Search the archive

YELP DATA REVEALS THE TOP 50 FASTEST GROWING BRANDS

Yelp ranked businesses based on internal data, including net-new location openings, consumer interest, and searches.

CODE SEARCH IS HARD

Code search is a hard problem to solve compared to normal search. Search tools often optimize for natural language, but these techniques make code search worse. Most big companies have custom code search tools, which is a sign of how difficult the problem is. Postgres has functions that can make code search easier, like its full text search options.

HOW I WRITE TESTS IN GO

Golang comes with its own testing library, which is not common among popular programming languages. Unit tests in Go should be written to run in parallel by default so that concurrency issues are surfaced early on. The testing package has a comprehensive set of flags and can provide code coverage reports, race condition detection, static analysis, shuffled test execution, and more.

SHELL HISTORY IS YOUR BEST PRODUCTIVITY TOOL

This blog demonstrates how customizing ZSH shell history configurations can be a boon for your productivity. These tips make it easier to find and reuse past commands quickly. You can efficiently access a personalized command library by tuning history storage and implementing advanced search techniques such as ‘fzf’ and ‘ag’.

DON'T REQUIRE PEOPLE TO CHANGE 'SOURCE CODE' TO CONFIGURE YOUR PROGRAMS

Programs should avoid requiring users to modify source code files (like Makefiles or header files) for configuration purposes. This is because it can often lead to merge conflicts during updates later on, which can be difficult to debug. Configurations should be handled through other methods instead.

CODEGEMMA

CodeGemma, a collaborative effort between Google and Hugging Face, is a family of open-access, code-specialized LLMs. The models are available in 2B and 7B configurations. They have been fine-tuned with a lot of data to bring better performance in logical reasoning and programming suggestions.

AIDER

Aider is a command-line tool that lets you directly edit code in your files while pair-programming with GPT. It will git commit changes with AI-generated commit messages.

DECO

Deco is a real-time TypeScript web editor. You can code directly on the web or on your machine and get instant feedback.

GEMINI 1.5 AND GOOGLE’S NATURE

The recent Google Cloud Next keynote showcased Google’s vast, scalable AI infrastructure, which is difficult to replicate. Gemini models can now be integrated with Google Search and enterprise data sources, reducing hallucinations and creating better answers. Gemini 1.5 Pro can process up to 1 million tokens of information, which is a breakthrough in long context understanding. Google is integrating Gemini into many of its enterprise products, such as Google Workspace and code development tools, which is another competitive advantage for them.

NOTES ON GIT'S ERROR MESSAGES

Git’s error messages can be confusing to understand, but there are some great strategies to help decode them. Git status clarifies the current branch state, and git log helps examine commits and identify divergence points. You can also customize your shell prompt to display branch status, which can be helpful.

WHY CAN'T MY MOM EMAIL ME?

This author was receiving encrypted emails with blank bodies from his mother. The issue was Proton's use of the Web Key Directory (WKD), which automatically encrypts emails if a public key is found for the recipient's address, even if the user didn't explicitly enroll in the service. The author’s email address was enrolled, so he received blank emails.

OH MY GIT!

An open-source game to learn and understand Git better.

GPT-4 TURBO WITH VISION NOW AVAILABLE VIA API

Some interesting use cases for GPT-4 with Vision include nutrition insights from pictures, UI replication from screenshots, web scraping, and data extraction.

"BATBADBUT" VULNERABILITY DISCOVERED IN RUST STANDARD LIBRARY ON WINDOWS

A vulnerability named "BatBadBut" was discovered in the Rust standard library on Windows that allows arbitrary command execution due to improper argument escaping when invoking batch files.

IMPROVING OUR JEST EXECUTION TIME BY 300%

A look into how one developer improved slow Jest tests by fixing memory leaks, disabling redundant type checking, and optimizing some functions.

WITH VIDS, GOOGLE THINKS IT HAS THE NEXT BIG PRODUCTIVITY TOOL FOR WORK

Google's new Vids app can create collaborative shareable videos for things people do at work. The app makes making videos as easy as making slides - no video production is required. Users line up assets inside the app and edit it all into a finished video. They can choose to do all of this themselves or ask Google's Gemini AI to build storyboards, write scripts, read the scripts aloud using text-to-speech, and create images to use in the video. People with access to the video can add comments, leave notes, and make edits. Vids will launch in public beta this summer.

INTEL’S GAUDI 3 GOES AFTER NVIDIA

Intel gave the first architectural details of its Gaudi 3 third-generation AI processor at Vision 2024 this week in Phoenix, Arizona. Gaudi 3 is made up of two identical silicon dies, each with a central region of 48 megabytes of cache memory, joined by a high-bandwidth connection, surrounded by four engines for matrix multiplication and 32 programmable units called tensor processor cores. It produces double the AI compute of Gaudi 2 using 8-bit floating-point infrastructure. It also provides a fourfold boost for computations using the BFloat 16 number format.

META CONFIRMS THAT ITS LLAMA 3 OPEN SOURCE LLM IS COMING IN THE NEXT MONTH

Meta plans to release an initial version of its next-generation Llama 3 large language model within the next month. The company will release a number of different models with different capabilities and versatilities during the course of the year. Llama 3 will be able to answer a wider range of questions compared to its predecessor, including questions regarding more controversial topics. Meta has not released any details about the model's size, but it is expected to have about 140 billion parameters - the biggest Llama 2 model has 70 billion.

GABE NEWELL'S NEURAL INTERFACE COMPANY UPDATED ITS WEBSITE

Starfish, a neural interface company co-founded by Gabe Newell, recently updated its website, indicating that its team is working hard on its brain-computer solutions. The company is developing minimally invasive one-dimensional implants for neuromodulation and neural recording. Newell's interest in the field goes back to at least 2021, when he brought up the idea of using brain-computer interfaces for gaming. Valve has previously worked to create open source software to allow developers to understand signals coming from people's brains.

GOOGLE LAUNCHES CODE ASSIST, ITS LATEST CHALLENGER TO GITHUB’S COPILOT

Google's Gemini Code Assist is an enterprise-focused AI code completion and assistance tool. It was previously offered under the now-defunct Duet AI branding which became generally available in late 2023. Code Assist is both a rebrand and a major update. It uses Gemini 1.5 Pro, which has a million-token context window. Code Assist will be available through plug-ins for popular editors like VS Code.

CHRONON

Chronon is a platform that allows organizations to power AI/ML projects without needing to worry about orchestration by abstracting away the complexity of data computation and serving for AI/ML applications. It can perform batch and streaming computation, scalable backfills, low-latency serving, and more. Chronon can utilize all of the data within an organization.

BEEPER WAS JUST ACQUIRED BY AUTOMATTIC, WHICH HAS BIG PLANS FOR THE FUTURE OF MESSAGING

Beeper, the company that attempted to launch an app that let Android users use iMessage a few months ago, has been acquired by Automattic, the giant that owns WordPress. As part of the deal, Beeper is opening up its messaging app, which attempts to corral all messaging services into one inbox, to everyone across platforms, and shutting down its waitlist for good. Automattic's CEO says that messaging will be the next big pillar of the company. Its team intends to replace a lot of messaging methods with an open-source system.

APPLE PLANNING TO OPEN LARGE OFFICE IN MIAMI

Apple is leasing an office building in the Miami area. The 45,000 square feet of space, located at The Plaza Coral Gables, is still under construction. Apple is also planning to open a new retail store at the Miami Worldcenter in the heart of the city. Multiple tech companies have expanded to South Florida, including Amazon and Microsoft.

GOOGLE’S NEW ARM-BASED CPU WILL CHALLENGE MICROSOFT AND AMAZON IN THE AI RACE

Google's new Arm-based CPU, Axion, will be used to support Google's AI workloads before it rolls out to Google Cloud business customers later this year.

BUILDING RELIABLE SYSTEMS OUT OF UNRELIABLE AGENTS

It's possible to build reliable systems out of AI models by writing simple prompts, building an eval system to do prompt engineering and improve performance in a principled way, deploying AI systems with good observability, investing in Retrieval Augmented Generation (RAG), and fine-tuning models using data gathered from the process.

OPENAI MAKES GPT-4 TURBO WITH VISION AVAILABLE TO DEVELOPERS TO UNLOCK NEW AI APPS

GPT-4 Turbo with Vision allows developers to call on a single model for both text and image processing.

LIFEHACKS

This post contains a list of 130 tips and mindsets for living a better life.

THE THREAT TO OPEN SOURCE COMES FROM WITHIN

The fragile balance of open source will be disrupted if the community comes to believe that maintaining that balance is more trouble than it's worth.

GOOGLE GEMMA EXPANDED MODELS

Google has trained a code Gemma and Recurrent Gemma models. They get competitive performance and include FIM capabilities. The recurrent model is much faster and more memory efficient.

META CONFIRMS THAT LLAMA 3 IS COMING NEXT MONTH

Meta has confirmed plans to release Llama 3, the next generation of its large language model for generative AI assistants, within the next month.

FOUR TAKEAWAYS ON THE RACE TO AMASS DATA FOR A.I.

SWAP ANYTHING

SwapAnything is a new algorithm that can replace things in a picture with other things you choose, without changing the rest of the picture. It's better than other tools because it can swap any object, not just the main subject, and it's really good at making the new object fit perfectly into the original picture. It uses inversion, concept vectors, and a pretrained diffusion model.

IMPROVING STABLE DIFFUSION WITH UNIFIED FEEDBACK

ENHANCING OBJECT DETECTION IN SINGLE-DOMAIN

Addressing the challenge of single-domain generalization (S-DG) in object detection, the new OA-DG method introduces OA-Mix for data augmentation and OA-Loss for training.

Visit the R2R GitHub ( [11]⭐ 1.1K stars) to get started today.

VISUAL AUTOREGRESSIVE MODELS

Code for the recent “next-resolution prediction” work that frames image generation as progressive prediction of increasing resolution. The repository includes inference scripts and a demo notebook. The training code will be released soon.

CHEMISTRY BENCH FOR LANGUAGE MODEL

A Big-Bench compatible benchmark for evaluating large language model capabilities on chemistry questions. This benchmark can help measure the scientific ability of various language models.

THE SQUEEZEATTENTION FOR LLMS

A new method called SqueezeAttention has been developed to optimize the Key-Value cache of large language models, significantly reducing memory usage by 30% to 70% and doubling the throughput.

INTEL DETAILS GAUDI 3 AT VISION 2024

Intel has announced its new Gaudi 3 AI processors, claiming up to 1.7X the training performance, 50% better inference, and 40% better efficiency than Nvidia's H100 processors at a lower cost.

‘IT’S VERY EASY TO STEAL SOMEONE’S VOICE’: HOW AI IS AFFECTING VIDEO GAME ACTORS

Cissy Jones, a renowned voice artist, co-founded Morpheme to offer ethical AI voice modeling after her voice was used without consent. Morpheme ensures actors consent and are compensated for AI-generated vocal content. The broader industry's rush to AI adoption may sideline human talent and ignore actors' rights. As the video game industry integrates AI, Sag-Aftra union negotiations aim to protect performers' consent and fair compensation.

CAN DEMIS HASSABIS SAVE GOOGLE?

DeepMind founder Demis Hassabis now leads Google's unified AI research arm, aiming to maintain the tech giant's edge in the AI landscape with breakthroughs like AlphaGo and AlphaFold. Despite the success, challenges in integrating AI into tangible products and competition from entities such as OpenAI's ChatGPT persist. Hassabis, recognized for his significant contributions to AI, must now navigate Google's product strategy to leverage DeepMind's research advancements.

APPLE’S NEW AI MODEL COULD HELP SIRI SEE HOW IOS APPS WORK

Apple has developed a new multimodal LLM called Ferret-UI that has the potential to understand the user interfaces of mobile displays, enabling Siri to comprehend and interact with on-screen elements effectively.

AEROSPACE AI HACKATHON PROJECTS

200 AI and aerospace engineers gathered to prototype innovative solutions for the Aviation and Space industries and built an impressive array of tools ranging from AI air traffic controllers to AI flight planners to Apple Vision Pro flight simulators.

LONGCONTEXT ALPACA TRAINING (COLAB NOTEBOOK)

Train over 200k context windows on an H100 with a new gradient accumulation offloading.

INACTIVE BITCOIN SUPPLY DECREASES AS HOLDERS SELL