Search the archive

ingredients. Special offer for TLDR ↗️

OPERA BROWSER DEV BRANCH ROLLS OUT SUPPORT FOR RUNNING LLMS LOCALLY

Opera's Opera One Developer browser can now run 150 different large language models from 50 different large language model families without requiring an internet connection except to download the models.

GROQ CEO: ‘WE NO LONGER SELL HARDWARE’

Groq is now an AI cloud services provider.

"IF THIS ONE GUY GOT HIT BY A BUS, THE WORLD'S SOFTWARE WOULD FALL APART."

The entirety of modern computing infrastructure is built on top of thousands of projects that are built, maintained, and run entirely by one person or by extremely small teams that are more often than not unpaid volunteers.

SCOOP: Y COMBINATOR INVITES ALUMS TO INVEST IN NEW FUNDS

Alumni must invest a minimum of $250,000 total, which will be split among three funds that Y Combinator is currently raising.

INVESTORS IN TALKS TO HELP ELON MUSK’S XAI RAISE $3 BILLION, ADDING TO INDUSTRY ARMS RACE

If the funding round is successful, xAI would be valued at $18 billion.

AMD TO OPEN SOURCE MES FIRMWARE

AMD has agreed to open source the firmware for its Radeon GPUs. This means the community can quickly improve AI frameworks and potentially AMD adoption.

XAI LOOKING TO RAISE $3B

Elon Musk's xAI is in talks with investors to raise $3B in a funding round that would value the artificial intelligence startup at $18B.

INTRODUCING COMMAND R+: A SCALABLE LLM BUILT FOR BUSINESS

Cohere has introduced Command R+, a powerful, scalable LLM designed for enterprise use cases, featuring advanced retrieval augmented generation with citation, multilingual coverage in 10 key languages, and tool use capabilities.

QWEN1.5-32B

The Qwen team has trained and released a 32B parameter model that achieves strong performance. It can be fit in more modest memory systems.

PROTECTING IMAGES FROM UNAUTHORIZED SEGMENTATION

"Anything Unsegmentable" is a new approach designed to shield digital images from being segmented by powerful AI models, addressing potential copyright and privacy issues.

A BENCHMARK FOR DETECTING LLM ERRORS

Researchers have launched ReaLMistake, a benchmark aimed at systematically detecting errors in large language model responses.

more >>

SWE-AGENT

SWE-agent turns LLMs (e.g. GPT-4) into software engineering agents that can fix bugs and issues in real GitHub repositories. SWE-agent sets the state-of-the-art performance on the full SWE-bench benchmark, resolving 12.29% of issues.

SCHEDULE FREE OPTIMIZATION

Researchers from Meta have been teasing a new optimizer on X. They have now released the code along with various integrations. The optimizer has no LR schedule, which means you don’t need to know the full number of training steps beforehand. It has been shown empirically to work on a wide variety of problems including language models.

REPRESENTATION FINE-TUNING

ReFT is a new parameter-efficient method for fine-tuning language models. It is substantially cheaper than even PeFT while achieving strong performance.

FOUR TAKEAWAYS ON THE RACE TO AMASS DATA FOR AI

The development of artificial intelligence heavily relies on vast amounts of data, which is being rapidly consumed by tech companies faster than it is being produced, leading to predictions that high-quality digital data may be exhausted by 2026. In response, companies like OpenAI, Google, and Meta are exploring new methods to obtain more data, including using YouTube video transcripts, revising privacy policies, considering the purchase of major publishers, and investigating the use of "synthetic" data generated by AI models, despite the risk of compounding errors.

SPEED TESTS FOR LLAMA 2, STABLE DIFFUSION

MLPerf has updated its inferencing benchmarks to include large language models like Llama 2 70B and Stable Diffusion XL, reflecting the industry's shift to massive generative AI. In the latest tests, Nvidia's systems, particularly those equipped with the H200 processor, outperformed competitors from Intel and Qualcomm in both speed and efficiency. The benchmarks highlight the importance of high-bandwidth memory in handling large AI models, with Intel's Gaudi 2 and Qualcomm's Cloud AI 100 Ultra also showcasing significant performances in the generative AI space.

RABBIT PARTNERS WITH ELEVENLABS TO POWER VOICE COMMANDS ON ITS DEVICE

Rabbit partners with ElevenLabs to incorporate voice command technology into its upcoming r1 devices, enhancing human-device interaction with low latency models for a more natural experience. The first batch of r1 devices, which feature functionalities like chatbot interaction and bi-directional translation, is set to ship by March 31. ElevenLabs has recently raised $80 million, despite facing challenges with misuse of its voice cloning technology.

DALL-E NOW LETS YOU EDIT IMAGES IN CHATGPT

OpenAI's DALL-E has been integrated with ChatGPT - users can now edit DALL-E images in ChatGPT across web, iOS, and Android.

THE TOP 100 AI FOR WORK

A curated list of the top 100 AI tools to boost your work productivity.

LLOCALSEARCH

A completely locally running search engine using LLM Agents.

COINBASE CLEARED IN LAWSUIT OVER CRYPTO TRANSACTIONS

The U.S. Court of Appeals for the Second Circuit confirmed that Coinbase did not violate the Securities Exchange Act by engaging in secondary sales of cryptocurrencies. This case was related to a lawsuit from Coinbase users accusing the platform of selling unregistered securities and violating related laws – not the recent SEC case alleging they are an unlicensed securities exchange and brokerage. The court ruled that without a contract that can be rescinded, there is no Exchange Act violation, meaning trading cryptocurrencies on Coinbase does not constitute securities trading.

FTX ESTATE SELLS $1.9 BILLION WORTH OF LOCKED UP SOL

The FTX estate has sold approximately $1.9 billion worth of locked-up Solana (SOL) tokens at a discounted price of $64 each to help clear the collapsed crypto exchange's debt. Galaxy Trading and other investors have shown significant interest in purchasing FTX's discounted SOL holdings.

DYDX LAUNCHES LIQUID STAKING

Stride has partnered with dYdX to launch a one-click liquid staking system for dYdX tokens. The system automates compounding USDC rewards into dYdX and distributes staked tokens across validators to strengthen network security.

JAPAN’S SONY BANK TESTS YEN-BACKED STABLECOIN

Sony Bank, a division of the Sony Group Corporation, is testing the issuance of a yen-backed stablecoin on the Polygon blockchain. The stablecoin is being considered as a potential payment method for businesses utilizing Sony Group's intellectual property, such as games and sports.

QUIC VULNERABILITY SOLANA

This thread discusses the QUIC protocol's challenges in blockchain networking, including susceptibility to packet overload but not inherent DDoS vulnerability. The author suggests improving QUIC server resilience and packet prioritization to manage the high packet traffic on Solana's network rather than changing protocols.

DEEP DIVE INTO BITTENSOR

Bittensor is an open-source protocol that aims to decentralize the development of artificial intelligence using economic incentives similar to Bitcoin's to build a market for synthetic intelligence. Through its unique network design, Bittensor allows contributors to earn a native token (TAO) by participating in a variety of subnets catering to different AI applications, from text generation to financial predictions, democratizing AI development beyond the control of centralized entities.

WHAT CRYPTO VCS ARE THINKING ABOUT

Crypto VCs are currently debating the merits of investing in "useful" versus meme-driven projects, with some leaning into the hype while others focus on more substantive areas like stablecoins, real-world assets, and social applications. There is also debate around whether mainstream adoption will be driven by Web 2.5 apps that abstract away crypto complexities or Web3 apps with tokens and protocols front and center.

DAO GOVERNANCE CATEGORIES

MakerDAO founder Rune outlines the two options for DAO decentralized governance: minimization of simple systems when oversight is not required and data-informed governance in complex systems - popularity-driven governance is dangerous, as it rewards the loudest solutions instead of those that may be the best.

ROBINHOOD CRYPTO HOLDINGS UPDATE

There has been a 14% increase in Robinhood's BTC holdings from October 2023 to now, a 9% decrease in ETH holdings, and a 100% increase in LINK holdings.

A SIMPLE EXPLANATION OF WHY YOUR SOLANA TRANSACTIONS ARE FAILING

Solana nodes are overwhelmed and can't process every tx - bots are spamming txs to the network in order to maximize the chance that a tx gets processed, hurting real users who can't spam at the same rate.

TENSOR NFT MARKETPLACE REVEALS SOLANA TOKEN LAUNCH, COINBASE TO LIST

This article discusses Tensor’s token launch, including details about its points program and NFT collection Tensorians.

IPHONE 16: LEAKED PICS SUGGEST IT'LL GO WILD WITH BUTTONS

According to photos of alleged dummy iPhone 16 units appearing on X, Apple seems to be adding a programmable action button to every iPhone 16 model. There's a new button on the right side called "Capture," which may be used for zooming and snapping photos more intuitively than currently possible on iPhones. 9to5Mac reports that both iPhone 16 Pro models will be bigger than last year, with the iPhone 16 Pro measuring 6.3 inches and the iPhone 16 Pro Max measuring 6.9 inches.

NASA PICKS THREE TEAMS TO DESIGN THE NEXT GENERATION OF MOON BUGGY

NASA has invited three space companies, Intuitive Machines, Lunar Outpost, and Venturi Astrolab, to build rugged vehicles astronauts can drive on the moon's surface. Next year, the agency will select a winner based on specifications such as range, battery technology, and life span. Using the vehicles, astronauts can carry scientific equipment, collect samples from the surface, and travel farther.

COCA-COLA EMBRACES IMITATION WITH THE ‘EVERY COCA-COLA IS WELCOME’ CAMPAIGN

Many brands move quickly to protect their copyrighted materials and likenesses. Coca-Cola has recently launched a campaign that does the opposite, celebrating the many interpretations of its logo created worldwide for generations. The ‘Every Coca-Cola is Welcome’ campaign embraces and showcases grassroots interpretations of the company's wordmark created by bodegas, shopkeepers, and local artists instead of slapping these creators with lawsuits.

THE VALUE OF DESIGN TOKENS IN MODERN WEB DEVELOPMENT

Design tokens promote collaboration and save team time by providing a common language for design decisions. By providing a scalable system of colors, fonts, spacing, and sizing, they aid product teams in maintaining consistency and allow them to update their styles more seamlessly. This thoughtful article explores a cross-functional workflow using design tools like Figma and web components.

RUNNING A DESIGN MEETING THAT DOESN’T WASTE TIME

Most meetings are time-consuming sessions with no clear direction or practical results. However, sometimes they drive collaboration and product innovation, bringing the team together to fine-tune design and product visions and drive projects forward. To design successful meetings, it's important to include clear objectives, a pre-meeting brief, time management, cooperation rules, and a clear action plan.

PLANNING RESEARCH WITH GENERATIVE AI

With context, examples, and guidance, AI tools like ChatGPT and Microsoft Copilot can generate research questions, tasks, interview prompts, and study materials much faster than creating them from scratch. Research leads and ResearchOps personnel can assist junior researchers and PWDRs by providing examples and advice for AI agents. Experienced researchers can use AI to expedite research planning and find inspiration.

EDITABLE HIGH-QUALITY ILLUSTRATIONS (FIGMA PLUGIN)

Storyset is a collection of thousands of editable illustrated concepts.

CREATE A WAITLIST IN THREE CLICKS

Waitforit lets you easily and quickly get customizable waitlist forms.

BUILD HEADLESS SHOPIFY STORES IN WEBFLOW

Shopyflow allows you to build Shopify stores using native Webflow elements and workflows.

THE DETAIL, THE QUALITY

Designers must think beyond the structures they impose on themselves. They must add the final touch to the product, making it special, unique, and, above all, consider the customer's journey. The experience should be savored and remembered.

IMMERSIVE EXPERIENCE DESIGN: EXPERT INSIGHTS AND TECHNIQUES

In immersive experience design, various disciplines, including augmented reality (AR), virtual reality (VR), interior architecture, sound design, visual storytelling, and more, are combined to create immersive experiences. Tech giants' adoption of immersive experiences makes designing for new realities increasingly plausible. This Q&A with two experts shares examples and ideas of how to bring immersive experiences to life.

SENSORY DESIGN IN ARCHITECTURE: WHAT ARE ITS CORE PRINCIPLES

Sensory design considers all the things a person feels inside a building, like sounds, textures, aromas, and tastes. The right combination of these elements creates environments that foster emotional connection, cognitive engagement, and well-being. In this article, architects demonstrate how they create spaces that resonate with occupants through sight, sound, touch, scent, and taste.

DESIGN IS NOT ART, BUT DO NOT FORGET THE ARTIST WITHIN YOU

To excel as a designer, amplify your artistic essence and incorporate it into your proposals, using emotional and visual design to establish connections and imbue your work with depth and meaning.

A COMPLETE GUIDE TO CSS LOGICAL PROPERTIES, WITH CHEAT SHEET

Logical properties are based on text direction, offering a way to adapt more effectively to changes in text direction, which is particularly useful when a site supports multiple languages.