Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Nvidia has announced a new GPU called the Rubin CPX designed for context windows larger than 1 million tokens. The GPU, meant to be used as part of a broader 'disaggregated inference' infrastructure approach, is optimized for the processing of large sequences of context. It performs better on long-context tasks like video generation and software development. The Rubin CPX will be available at the end of 2026.
OpenAI's Responses API replaces the previous /chat/completions API for inference. The new API has a lot more features, but the main difference is that it is stateful. Users no longer have to pass the entire conversation history with each request, they just have to pass an ID representing the state of the conversation and the provider keeps models up to date. This allows OpenAI to keep its reasoning traces secret.
Every serious AI company will eventually train its own models. The barrier to doing so is collapsing. Distillation, fine-tuning, and post-containing get easier every month. Soon, the only way to stay relevant will be to own your own models.
Production monitoring reveals real issues that pre-deployment evals will inevitably miss, especially as AI products become more unpredictable and personalized. Evals are collections of already-known failure cases, but agents can and often do fail in ways that produce no error codes.
Sponsored
Get your data ready for the future of analytics and AI at CData Foundations 2025 [12]! One free registration gets you two virtual summit days - Analytics on September 17 and AI on September 24. Join for expert insights, real-world success stories, and practical strategies to modernize your data architecture. SAVE YOUR SPOT (FREE)
ByteDance's REER (REverse-Engineered Reasoning) is a paradigm that derives step-by-step reasoning from known good answers rather than building it forward via RL or imitation.
The Model Context Protocol Registry provides a standardized way to distribute and discover MCP servers. A community-driven project, it allows organizations to create private enterprise registries while maintaining compatibility through shared API schemas and moderation guidelines.
Gemini Web has rolled out its Tools redesign. The prompt bar is now centered with a new 'Tools' menu that lets users access Deep Research, Veo, image creation, Canvas, Guided Learning, and Deep Think. Gemini now groups suggested prompts together for a cleaner look. Screenshots of the new interface are available in the article.
AI companies face varied gross margins across sectors. Chips maintain around 70% margins, while cloud services see margins pressured by AI investments, estimated between 50-55%. Application-level margins range widely, with AI "Supernovas" starting at 25%, potentially negative, while others hit 60%, highlighting pricing strategies and diversified revenue models to improve margins over time.
Sponsored
Warp Code adds a built-in editor and code review to proven CLI and MCP tools, bringing everything together in one powerful app. Top-rated
OpenAI announced a $50 million fund to support nonprofits working on AI literacy, community-driven innovation, and equitable economic opportunity.
The main question is whether smaller models will continue to lag closed source models by 6-12 months, or if the gap will expand as capabilities advance.
Nvidia CEO Jensen Huang predicts AI and robotics will increase human workloads, countering Elon Musk's vision of more leisure time.
Nasdaq has filed a proposal with the SEC seeking permission to offer tokenized stock trading alongside traditional equity trading, allowing customers to choose between legacy methods and onchain transactions with equal priority treatment. The tokenized stocks would clear and settle through the existing Depository Trust Company infrastructure, with token holders receiving full shareholder rights, including voting and liquidation privileges. This move follows similar tokenization efforts by Robinhood in Europe and various crypto exchanges. The proposal aligns with SEC Chairman Paul Atkins' stated priority of clearing a path for tokenized securities markets. The regulator compares the transition to the historical shift from analog to digital audio formats.
SharpLink Gaming (SBET) has begun executing its $1.5 billion share repurchase plan, buying back 939,000 shares at an average $15.98. With 837,000 ETH ($3.6B) staked and no debt, the company says repurchases are “immediately accretive” while shares trade below NAV, boosting ETH per share for remaining holders. Staking income of $67M–$151M annually plus cash reserves will also help fund additional buybacks as SharpLink prioritizes capital allocation over equity issuance.
Aerodrome is rethinking token launches by eliminating the usual pre-sale, market maker payments, and CEX lobbying. Instead, projects seed liquidity directly on Aerodrome, incentivize emissions via veAERO votes, and let community liquidity providers and traders handle price discovery from day one. With Coinbase's new DEX integration, this model now connects transparent, community-driven launches to millions of retail users, avoiding the extractive dynamics of traditional launches.
Coinbase has acquihired the founders of Sensible, a DeFi-focused startup, to accelerate its onchain consumer strategy. The new team will help simplify access to DeFi and expand consumer use cases, supporting Coinbase's push to become an “everything exchange” where users can trade, stake, borrow, and spend seamlessly. Coinbase framed the move as part of its broader mission to make onchain finance the default for everyday money management.
Multiple stablecoin issuers are competing to partner with Hyperliquid for its upcoming USDH stablecoin, with GENIUS Act compliance and revenue sharing as baseline requirements, but the real contest centers on who can present the most credible vision for growing USDH beyond margin trading on HyperCore. While established issuers offer regulatory support and distribution, USDH would likely never be their top priority, whereas native teams like Native Markets (using Stripe's Bridge for compliance) could offer maximum ecosystem alignment despite higher execution risk. The GENIUS Act has effectively commoditized regulated stablecoins, shifting the competitive advantage from regulatory compliance to demonstrated long-term commitment and strategic alignment with the Hyperliquid ecosystem.
Stablecoins inadvertently support US monetary policy by creating global dollar demand through cryptocurrency trading, enabling the government to fund massive deficits through money printing. This printed capital flows primarily into AI development, which is viewed as critical for maintaining US economic dominance in competition with China. Investments should focus on crypto infrastructure ("shovels"), including stable blockchain networks, exchanges, and major cryptocurrencies like Bitcoin and Ethereum, rather than speculative tokens that primarily serve as liquidity-churning mechanisms. AI-driven media will amplify this cycle by encouraging more retail participation in onchain trading, ultimately reinforcing dollar demand through stablecoin usage.
MEV evolved from an initial design flaw that allowed transaction ordering to affect blockchain state into a systematized value extraction mechanism. Over $1.1 billion has been siphoned from Ethereum users since the Merge through block builders and searchers. Rather than solving the problem, Layer 2 rollups created new centralized extraction points through sequencer control, with networks like Base generating an estimated $50-200 million annually and major L2s capturing tens of millions from transaction ordering privileges. The economic incentives have created a permanent "invisible VAT" on DeFi interactions, where centralized sequencers have no motivation to decentralize due to the massive revenue streams from controlling transaction flow. This represents the entrenchment of value extraction at the infrastructure layer, where owning transaction ordering mechanisms has become a highly profitable rent-seeking business model with little accountability or competition.
Meta's stablecoin project came much closer to regulatory approval than commonly believed. The Swiss regulator FINMA was ready to approve Meta's license before political intervention killed the project. Stripe's Tempo faces the same fundamental flaw that doomed Libra: corporate chains with centralized control will inevitably break promises about neutrality and permissionlessness due to economic incentives to favor their own products over competitors. The core tension is between crypto's original vision of truly open, permissionless networks versus the pragmatic appeal of "slick, branded, proprietary chains" controlled by fintech giants. The success of corporate blockchain projects like Tempo would represent the failure of crypto as a revolution, simply replacing traditional financial monarchies with new centralized empires controlled by tech companies, potentially splitting along geopolitical lines rather than creating a unified global infrastructure.
Over the past five years, Strategy (formerly MicroStrategy) has outperformed BTC by 3x.
Ant Digital Technologies, the blockchain arm of Ant Group, plans to put over $8.4 billion in energy assets onchain.
Ripple is expanding in Spain via a new custody agreement with banking giant BBVA, which just rolled out retail crypto trading and custody for BTC and ETH.
Tempo, Stripe's new L1, is a Trojan horse for the global financial system.
Snapchat's Imagine Lens allows Snapchat+ Platinum and Lens+ subscribers to create and edit images using text prompts like "turn me into an alien." The open prompt image-generation Lens features both custom prompts and pre-loaded options for comic panels or caricatures. It costs $15.99 per month for Platinum or $8.99 for Lens+ subscriptions. The release demonstrates Snap's continued AI investment alongside its AR leadership.
NVIDIA released an AI Blueprint for 3D object generation that enables artists to create up to 20 3D objects from text prompts, streamlining the traditionally time-consuming prototyping process. The workflow integrates Microsoft's TRELLIS model via a new NVIDIA NIM microservice, which generates high-quality 3D assets 20% faster than the native model. Artists can automatically export generated objects to Blender or other 3D applications. The system works on NVIDIA GeForce RTX GPUs with 16GB or more memory.
Adobe is facing a new class action lawsuit accusing it of misleading customers with unclear subscription terms, hidden fees, and intentionally difficult cancellation processes that leave users feeling trapped. The case reflects growing backlash over Adobe's pricing, AI changes, and business practices, threatening the company's reputation among creatives.
Generative AI offers rapid creative output, but designers must resist treating initial AI results as finished work and instead focus on disciplined iteration and craft. The "EDEN Loop" framework (Explore, Diagnose, Elevate, and Nurture) helps designers harness AI as a creative multiplier rather than replacement, emphasizing the need for human judgment in art direction and brand storytelling. The approach cautions against allowing emerging designers to master AI prompts before they have learned fundamental design principles, instead advocating for AI as an amplification tool that still requires human pilots to navigate toward authentic, culturally sensitive brand work.
Slack's Design Operations team demonstrates how a small three-person team can drive significant impact by strategically dividing responsibilities across business strategy, people programs, and product systems. The team operates through focused leadership alignment, thoughtful designer experience programs, and scalable processes that provide structure while maintaining flexibility. Their collaborative approach emphasizes working smarter rather than larger, proving that lean teams can create meaningful organizational change when they prioritize strategic focus and cross-functional partnership.
Eurovision commissioned Sheffield-based agency PALS to refresh its brand ahead of the contest's 70th anniversary, embracing the studio's freelance “hub and spoke” model to create a flexible, digitally adaptable identity with a new beating-heart logo and custom typeface. The project highlights a shift away from traditional agencies toward diverse, freelance-driven teams, though fan reactions to the new look have been mixed.
Transform your ideas and documents into captivating slides with AI.
Start, iterate, and launch a website all in one place - no code, no drag-and-drop, just prompt.
Kelp is a UI library for people who love HTML powered by modern CSS and Web Components.
IDC research forecasts the global software design workforce will expand by over 30% from 107 million to 144 million people between 2025 and 2029. UX design roles are expected to grow fastest at 7.6% annually, outpacing other knowledge worker categories due to increasing demand for design-intensive solutions. Generative AI will accelerate product development while creating pressure for designers to deliver highly usable and visually appealing AI features.
The Iceberg UX Model emphasizes that 90% of an AI product's user experience lies beneath the visible interface, focusing on error handling, trust mechanisms, and failure recovery rather than just polished UI design. Most AI product teams fail by over-prioritizing visual design while neglecting invisible layers like fallback flows, confidence indicators, and system transparency that determine user trust and retention. A successful AI-first UX workflow should start with user needs, map failure states, test trust patterns, and build foundational invisible layers before designing the interface.
Paula Mastrangelo and Ramiro Espinoza launched “Brutal Types” as a separate foundry from Retype to focus on expressive display fonts inspired by Europe's interwar lettering, especially the Dutch avant-garde. Drawing on fieldwork, archives, and overlooked sources, they and their collaborators reinterpret historic alphabets into contemporary typefaces, offering designers distinctive alternatives for branding and media while planning future releases and a book on monumental Dutch lettering.
Effective animations serve clear purposes like explaining features, providing feedback, or creating spatial consistency, but they can harm user experience when overused or poorly implemented.
Illustrator Edie Medley transforms overheard pub conversations into playful drawings, capturing both the humor and intimacy of everyday social life while celebrating pubs as spaces of community.
Sponsored
Free trials help AI apps grow, but bots and fake accounts exploit them. They steal tokens, burn compute, and disrupt real users.
Every article about Ramp starts with the same question: Can it really keep growing this fast? It was doubling at $100 million, doubling at $500 million, and somehow still doubling at $1 billion in revenue, all while being cash flow positive. This piece digs into why its growth hasn't hit the wall everyone expects.
The unicorn market has split into rockets and zombies, with no middle ground left. Out of 1,609 unicorns worth $6.2 trillion, winners like Databricks ($100 billion valuation, 50% growth) and Supabase (jumping from $900 million to $5 billion) are staying private to keep the spoils for themselves, while 28% aren't growing at all and over half fail basic financial health checks. The best companies are avoiding IPOs because they're growing so fast they don't need public money, while the zombies have nowhere to go but down.
Cognition raised $400M at a $10.2B valuation in a funding round led by Founders Fund, with contributions from existing and new investors. The company's acquisition of Windsurf significantly boosted its ARR and product offerings in AI coding. The merger accelerated growth, positioning Cognition as a leader in AI-powered engineering solutions for top-tier customers like Goldman Sachs and Dell.
Snap has restructured into "startup squads" of 10 to 15 people to address stalled ad revenue and user decline. Snapchat+ subscriptions generate over $700 million annually, marking a significant direct revenue source. Snap is also developing AR glasses, aiming to revolutionize human-centered computing.
Bootstrapping allows founders to maintain control and minimize equity dilution using personal funds, but it poses financial risks and may limit growth. Raising venture capital provides cash, expertise, and validation but involves equity dilution and loss of control. With a hybrid approach, companies can bootstrap initially to prove their concept, then seek external capital for accelerated growth.
Ramp's data shows the "996" workweek, a 9 a.m. to 9 p.m., six-day-a-week schedule, is becoming prevalent in San Francisco, affecting multiple sectors beyond just tech. This shift creates intense competition, demanding faster product development and heightened responsiveness, especially for B2B companies. Leaders must decide whether to adopt 996, focus on efficiency, target different markets, or innovate hybrid models to maintain a competitive edge.
Sponsored
SCIM lets identity providers like Okta communicate user changes to your B2B app. This developer guide [13] covers the real-world gotchas. Want to add SCIM to your app WITHOUT building it? Talk to Tesseral