Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
OpenAI recently released an impressive-sounding case study about how it helped Asana save $5.9 million with a single migration. The report claimed Asana cleared five years of engineering work in two weeks with Codex. However, the estimate of four engineers spending five years on the project seems inflated. Regardless, it is clear that AI makes previously impractical migration viable.
Mark Zuckerberg recently indirectly took aim at rival Anthropic in an essay, saying that leading AI labs were trying to consolidate power while painting the future as filled with doom. The 6,500-word essay claims the balance of power will favor larger institutions over individuals if those labs lead. Meta is a heavy user of Anthropic's AI products. It is internally projected to spend as much as $10 billion annually on Anthropic's services, making it one of the AI lab's largest customers.
Demand for frontier-level models is likely to keep compounding. However, the demand for fast, cheap, and 'good enough' models is just about to take off. Most people just want something responsive that handles basic tasks. There's still a lot of work needed to make these models a reality for business, but it is just a matter of time.
Sponsored
Not sure your new software is paying off? This free Capterra guide reveals 4 signs your implementation is succeeding — and what to fix if it's not. Get your guide now [12].
Taste is about making choices, and whether those choices are good depends on who they are made for.
OpenRouter allows Stripe to be the bank that every AI company is currently forced to build for itself.
Chroma's Foundation is a memory layer that operates through a swarm of agents modifying shared state, ingesting coding agent traces and company data to build a durable record and index.
Each passing quarter with continued hypergrowth in frontier AI revenue provides important evidence on AI's trajectory.
Harness engineering is the practice of surrounding AI-assisted code generation with deterministic tooling, agent-based review, and periodic entropy checks so that AI-generated code stays correct and coherent over time.
This post talks about five successive changes Cloudflare made to how cache entries are stored in memory that resulted in the per-entry footprint being cut by over 50%.
Sponsored
Access 500+ MODELS FROM 80+ PROVIDERS, with 200T+ TOKENS ROUTED MONTHLY through OPENROUTER [5]. > BEST EXECUTION: intelligent routing (price / latency / quality) and Auto Router. > BUILT FOR RELIABILITY: multi-provider failover and degradation routing. > MORE CONTROL: per-request policies, team controls, observability and auditability. > EASY TO SWITCH: OpenAI-compatible API; keep existing SDKs and switch models with a parameter. START BUILDING NOW
Anthropic opened a research preview of the Model Hardware Standard, a model-agnostic specification for letting AI agents operate scientific and manufacturing equipment.
H3 Max is a post-trained version of MiniMax H3 optimized for maximum speed. It can generate a 5-second-long video in under 3 seconds. The model is now available on fal for 50% off for the first week. This article discusses how the model was trained.
Cohere Parse converts complex, multimodal files into structured machine-readable data. A cost-effective vision language model for processing large volumes of enterprise documents, it detects and understands key visual elements and can process documents and images across nine major world languages. Parse can be accessed through the Cohere API for just $1.50 per 1,000 pages. A free version is available in Cohere Space for users to try.
OpenAI introduced large discounts on some of its models between July 27 and August 14. This resulted in Luna token usage jumping 13.8x and Terra token usage rising 5.6x while Sol, which remained at list price, only saw a bump of 1.1x. Most of the share gained by OpenAI's discounts came from other labs as opposed to cannibalization within the OpenAI family of models. Nearly a third of users who tried a discounted OpenAI model kept using it after the discounts expired.
A detailed benchmark of MiniMax-H3 video generation on 8× H200 GPUs showed SGLang Diffusion reaching 1.95x lossless speedups and as much as 6.24x with step reuse and sparse attention.
Anthropic and OpenAI show impressive revenue growth, with expectations for continued rapid expansion. Forecasts indicate moderated growth for semiconductor stocks but an uptick for software stocks, with AI's impact on the software sector remaining a key consideration.
AI is driving more distributed business formation in the US, with a rise in new companies forming outside major cities and in outer suburbs. This trend began during COVID-19 with remote work and accelerated with AI, yet business formation does not correlate directly with work-from-home levels. AI labs and product companies still cluster in major metro areas like San Francisco, highlighting the continued value of agglomeration for innovation.
Sponsored
Gartner has the data [13]. Every year, they interact with 10K+ CIOs and their teams, 80K+ business executives, and 5K+ technology providers. Their AI Hub already has over 4,000 use cases and case studies. You don't have to guess your way to AI success. Explore the
Adding scaffolding around text-to-SQL models improves performance somewhat, but this is limited by the base models' capabilities. This suggests that the task knowledge contained in the scaffold belongs in the training signal instead. Expert judgment needs to be part of the training process. Putting task expertise into task-specific training is ultimately what scales.
Codex has added 'persistent' to the reasoning-effort protocol and TypeScript SDK types. When a user targets a custom Responses-compatible provider whose model-defined effort is literally persistent, this now deserializes to the new Persistent variant and is unconditionally rewritten to disabled. Previously, it remained Custom("persistent") and was forwarded unchanged. Existing configurations and resumed sessions can therefore send a different value and change behavior or be rejected, so this alias should be limited to providers that define the translation.
Google introduced Gemini Omni 1.1 Flash with new controls for extending scenes, interpolating first and last frames, 4K upscaling, and faster video iteration through the Gemini API.
Halo Neuro introduced Sopro V2 and open-sourced Sopro V2 Turbo, a 120M-parameter multilingual voice-cloning model that can stream on laptop CPUs and in browsers.
The average sell-side estimates for Nvidia's FY 2028 revenue a year ago were around $310 billion. Nvidia guided for around 70% revenue growth in FY 2028 yesterday, which translates to approaching around $700 billion in revenue next year. While analysts kept updating the estimates throughout last year, they were still short by around $125 billion. Nvidia's revenue growth of around 70% could be the floor for next year.
OpenAI and Anthropic are seeing unprecedented growth, with revenues surpassing $100 billion combined, fueled by rapid adoption and technical advancements. This swift rise challenges expectations, as typical tech growth slows at such scales, potentially reshaping the economy if sustained.
Sponsored
They map how work happens before automating it — no surveys, no
Terminal-Bench-Science evaluates AI agents on workflows from researchers' own work.
Anthropic is putting a great deal of faith in Claude Code's auto mode for protecting its coding agent users against prompt injection attacks.
DeepMind introduced the world's first double-blind evaluation for AI models, using cryptographic environments to prevent benchmark contamination.
Anthropic seeks a new Head of National Security Sales to revive its defense contracts, following previous blacklisting by the Department of Defense over technology use concerns.
DeepSeek is seeking to raise $7.4 billion to bankroll research and development and the buildout of computing infrastructure.
Analysts have identified an $81,000-$86,000 supply zone as the key barrier Bitcoin must clear to recapture its January 2026 highs, with four factors overlapping in that band: long-term holders sitting near breakeven, self-custody accumulation starting around $80,800, options dealer hedges activating near $82,300, and dense short liquidation clusters between $82,000 and $86,000. The current recovery was catalyzed by the US Treasury's buyback expansion announcement, followed by the largest short liquidation event since 2019 on August 19, which compressed futures open interest 11% in coin terms. Eight consecutive days of ETF inflows totaling $2.8 billion support the rally, with a close above $83,300 alongside continued inflows serving as the signal that overhead supply is being absorbed. A retreat from this zone would expose $62,900, the level that would erase the entire recovery.
Security firms estimate an $8.7M loss after an attacker manipulated the price of the illiquid MAMO token used as collateral, then borrowed real cbBTC against it. Borrowing/supply on Base Core Markets has been capped to 1 wei as a precaution. WELL is down ~13%, MAMO down ~9% in 24 hours.
Revolut, the European neobank with 80 million customers globally, announced EURR, a Euro-denominated stablecoin deployed on Ethereum mainnet and targeted at European Economic Area users. The launch adds a high-distribution issuer to Ethereum's Euro stablecoin segment, which has lagged USD-pegged counterparts in adoption and liquidity depth. Revolut's entry follows the PayPal PYUSD model of a regulated fintech issuing its own stablecoin to capture payment flows onchain, and arrives as MiCA's stablecoin provisions create clearer issuance pathways across the EU. If Revolut integrates EURR into its core app for transfers and spending, its existing euro user base could drive adoption that reshapes Euro stablecoin market share on Ethereum.
LayerZero is launching a blockchain-based exchange targeting financial institutions this fall, backed by Citadel Securities, DTCC, and Intercontinental Exchange. The trio covers primary market-making, central US securities clearing, and exchange operations, positioning LayerZero's cross-chain messaging infrastructure as a settlement backbone for institutional capital markets.
A breakdown of neobank revenue streams beyond interchange: on/off-ramp spreads, FX markup, yield/treasury products, credit cards, float on customer deposits/stablecoin balances, lending spreads, payment processing, wire/transfer fees, payroll, travel commissions, subscriptions, brokerage/investing, and embedded verticalized software (like AI billing or AR-collection agents). The core thesis is that the more of a customer's financial life a neobank owns, the more monetizable surfaces it creates and the harder the platform becomes to replace.
A study of 293,273 capacity scenarios across 78 Ethereum token families finds that Uniswap v3 pool multiplicity delivers almost no distributed execution capacity in practice. The median HHI across scenarios sits at 0.930, the top pool captures a 96.3% median share, and only 0.67% of scenarios achieve an effective pool count above 2, with multi-pool routing adding just 3.2% capacity on average. Reconstructed daily pool states show actual trading concentration exceeds the already-concentrated liquidity-layer capacity on 67.6% of matched family-days, with a Spearman correlation of 0.841 between observed and modeled distributions. The analysis covers liquidity-layer capacity only and excludes gas costs, routing friction, and MEV, meaning measured concentration represents a floor, not a ceiling.
The Vector co-founder traces how the Solana social trading app, acquired by Coinbase in late 2025, peaked at $20M daily volume and ~$1B all-time volume before stalling when it pivoted to pro desktop tools after memecoin activity concentrated in ~5% of its user base. Fomo pursued the inverse strategy, targeting TikTok and Instagram audiences new to onchain trading rather than competing for existing crypto-native traders, and recently crossed $100M in daily spot volume. Crypto Twitter is a beachhead that validated the social-signal-plus-execution thesis but represented only a fraction of the eventual market. As equities, RWAs, and prediction markets migrate onchain and AI sharpens signal discovery, a social trading graph paired with execution rails could scale to a $1T category.
Ethlabs, a 9-week-old nonprofit R&D lab focused on making Ethereum the global settlement layer, emerged from its first in-person offsite with four concrete priorities. EIP-8198 (Quick Slots), which makes slot time configurable to bring the 12-second block time down in stages, is advancing in the Hegotá fork discussion, and a major bridge is preparing to adopt the Fast Confirmation Rule (FCR) in production, promising up to 32x faster deposit confirmations in best-case scenarios. On interop, the team landed on a 1-second cross-chain movement goal and separated problems requiring protocol changes from those needing ecosystem coordination alone. The lab is also committing to substantive work on ETH as an asset, a topic protocol organizations have long sidestepped in favor of Ethereum-as-platform framing.
A dozen-plus banks are moving forward with a joint stablecoin venture after a year of lobbying against stablecoins in favor of tokenized deposits.
A Seoul court ruled in Bithumb's favor in one of four lawsuits tied to February's error, where the exchange mistakenly credited 620,000 BTC (~$43B at the time) to users in a promo campaign.
DeFi is rebuilding TradFi cheaper/faster.
KAST's new business card provides 3% cashback uncapped on the free tier (2% after a 2026 launch bonus expires), 4% on paid plan up to $1M then 3% uncapped, up to 12% APY on idle cash (8% post-bonus), and 5% ad-spend cashback for the first 100 approved businesses.
The smartphone industry appears to be coalescing around the book-style foldable design, where a phone unfolds into a small tablet, rather than the flip-phone format that folds a standard phone in half. Samsung's latest Galaxy Z Fold, Apple's rumored iPhone Ultra, and leaked images of Xiaomi's upcoming 18 Fold all point to growing momentum behind larger-screen foldables, suggesting that this form factor is emerging as the preferred direction for premium folding phones.
Adobe has released a major Photoshop update focused on making image editing faster and more intuitive, with new features including a Light Adjustment Layer that brings Camera Raw-style lighting controls directly into Photoshop, Firefly-powered editing tools that can automatically mask and modify specific parts of an image from text instructions, and AI Markup tools for visually communicating edit requests. The update also introduces a beta AI-Assisted Editor, allowing users to make complex edits through natural-language prompts while Photoshop automatically selects and combines the appropriate tools. Adobe says the goal is to let creators spend less time navigating software and more time focusing on creative decisions.
Presentation startup Gamma has acquired Accel-backed Lica to build a design research lab, which Lica's co-founders will lead. Lica raised $4 million in 2024 and worked with e-commerce sites on brand-compliant marketing videos. Gamma now wants to explore fluid, multimodal presentation formats through the new division.
Generative AI can deliver a solution before there is time to explore the problem, removing the detours where creative learning actually happens. Photoshop began stripping away that friction. AI takes it much further, especially for students and young designers. The proposal is not banning AI but designing for Learning Experience alongside User Experience.
Design system pattern names like Overview and PageHeader are shorthand for decisions a team should not have to reopen every time. Learning that vocabulary takes a person months, but an agent can read the repository and retrieve the code, guidance, and examples behind each name. That lets more people describe a screen in the system's own language and judge the result.