Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Five ways to reduce HTML file size to improve website performance and Core Web Vitals include checking file sizes, minifying and compressing files, breaking up long pages, limiting large inline resources, and avoiding excessive responsive markup.
Internal Facebook emails revealed during an antitrust trial show that Facebook executives, including Mark Zuckerberg, closely monitored Instagram's rapid growth in the early 2010s and considered acquiring it to neutralize the competitive threat.
Sponsored
Agentic AI is inseparable from the APIs that enable AI agents to work their magic. How do you build software that's ready for a whole new generation of AI-powered and AI-enable users?
A federal judge ruled yesterday that Google created a monopoly that allowed it to control parts of the online advertising industry. It is the second time in eight months that Google has been labeled an illegal monopolist. The company could be forced to sell some of its advertising products. Google plans to appeal the ruling.
SpaceX and two partners, Anduril and Palantir, are frontrunners in a race to win a crucial part of President Trump's Golden Dome missile defense shield. They have proposed a constellation of 400 to more than 1,000 missile defense satellites and a subscription model for missile defense. Trump has cited a missile attack as the most catastrophic threat facing the US. The decision process for the Golden Dome is in its early stages and its ultimate structure and who is selected to work on it could change drastically in the coming months.
Orforglipron is a daily pill by Eli Lilly that may be as effective in lowering blood sugar and aiding weight loss in people with Type 2 diabetes as Ozempic. It is part of the GLP-1 class of drugs that have become famous for their weight-loss effects. While the pill is still expected to be expensive, it has the potential to become far more widely used than current GLP-1s, which must be kept refrigerated and must be injected. Eli Lilly will seek approval from the FDA later this year to market the drug for obesity and early in 2026 for diabetes.
Surgeons at the National Taiwan University Hospital have performed a heart transplant during which the donor's heart continued beating between the organ removal and transplantation stages. Donor hearts are usually removed and preserved in cold storage, which can damage muscle tissue, reducing function and health once transplanted. The operation was possible due to a specially designed organ maintenance system that kept the donor heart pumping with oxygenated blood throughout the process. Followup appointments with the patient showed that they maintain a low level of cardiac enzyme, something that usually spikes during typical transplant conditions.
Agents are AI systems that can perform workflows on users' behalf with a high degree of independence. This guide, designed for product and engineering teams, explores how to build agents with frameworks for identifying promise use cases, clear patterns for designing agent logic and orchestration, and best practices to ensure agents run safely, predictably, and effectively. It provides the foundational knowledge needed to confidently start building agents. Building reliable agents means starting with strong foundations - start small, validate with real users, and grow capabilities over time.
OpenAI's Flex processing is a new API option that provides lower AI model usage prices in exchange for slower response times and occasional resource unavailability. It is available in beta for the company's recently released o3 and o4-mini reasoning models. Aimed at lower-priority and non-production tasks, the feature reduces API costs by exactly half. In the email announcing the launch of Flex processing, OpenAI also indicated that developers in tiers 1 to 3 of its usage tiers hierarchy will have to complete ID verification to access o3.
The AI industry is now shifting its focus from solving problems to defining problems. Evaluation is becoming more important than training - people are now looking at how to measure real progress. A shift in mindset and skill set is required to thrive in this new era. Players in the game can now build billion- or trillion-dollar companies by making useful products out of intelligence.
Lawyers for the FTC pressed Sheryl Sandberg, who was second in command at Meta from 2008 to 2022, on the witness stand on Wednesday during the company's landmark antitrust trial. She was asked to explain an email chain in which executives disallowed Google+ from advertising with the company. She said that it was a scary time for the company and that Google's new tool was being promoted on its main search page. The FTC's volume of internal communication evidence is compelling as it shows that Mark Zuckerberg had an interest in buying a competitor.
Apple is considering expanding the capacity of its facilities in Brazil in partnership with Foxconn to avoid US import tariffs.
Gemini 2.5 Flash is now rolling out widely in preview at prices much lower than competing products.
Developers are taking supervisory roles, guiding agents doing work on their behalf, but they remain responsible for what gets deployed.
Anysphere, the startup that sells Cursor, was reportedly in talks last month to raise funding at a valuation close to $10 billion.
Some of Meta's apps now only offer a 'Write with AI' option that leads to a Meta AI window rather than offering Apple Intelligence writing tools.
OpenAI's memory upgrade marks the end of manual prompts and the start of truly personalized, proactive experiences that anticipate users' needs.
Sponsored
You've built something powerful with AI. Now you're asking: how do we price it? Many startups are getting the answer wrong right now—and putting their business's viability at risk.
Mistral, a French AI startup, has launched a new product that allows users to very quickly build and deploy custom classifiers for a whole variety of tasks (e.g., spam, moderation, and more).
Goodfire is a mechanistic interpretability company with strong expertise in SAEs, among other things. It is working closely with closed and open model providers to steer, control, and understand model motivations and behavior.
OpenAI's latest visual models can reason with images through tool-augmented transformations, enabling a new level of multimodal understanding and step-by-step visual problem-solving.
Novel efficient long-context fine-grained ID preservation framework for line art colorization, achieving high precision, efficiency, and flexible usability for comic colorization. It transforms black-and-white line art into vibrant illustrations by effectively integrating extensive contextual references.
3D CoCa is a unified framework that combines vision-language contrastive learning and captioning for 3D scenes.
JudgeLRM is a family of LLMs trained with reinforcement learning for judgment tasks. Unlike SFT, it excels in reasoning-heavy evaluations, outperforming models like GPT-4 and DeepSeek-R1.
The DeepSpeed team has worked to bring compilation to their distributed training efforts. This compilation speeds up various bottlenecked operations by many times. It makes use of a patched version of torch compile.
SIFT-50M (Speech Instruction Fine-Tuning) is a 50-million-example dataset designed for instruction fine-tuning and pre-training of speech-text large language models (LLMs). It is built from publicly available speech corpora containing a total of 14K hours of speech and leverages LLMs and off-the-shelf expert models. The dataset spans five languages, covering diverse aspects of speech understanding and controllable speech generation instructions. SIFT-50M augments existing speech datasets with instruction-based question-answer (QA) pairs for speech understanding and includes approximately 5 million examples for controllable speech generation.
Meta has released an image Encoder, a VLM, a 3D object localization model based on JEPA, and weights for a BLT model that operates directly on bytes without tokenization.
Cohere became the first model creator to directly host and serve its enterprise-focused AI models on Hugging Face.
YouTube's Dream Track is now available in the U.S. on YouTube Shorts and the YouTube Create app, enabling AI-generated instrumental soundtracks for content creators. These tracks can be remixed globally to create unique Shorts, fostering a collaborative ecosystem. The feature integrates directly with YouTube's creation tools and adheres to community guidelines.
OpenAI has introduced Flex processing, a cost-saving API option that trades slower response times and intermittent availability for lower prices, ideal for non-production tasks.
C2S-Scale is a new family of LLMs that interprets single-cell data and translates biological signals into natural language for applications in personalized medicine and drug discovery.
Anthropic has launched new Claude features: Research for autonomous multi-step search with citations and Google Workspace integration for context-aware assistance.
A new memo from the DOJ signals a shift away from prosecuting crypto intermediaries, narrowing enforcement to clear cases of fraud or national security risk and potentially leading to the closure or dismissal of ongoing investigations launched under the Biden administration. While industry advocates see this as a relief, critics warn it may weaken oversight and embolden misconduct, with civil agencies like the SEC and CFTC also adjusting their approach in response.
VanEck has secured SEC approval to launch the Onchain Economy ETF (ticker NODE), with a planned debut on May 14. The actively managed fund will hold 30-60 assets tied to the blockchain economy, including exchanges, miners, data centers, and firms with crypto assets. This offering joins a surge of digital asset ETF filings timed around the new pro-crypto administration and SEC overhaul.
EigenLayer has activated slashing on mainnet, making the protocol feature-complete and enabling economic enforcement for Autonomous Verifiable Services (AVSs), where operators and stakers are rewarded for meeting commitments or penalized if they fall short. This milestone brings verifiable trust to decentralized services, with major partners like LayerZero and Infura integrating slashing to enforce service-level guarantees and increase reliability across blockchain infrastructure.
Eliza Labs has launched auto.fun, a no-code platform that lets users deploy AI agents across Web3 services, supported by a token model that combines bonding curves and fair launch principles. The platform aims to democratize AI and Web3 access with transparent, open-source agents and sustainable tokenomics that align incentives between developers and users.
Converge is an institutional-grade EVM chain with support for permissioned RWAs and permissionless DeFi from Ethena and Securitize. Its recently unveiled technical roadmap details a high-performance EVM chain with 100ms block times and novel security features through the Converge Validator Network (CVN) secured by staked ENA tokens. Partnering with Arbitrum and Celestia, Converge aims to offer frictionless usability and institutional-grade infrastructure, including support for stablecoins like USDe and USDtb as gas tokens.
DEX trading hit an all-time high of $380 billion in January 2025 before plunging 35% by February. The top ten DEXs accounted for 80% of volume, with Uniswap and PancakeSwap driving 40% and five of the top ten being Solana-native. Base's DEX share is 7%, Ethereum's is 26%, and Solana's is 21%. Centralized exchanges captured 80% of spot volume at the January peak and Hyperliquid held 60% of the perpetuals market share.
Timeboost is a new optional transaction ordering system for Arbitrum chains that uses sealed-bid auctions to allocate short-term “express lane” privileges, allowing high-value transactions to bypass delays while preserving fast block times and privacy for all users. It helps chain owners capture some MEV revenue and reduces spam while protecting users from harmful MEV like front-running
Verifiable AI agents tackle both AI's trust problem and crypto's app problem by combining blockchain verifiability with domain-specific intelligence. OpenLedger trains specialized language models on high-quality, sector-focused data and uses Proof of Attribution to trace how individual data points influence outputs and detect bias. By layering this model-level verification with onchain execution checks from EigenLayer, agents become fully auditable from reasoning to execution.
Coinbase has significantly improved its support for Solana, increasing transaction throughput fivefold and providing a smoother experience for Solana users on the platform.
A battle is unfolding over U.S. stablecoin regulation, with Bank of America pushing to limit issuance to banks, while Circle promotes itself as a compliant U.S. alternative to offshore Tether.
Safe is undergoing a major internal restructuring and has parted ways with 14 team members as part of efforts to streamline operations and improve agility as the organization evolves.
A new render offers a closer look at the iPhone 17 Pro's unibody rear, which showcases a sleek ‘camera bar' design with a uniform color across the back. Unlike previous models, there's no two-tone finish. The camera bump may include a larger Xenon flash, a higher-resolution LiDAR scanner, or an additional sensor. This redesign is part of Apple's iPhone 17 lineup, which is expected to debut this fall.
OpenAI has added an image library to ChatGPT that allows users to easily access AI-generated images via a sidebar. Available on both web and mobile platforms, it supports all users. Though simple, it could signal OpenAI's push toward a social platform for image sharing, boosting visibility, and gathering real-time user data for future developments.
The 2026 Winter Olympic torches, called Essential, feature a minimalist, open-top design that emphasizes the flame over the torch itself. Designed by Studio Carlo Ratti Associati and made with recycled materials and renewable fuel, the torches are lightweight, reusable, and produce a warm, dynamic flame. The design reflects Italy's modern design legacy and sets the tone for the Games' overall aesthetic.
Adobe plans to integrate agentic AI across its tools to boost creativity and productivity. These AI agents will assist with tasks, offer suggestions, and streamline workflows, empowering users at all skill levels. From Acrobat to Photoshop and Premiere Pro, Adobe envisions agents that enhance human imagination, reduce tedious work, and accelerate high-impact content creation.