Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We're looking for a builder who has shipped real products/systems with LLMs. Click here to learn more [17].
OpenAI's model is a looped transformer, which means it reuses layers in the transformer block to increase capacity without adding parameters. This can significantly increase the size of the model without increasing the amount of storage and RAM needed to host it. However, it also increases the costs of running the model as embedded text has to run through more layers. The looped transformer aspect is just a small architectural tweak and is not the reason why Astra is a really good model.
World models could become the next major AI paradigm by helping systems represent environments, predict outcomes, simulate possibilities, plan, and act. The convergence of Yann LeCun, Demis Hassabis, and Fei-Fei Li suggests growing momentum around models built for decisions, not just generation.
Nvidia and CrowdStrike have introduced a new family of agentic AI models dubbed SafeMind that can both find and close attack paths for customers.
TxBench-AB is a novel AI benchmark that assesses LLMs' effectiveness in biomedical research.
Shamez Hemani, a senior OpenAI data center employee who left the company in April to join Meta's dedicated compute team, is now a member of Anthropic's technical staff.
This tool identifies the text watermark Claude uses to produce files.
Unit 42 investigated a ransomware attack using frontier AI, where a human attacker breached an enterprise network with unprecedented speed.
Goldman Sachs, Bank of America, and several other major financial institutions in the US are forming a consortium to jointly issue a dollar-backed stablecoin, targeting a 2027 launch. The collaborative model suggests the banks aim to share infrastructure costs and regulatory compliance burdens rather than build competing products in parallel. A coordinated TradFi entry from systemically important institutions would put the product in direct competition with USDT and USDC, which together account for the majority of global dollar stablecoin supply. The move reflects broader bank momentum toward stablecoin rails as US federal stablecoin legislation has provided clearer issuer requirements and institutional on-ramps.
CEO Phong Le defended Strategy's sale of ~7,000 BTC at $60K-$65K (to fund preferred dividends) ahead of resuming purchases near $80K, framing both moves as driven by capital costs, not price predictions. During the ~10-week pause, the company cut net debt from ~$7B to zero and built ~$7B in cash. The sold BTC was under 1% of holdings. Strategy grew its stack ~30% this year and says it would keep buying at $90K-$130K if capital economics stay favorable. Bernstein kept an Outperform rating but cut its price target to $350 from $450.
3Jane launched a product letting users earn 20%+ APY on Aave USDC/USDT deposits by making that capital "callable". This means staking $75K as margin while committing up to $1M of capital that stays usable elsewhere until called, earning Aave's base ~3.5% plus a 1.25% standby fee on the full commitment. If called, users have 9 days to fund it, with a 35-day cooldown after. Framed as a sign of DeFi maturing beyond simple lending/swaps into markets for balance-sheet capacity itself.
EthenaPay is a crypto neobank built on Avalanche that pairs a traditional IBAN with self-custodial stablecoin accounts, letting users receive funds via standard banking rails and hold them in stablecoins. The product offers a 6% dollar savings rate, 5% card spend cashback, free global onramps in USD, GBP, EUR, and local currencies, and borderless instant transfers. The iOS app is live now, with a "Buy Now Pay Never" feature that routes savings yield to cover card balances automatically, positioning Ethena's USDe yield infrastructure as a direct-to-consumer fintech layer.
Real-economy stablecoin payments were only about $390B in 2025, and B2B accounts for ~$226B of that, dominated by USDT and driven by small and mid-sized importers, not crypto-native firms. Asia is the current volume center, LATAM shows strong institutional B2B rails despite a declining US-Mexico remittance corridor, Africa is the least-served and highest-margin region due to thin payout coverage, and MiCA rules make EU-touching corridors route around USDT. Infrastructure is consolidating fast into incumbents, leaving genuine white space only in Africa payouts, intra-LATAM, Turkey/CEE, and mid-market Asia sourcing: corridors with the widest cost gap and the thinnest coverage.
Argentina's stablecoin adoption traces to a 2001 banking crisis in which the government froze deposits and forcibly converted dollar accounts to pesos, wiping out roughly three-quarters of the peso's dollar value, with usage accelerating after 2019 capital controls capped official dollar purchases at $200/month. Today, 94% of peso crypto trading flows to stablecoins, the highest share of any major currency Artemis tracks, while downloads of the country's top 15 crypto apps rose 93% in 2024 and 1 in 5 Argentines now use crypto. After Argentina lifted most capital controls in April 2025, the official and parallel exchange rates converged to within ~4%, removing the structural arbitrage that originally made stablecoins essential for dollar access. Lemon wallet downloads climbed each quarter even as monthly inflation fell from 25.5% to 2.1%, and USDC contractor payroll share leveled off rather than dropped, indicating stablecoins have shifted from crisis hedge to standing financial preference.
Robinhood Wallet and social trading app Fomo allow users to purchase meme coins, including WIF, through Crossmint using Visa, Mastercard, Apple Pay, and Google Pay without separate KYC verification, according to a The Block investigation. Crossmint routes these transactions as "digital goods/media" rather than crypto purchases, a classification that bypasses standard credit card network crypto restrictions and qualifies buyers for rewards points or cash back they would not earn on a properly coded crypto transaction. Chase flagged the Visa transactions as incorrectly classified and has asked Visa to investigate, while the New York Attorney General's Office confirmed it is reviewing the matter. Crossmint has defended its classification approach as scrutiny from both a major bank and a state regulator mounts.
Robinhood Chain's daily fees doubled to $2.13 million, sending ARB up 25% to $0.1087 as the Arbitrum DAO's 10% licensing cut translates to roughly $192,000 per day at current rates. Chain DEX volume climbed 89.5% weekly to $1.56 billion, and TVL reached $738.5 million, though memecoin speculation rather than tokenized equity trading drives the bulk of that activity. The move lifted adjacent tokens across the L2 and DeFi sector, with CRV gaining 14.1% and OP rising 13.9%, while Bitcoin held near $78,000 and total crypto market cap stood at $2.72 trillion. ARB remains 95.5% below its January 2024 peak of $2.39, placing market cap at roughly $726 million despite the week's gains.
New self-custodial card from Ethena Pay offers tiered cashback and up to 6% APY on balances.
Privy launched native fiat on/off-ramp orchestration via its wallet APIs, letting developers move funds between bank accounts and stablecoins/Privy wallets.
An autonomous AI agent experiment failed to gain traction selling to human customers but found small-scale success pivoting to audit agent-facing products, pointing to agent-to-agent commerce as an emerging niche.
Combined Kalshi and Polymarket monthly volume fell 14.5% to $45.33B in August, the first decline in a year, following a World Cup-driven summer spike.
Sponsored
They're all showing up to OPENSEARCHCON NORTH AMERICA [5] to share how they run search, observability, and data infrastructure at massive scale. WILL YOU BE THERE TOO? Get your ticket now [6] for three days of practical learning and direct connections with other technical leaders. Here's what to expect: * REAL-WORLD IMPLEMENTATION STRATEGIES: vector search, agentic AI workflows, observability pipelines * TECHNICAL SESSIONS: search relevance tuning, plugin development, and vector database optimization * HANDS-ON WORKSHOPS: build a full observability stack with OpenTelemetry, or a working agentic AI search application. * $250 CERTIFICATION VOUCHER toward the Linux Foundation's OpenSearch Certified Associate exam.
Netflix's member-sessionizer pipeline turns tens of billions of hourly client events into viewing sessions, but scale bottlenecks made ordinary tuning painfully slow. Instead of optimizing one Spark run at a time, the team built a parallel experiment loop that launched, monitored, and compared multiple production-scale tests to expose memory and shuffle bottlenecks faster.
Lyft retired its internally built Flink Kubernetes operator and migrated its streaming fleet to the open-source Apache Flink Kubernetes Operator. The useful lesson is the migration pattern: run both operators in parallel, preserve the user-facing deployment contract, patch gaps such as default restart behavior, and then use the move to simplify ownership and upgrades.
Databricks traced silent MCP tool failures that caused agents to repeatedly retry, revealing seven small bugs responsible for an estimated $499,000 in wasted tokens and $1.2 million in annual lost productivity. Fixing them took about an hour, highlighting the value of tracing tool calls and designing tools to accept the reasonable input variations generated by LLMs.
Modern apps can write successfully, invalidate cached state, and immediately read from a replica that has not replayed the change yet. PostgreSQL 19's WAIT FOR LSN lets applications route reads to replicas while waiting for a specific WAL position, turning read-your-writes consistency into a measured timeout budget instead of primary pinning or sleep hacks.
AWS bought DuckLabs to gain its core team and influence DuckDB's roadmap, positioning DuckDB, DuckLake and Quack as foundations for S3-based analytics and AI agent workloads. DuckDB remains independently governed and open source, while MotherDuck will likely differentiate through higher-level tools for pipelines, agents, and data applications.
NVIDIA demonstrates how modern CUDA tools can make GPU code safer and dramatically faster by combining better debugging, optimized CUB algorithms, pooled memory, pinned transfers, and asynchronous streams. These incremental changes reduced an image-processing pipeline from 6.8 seconds to 23 milliseconds, a roughly 300x speedup.
Named queries turn SPARQL from an expert-only endpoint into a governed API surface. By registering queries as RDF resources with parameters, output contracts, metadata, and ACLs, teams can expose graph data to applications and agents without giving every caller raw query access or forcing them to understand the whole ontology.
FrontierHarness found that coding-agent harnesses achieved relatively similar pass rates but varied by up to 17x in cost, showing that tooling, caching, and model compatibility strongly affect efficiency. Codex had the highest pass rate at 66.7% with moderate cost, while Exo was cheapest and DSH Minimal was fastest.
MLPerf Storage v3.0 adds benchmark tests for KV cache and vector database workloads, plus an S3 object-storage access layer alongside POSIX. That makes the suite more relevant to modern AI systems, where inference caches, vector indexes, checkpointing, and object storage can be the real bottlenecks behind model throughput and cost.
dbt doctor is an open-source static analysis tool that scans dbt projects for missing documentation and tests, schema drift, stale models, DAG complexity, and governance gaps. It can run locally, through coding agents, or in CI to score project health and block changes that introduce quality issues.
Large-scale operators often default to line charts even when they need to answer different questions about capacity, latency, fleet sizing, and distributions. A broader telemetry visualization toolbox is needed that matches the operational question instead of forcing every signal into time-series lines.
Cloudflare's Cache Transcoding prototype compresses eligible text assets with Zstandard before storing and transferring them, reducing their size to roughly one-third on average. This could unlock petabytes of additional cache capacity and reduce bandwidth usage for only a small increase in CPU consumption.
Google's TimesFM-3 extends time-series foundation models from univariate forecasts to multiple interacting series with past and future covariates, aiming to make retail, observability, finance, and operations forecasting work without task-specific fine-tuning.
Keenable SELECT lets agents research the live web using read-only DuckDB queries that combine search, filtering, and AI-powered extraction.
Google Pics, an image creation and editing tool built on the Nano Banana model, reaches Google AI Pro and Ultra subscribers and most Workspace business customers. Object segmentation, in-image text editing and translation, and collaborative editing give precise control over an image rather than regenerating it from scratch. Integration is available in Docs and Slides, with Drive arriving in the coming weeks.
Custom generative plugins and shaders, built by prompting the Figma agent, gain animated and interactive shaders that respond to motion and mouse input. Creations can now be published to the Community or to an organization, and a code viewer exposes the underlying shader and plugin code. MCP updates let third-party agents modify plugins and shaders, with shaders rendering correctly in generated React code.
Leaks continue to point to Dark Cherry/Burgundy and Sky Blue as two of the iPhone 18 Pro's launch colors. The main uncertainty is the third finish, with some well-known leakers claiming Apple will offer black while others believe silver is more likely. Until Apple's announcement next week, both options remain plausible, though recent reports suggest silver and black are the final contenders.
Ecosystem design pushes past the product boundary, shaping the conditions and capabilities people need to succeed on a platform rather than the interface itself. Integrations, partnerships, marketplaces, developer platforms, community development, and corporate development are the six surfaces where a product becomes a world other people build on. Five practices follow: design the whole customer loop, treat third parties as users with business models, build trust, tend community, and treat docs as a product.
Product discovery works as weekly customer touchpoints by the team building the product, not as a kickoff phase that hands a backlog to engineering. Discovery and delivery are two kinds of work on one team, and a high kill rate is the test that discovery is real. Recruiting is the bottleneck: a Monday-to-Friday cadence survives only when a conversation is already on the calendar, even a customer-support ride-along.
AI can dramatically speed up individual tasks like research synthesis, prototyping, and code generation, but that doesn't automatically make the entire product development process faster. Often the work doesn't disappear—it shifts into review, verification, refactoring, or downstream teams, creating hidden costs that can erode the initial gains. The real question isn't whether AI increases local productivity, but whether it improves the whole system's ability to turn ideas into reliable products without accumulating technical debt, instability, or rework.
Visual precision for building with coding agents. Measure, inspect, and capture exactly what you see in the browser.
Cursor Craft is a native macOS app to browse, install, and switch custom pointer sets — minimal, gaming, anime, nature, and more.
The best branding projects from around the internet.
A ghosted double image in DuckDuckGo for iOS's floating toolbar survived three rounds of retiming because the bug was view ownership, not animation timing. Recording the transition at 60fps and stepping through it with ffmpeg showed the tab-count badge fully visible 24 frames before the snapshot animation began. Two failure modes generalize: snapshotting a view's bounds instead of what it draws, and resizing a snapshot instead of scaling it.
A new typeface family explores how far geometric reduction can be pushed without sacrificing readability, drawing inspiration from early modernist experiments in typography and visual communication. Built as both proportional and monospaced families with extensive stylistic alternates, it reimagines the unrealized radical potential of classics like Futura while balancing mathematical purity with practical usability. The release extends beyond the typeface itself through a coordinated visual identity of 3D animation, motion design, and sound, all expressing the same philosophy of clarity, reduction, and geometric transformation.
Disney World's upcoming 55th-anniversary logo is earning praise for cleverly combining two “5”s into unmistakable Mickey Mouse ears while reviving elements of the resort's original 1971 globe logo. Fans have responded far more positively than they did to the 50th-anniversary mark, praising its simplicity, nostalgia, and stronger visual concept.
Sand Castle University, founded in 2017 and now staffed by five instructors, turns a childhood beach pastime into architectural sculpture packed with windows, brickwork, and staircases.
In Pursuit of Magic took the London studio Art&Graft two years and 42 people, all answering one question: what is the pursuit of magic for you?