMING IS VIBE CODING (BUT VIBE CODING ROCKS)

Vibe coding is building software with a large language model without reviewing the code it writes. Making an AI write code, then reviewing that code and testing it thoroughly is not vibe coding - that's just software development. Vibe coding could grant millions of new people the ability to build their own custom tools. It also has a ton to offer experienced developers - it can help them build an intuition as to what AI can and cannot do for them.

ELON MUSK IS JOINING MICROSOFT IN $30 BILLION DATA CENTER PROJECT

Microsoft, BlackRock, and MGX are partnering with xAI on a $30 billion project to develop data centers and other infrastructure. The details of the new deal are unclear. The relationship between Microsoft and OpenAI has increasingly cooled as Microsoft has loosened its reliance on OpenAI and developed its own in-house AI models. OpenAI recently announced its own $100 billion infrastructure project with Oracle and SoftBank - an early sign that the companies were starting to drift apart.

AI IS "TEARING APART" COMPANIES, SURVEY FINDS

Executives across the US have been pushing AI as an inevitable revolution, but workers aren't buying it. Nearly all C-suite executives are unsatisfied with their company's current AI solution - over half say they're looking for new jobs with companies that are more innovative with generative AI. While almost all C-suite executives believe their companies have an AI strategy, just over half of employees say their companies have an AI strategy at all. The pushback from employees seems to stem from both fear of being replaced by AI and AI tools not being suited for work.

with DeepSeek R1. Read the blog

THE DOCUMENT CULTURE OF AMAZON

Amazon uses documents in a way that is incredibly unique - this article shares the process and explains how to implement document-based meetings in the workplace.