Search the archive

👉Join a live demo session of Augment →

HOW CHROME DOUBLED ITS SPEEDOMETER SCORES ON ANDROID

Chrome's Speedometer scores on Android have more than doubled since Chrome M112. These improvements came from build optimizations, including a 64-bit Arm build and profile-guided optimization, and enhancements to the V8 JavaScript and Blink rendering engines. There were also optimizations to Chrome's interaction with the operating system and hardware.

BETTER CODE RENDERING THROUGH VIRTUALIZATION

Sentry rebuilt Codecov's code renderer to handle extremely large files that previously caused crashes. The root cause of the crashes was due to React struggling to render a massive number of elements, so the team implemented virtualization using the `@tanstack/react-virtual` library. This improved performance by only rendering visible lines, and after some additional work, the renderer was capable of handling files with up to 500,000 lines of code.

DSQL VIGNETTE: READS AND COMPUTE

Aurora DSQL, AWS's new scalable SQL database, achieves scalability by independently scaling compute, read/write throughput, and storage using a disaggregated architecture that separates compute and storage. This architecture uses Firecracker MicroVMs for each transaction, resulting in dynamic scaling and optimized latency through proximity to clients. Reads are implemented using multiversion concurrency control allowing for strongly consistent, snapshot-isolated transactions without inter-node coordination.

WASHING YOUR CODE: DON'T MAKE ME THINK

Clean, readable code prioritizes clarity over brevity. Overly concise code snippets are difficult to understand. Devs should avoid obscure operators, unnecessary conditions, and shorthand properties that hinder readability.

TRY...CATCH AS LITTLE AS POSSIBLE

This blog post explains why `try...catch` statements should be minimized in code reviews. Instead of using `try...catch` to handle expected errors, potential issues should be proactively checked for and handled before they cause exceptions.

Schedule a demo to learn more 🐺

ASTRO 5

Astro 5.0 has been released, featuring a new Content Layer for managing content from various sources with a unified API and improved performance, and Server Islands to combine static and dynamic content for optimized caching and personalized experiences. The release also simplifies prerendering by merging hybrid and static options into a default static mode and includes a type-safe `astro:env` module for managing environment variables.

OUTERBASE

Outerbase Studio is a browser-based GUI for managing various SQL databases, including SQLite, MySQL, PostgreSQL, and others. It has features like a query editor, data editor, schema editor, and connection manager, with both web and desktop app versions available.

UNDICI

Undici is a high-performance HTTP/1.1 client for Node.js. It has various request methods, body mixins, and benchmark results better than other popular clients.

EDGEPIPES: THE ALTERNATIVE TO SSR AND RSCS

EdgePipes is an alternative to server-side rendering (SSR) and React Server Components (RSCs) that focuses on optimizing network requests, rather than generating complete pages. It works by moving fetch hooks and routing logic to an Edge Function or SharedWorker, creating a "pipe" for streamable responses. This approach uses the improved network conditions between edge servers and backend data centers. Unlike SSR and RSCs, EdgePipes are resource-efficient, avoid rehydration issues, and are compatible with various application architectures.

USING GENERATIVE AI TO PARSE WEB PAGES INTO DATA

GENIE 2: A LARGE-SCALE FOUNDATION WORLD MODEL

Google DeepMind's Genie 2 is a large-scale foundation world model that generates diverse, playable 3D environments from single image prompts for training and evaluating embodied AI agents.

THE CLOUD CONTAINER ICEBERG

This guide explores the range of options for running containers, from mainstream cloud providers to highly specialized and experimental platforms.

THOUGHTS ON TAILWIND 4

Tailwind CSS v4 has great improvements like the switch to LightningCSS and native CSS cascade layers, but the author has concerns about the performance implications of CSS variables, the potential for abuse with new descendant variants, and the lack of clarity in some class names.

See everything that was announced right here →

P.S. Most of these features are included for free with AuthKit up to 1 million MAU 🎉

THE FURIOUS CONTEST TO UNSEAT NVIDIA AS KING OF A.I. CHIPS

Amazon, AMD, and several other startups are starting to offer credible alternatives to Nvidia's chips. Nvidia's rivals are proving they can deliver faster speeds at lower prices for some crucial AI tasks. Many are emulating Nvidia's tactic of building complete computers so that customers can gain maximum power and performance out of their chips. Spending on computers without Nvidia chips by data center operators is expected to grow 49% this year to $126 billion.

META JOINS THE NUCLEAR-POWERED AI FRAY

Meta is turning to nuclear energy to support its data centers and the communities around them. It recently announced a 'request for proposals' for developers that can bring nuclear reactors online starting in the early 2030s. The company is open to many ideas on the size and type of reactors, the locations, and other details. It is seeking initial responses by February 7.

A NEW METHOD TO MAP THE BRAIN: INFECT IT WITH A VIRUS

E11 Bio has been working for the past three years to develop techniques to make brain mapping better, cheaper, and faster. One of these approaches involves making the brain label itself by injecting it with viruses that cause neurons to make special proteins that act like barcodes. The approach reduces the amount of brain slicing required and creates more accurate wiring maps that require less proofreading. E11 is now shifting from proof-of-concept mode to building out its imaging infrastructure so it can tackle an entire mouse brain.

USC STUDENT ROCKET GROUP SHATTERS INTERNATIONAL AMATEUR SPACE RECORD

The USC Rocket Propulsion Lab at the USC Viterbi School of Engineering has broken the international altitude record with its Aftershock II rocket. The rocket reached an altitude of 470,000 feet, higher than any non-governmental and non-commercial group has ever flown before. Aftershock II, at 330 pounds with a height of 13 ft and 8” diameter, reached a velocity of 5,283 ft/s and Mach 5.5. A video of the launch is available in the article.

group demo session to see Loom AI in action

IS AURORA DSQL YOUR NEXT DATABASE?

Amazon's new Aurora DSQL is a serverless, globally distributed database that is highly available, multi-regional, Postgres compatible, and fast. This post looks over its features to see if the product works as advertised. Aurora DSQL is still very new - the service is only available in two US regions at the moment. It is still missing a lot of features from Postgres and uses an unusual authentication method that involves using a URL-like token to connect to the database.

SKIA CANVAS

Skia Canvas is a browser-less implementation of the HTML Canvas drawing API for Node.js. It produces very similar results to Chrome's canvas element. Skia Canvas is best used on desktop machines that can render hardware-accelerated graphics to a window or on a server where it can output a variety of image formats.

THE NUMBER OF AMERICANS WANTING TO SWITCH JOBS HITS A 10-YEAR HIGH

More than half of US workers are watching for or actively seeking a new job. Job satisfaction has fallen to its lowest level in recent years. The cooling job market is causing employees to feel more stuck and frustrated. While employers can worry less about turnover, they risk employee burnout rising if workers feel stuck.

Y COMBINATOR AND POWER IN SILICON VALLEY

AdGrok's founders received notice that their company was being sued by Adchemy, a former employer, in September 2010. The legal action was not unexpected as the former employer had told them they were unhappy with the founders leaving and starting a new company. AdGrok had gone through Y Combinator. Its relationship with the accelerator program and Sam Altman resulted in the lawsuit being dismissed. Adchemy was attempting to court funding from Microsoft - Altman had a contact at Microsoft tell Adchemy that it would be problematic if the company was embroiled in litigation while discussing the deal, so Adchemy dropped the suit

DSQL VIGNETTE: AURORA DSQL, AND A PERSONAL STORY

Amazon recently announced Aurora DSQL, a serverless SQL database designed for the cloud and optimized for transaction processing.

AMAZON ANNOUNCES NOVA, A NEW FAMILY OF MULTIMODAL AI MODELS

Amazon's new family of multimodal generative AI models includes four text-generating models, an image-generation model, and a video-generation model.

SAMSUNG SMART GLASSES DEBUT IN JANUARY, SOFTWARE REVEAL ‘THIS MONTH'

Samsung plans to preview its smart glasses, which will work similarly to Ray-Ban Meta smart glasses, in January alongside the Galaxy S25 - the actual release will be in the third quarter of 2025.

THE NEXT PLATFORM

Possible contenders for the next platform for job security purposes are PaleoOps, Erlang/Darklang, Fly.io, Serverless/AWS Lambda, and V8 isolates.

THE STARTUP TRYING TO TURN THE WEB INTO A DATABASE

Exa crawls the web and encodes the contents of web pages into embeddings that can be processed by large language models.

WALMART BOUGHT VIZIO

The acquisition will help bolster Walmart's advertising business.

ELEVEN LABS CONVERSATIONAL AI

Eleven Labs has launched a new conversational AI service that is more of an end-to-end product for building conversational agents. It uses a variety of LLMs on the backend while integrating seamlessly to a wide variety of specialized voices.

CLAUDE MODELS NOW FASTER AND SMARTER ON AWS TRAINIUM2

Claude models are being optimized for AWS' advanced Trainium2 AI chips. This will enable faster and more efficient performance. Claude 3.5 Haiku is now available on AWS Trainium2 and model distillation in Amazon Bedrock.

AMAZON LAUNCHES NOVA

Amazon Nova introduces a suite of multimodal models designed for tasks like document analysis, visual understanding, and creative content generation. With a focus on customization and efficiency, Nova models cater to diverse enterprise needs while excelling in tasks that combine text, image, and video inputs.

DIFFUSION MEETS FLOW MATCHING: TWO SIDES OF THE SAME COIN

This post talks through some of the literature and shows that, mathematically, both flow matching and diffusion models are the same. Interestingly, flow matching still seems to scale better practically.

RESTRUCTURING VECTOR QUANTIZATION WITH THE ROTATION TRICK

Vector Quantization relies on gradient estimation through what's known as the Straight Through Gradient estimator. However, its direction is sometimes incorrect. This paper suggests rotation as a way to fix the gradients and improve code book utilization.

LAYOUT GENERATION WITH DIFFUSION GANS

DogLayout is a hybrid model that combines GANs and diffusion processes to overcome challenges in layout generation.

drops PLUS a chance to win a MacBook M4. No brainer!

HUNYUAN VIDEO MODEL

This state-of-the-art open video model from Tencent is particularly compelling due to its realistic motion and dual training as an image generation model. This dual training means the aesthetics of the output are higher quality, akin to image generation models like Flux.

SCENE TEXT RECOGNITION

TextSSR is a framework that uses diffusion-based methods to generate accurate and realistic synthetic text images for scene text recognition.

VIDEO UNDERSTANDING WITH MULTIMODAL LANGUAGE MODELS

T2Vid is a new method to enhance video understanding in Multimodal Large Language Models (MLLMs). It generates video-like samples to diversify training instructions.

A SYSTEM OF AGENTS BRINGS SERVICE-AS-SOFTWARE TO LIFE

AI is transitioning software from a tool to autonomous agents capable of executing tasks traditionally performed by humans, with a projected $4.6 trillion market opportunity. Innovations such as LLMs and agents enable AI systems to process unstructured data, make decisions, and act autonomously across sectors like sales and healthcare. The future of AI involves Systems of Agents that collaborate and learn from each other, functioning like a highly skilled team to deliver services.

AI MUSIC IS MORE REALISTIC THAN EVER: MEET SUNO'S NEW MODEL

Suno has become the fifth most-used generative AI service with its realistic AI music model V4, despite facing a copyright lawsuit. The model improves user experience by focusing on human preferences, offering enhanced sound quality and advanced composition skills. Suno aims to advance AI-human music collaboration while addressing copyright concerns with the recording industry.

BLUESKY'S OPEN API MEANS ANYONE CAN SCRAPE YOUR DATA FOR AI TRAINING

Bluesky isn't using user content to train AI, but third parties like Hugging Face have accessed posts for machine learning research. This highlights the public nature of user posts and the challenges in controlling third-party use. Bluesky is exploring ways for users to convey consent, but enforcement outside its platform remains a challenge.