Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Software factories shift engineers from coding locally toward supervising cloud agents and improving the systems that produce software. Warp expects engineers to own product quality while reducing human touches per PR through measurement, automation, and factory optimization.
OpenAI is in talks with investors to raise at least $30 billion in a funding round at a valuation of roughly $1.4 trillion. The company previously raised $122 billion in March in what was supposed to be the last private raise before an IPO. OpenAI has now ruled out a public debut in 2026 due to AI safety issues. The new fundraising will serve as a bridge round to the IPO.
Sponsored
CData ran 22 models on live enterprise data. All returned the same
Baseten is partnering with OpenAI to serve open models natively via Codex and the Responses API.
Embed 5 has major gains over Embed 4 on visually rich documents, financial filings, parsed PDFs, code, and multilingual retrieval.
Devin is now 30% to 40% cheaper in Fusion and Normal mode, 15% to 20% cheaper in Ultra, and up to 70% cheaper in Devin Review.
Sponsored
When every agent hits your raw sources, one question can burn 45,000 tokens. Guru verifies knowledge once, then serves it to every agent
There is no clear performance leader and high variance in model-task fit on web tasks.
Anthropic's testing found that attackers can bypass GLM-5.3's safeguards between 64% and 100% of the time with simple techniques.
AI agents pose security risks as they may inadvertently cross security boundaries while pursuing legitimate goals.
The SEC updated its crypto FAQ to clarify that decentralized token buyback programs, where no central party is involved, do not constitute investment contracts under the Howey test. The revision followed feedback from a16z Crypto General Counsel Miles Jennings, who warned that prior wording could cause a buyback announcement alone to trigger securities classification for the issuer. The broader September 25 FAQ batch also addressed how the SEC's March digital assets interpretive release applies to token functionality and staking receipt tokens, including a clarification that services to maintain or grow a functional network do not qualify as "essential managerial efforts" under Howey. The FAQs carry no legal force as staff views, but the responsive revision gives DeFi protocols running on-chain buyback mechanisms clearer grounds to structure those programs without securities registration concerns.
Coinbase received CFTC approval to operate as a Derivatives Clearing Organization (DCO), adding clearinghouse status to its existing exchange and broker licenses to form a vertically integrated regulated derivatives stack. The approval order restricts Coinbase to fully collateralized products, with leverage expected via a future amendment. The clearinghouse designation positions Coinbase to back permissioned institutional derivatives markets through Hyperliquid's HIP-3 framework, which allows licensed partners to operate compliant perp trading venues. Coinbase becomes the second major crypto exchange after Kraken to hold CFTC-regulated clearing capabilities, completing the exchange-broker-clearinghouse triad that institutional counterparties require for compliant derivatives exposure.
Chainlink launched CCIP 2.0 on September 28, adding Cross-Chain Verifiers (CCVs) that let enterprises attach custom security checks before destination-chain execution, built-in KYC, AML, and sanctions screening on every transaction, AWS and Google Cloud deployment kits, and additive verification from Infosys, Nethermind, and Deutsche Börse's Crypto Finance, bringing the protocol's secured token value above $84 billion. Ethlabs integrated its Fast Confirmation Rule (FCR) with CCIP 2.0, drawing quick-action signals from Ethereum's existing validator consensus to drop cross-chain confirmation times from approximately 13 minutes to 12-24 seconds (30x faster) without new economic assumptions or additional infrastructure. The combined update targets institutional RWA issuers and tokenized asset distributors requiring compliance enforcement and sub-30-second settlement across chains.
Proof is building a derivatives exchange around a new instrument class called Conditionals, which price assets contingent on specific events and settle to the underlying position if the condition is met or revert to zero PnL if not. The design addresses a gap between prediction markets and spot/perp trading: existing event markets tell traders whether something happens but do not reveal what an asset is worth if that event occurs. US equities generated over 60% of their excess returns on roughly 13% of trading days tied to scheduled macro announcements, a figure Proof cites to frame the scale of event-conditional demand that current instruments leave unmet. The protocol runs on testnet after a June v0 Challenge with 500+ paper-trading participants, with a capped real-money Mainnet Private Alpha rolling out shortly.
"Shielded Bitcoin" is a metaprotocol that ports Zcash's encrypted-note, ZK-proof privacy architecture to Bitcoin L1 via OP_RETURN data envelopes, requiring no soft fork or sidechain. The design defers trustless peg-in/peg-out to a future paper, leaving the bridge problem open, and depends on Bitcoin Core v30's relaxed OP_RETURN relay policy. The paper arrives as ZEC has climbed ~2,500% over the past year to a ~$26B market cap (ranked #9), with Grayscale's ZCSH ETF drawing $250M+ in net inflows since August 25 and public ZEC holdings disclosed by Paradigm co-founder Matt Huang and Multicoin. Citrea (Galaxy Ventures) acquired privacy wallet Crest to target the same problem on Bitcoin L2, while NEAR Protocol rolled out confidential perps and limit orders by default, with both teams citing AI-powered surveillance tools, including LLMs re-identifying pseudonymous users, as the core demand driver.
Citi, with $2.8T in assets and a payment network spanning 94 markets and 300+ clearing systems, has partnered with Coinbase to offer institutional clients stablecoin payment services covering fiat-to-stablecoin conversion, onchain transfers, and stablecoin-to-fiat settlement. The arrangement layers Coinbase's crypto custody and $246B asset platform onto Citi's regulated banking rails, giving institutional clients a compliant on/off ramp to stablecoin flows without depending on crypto-native intermediaries. The deal extends one of the largest global correspondent banking networks to onchain settlement, putting stablecoin payment reach within regulated multicurrency infrastructure that institutional counterparties already operate inside. Citi also projects stablecoin issuance will reach $1.9T by 2030 in its base case, indicating the bank is building this infrastructure ahead of anticipated supply growth rather than in response to it.
In Flex's new Elite tier, personal cashback starts at 3% and rises based on business deposits kept with Flex. $250,000 in deposits unlocks 4% cashback, and $1 million or more unlocks 5%. Flex funds the higher rewards from revenue generated on business deposits. Several $100 million-plus companies have already switched from competitors because of the structure, a rare direct link between a business banking relationship and personal rewards.
Aave increasingly looks less like a lending protocol and more like a credit layer, as tokenized stocks cross $3.5 billion in market cap and Coinbase's tokenized stocks become productive collateral on Aave. Stocks, tokenized funds, custodied BTC and ETH, RWAs, and crypto collateral are all increasingly being originated through Aave, while on the borrower side, Coinbase can lend against tokenized equities, institutions can access stablecoin credit, fintechs can embed Aave liquidity, and banks may eventually launch their own fronts for the Aave app, calling the addressable market for Aave closer to credit broadly than DeFi lending specifically.
Krak offered instant EUR transfers with no conversion, a personal IBAN, and zero deposit fee.
Top 10 crypto card projects by cumulative user spend are RedotPay, EtherFi Cash, KAST, Katra, Gnosis Pay, Tria, Karta Personal, Kolo, Wirex One, and MetaMask.
EtherFi's own numbers show last week's $868,000 in revenue came mostly from its Crypto Card ($507,000), followed by staking ($266,000), Liquid vaults ($58,000), borrowing ($28,000), and swaps ($9,000).
A Bitwise survey of 15 of the world's largest institutions on their crypto allocations demonstrated that none of the institutions cut exposure during a 50% drawdown, and some added to positions.
OpenAI has launched Dots, always-on personal agents powered by GPT-6 Astra that pursue user-defined goals in the background rather than staying tied to a single interface. Available to ChatGPT Pro and Business Premium users in eligible markets, Dots can be launched from ChatGPT or Codex and contacted through platforms such as Slack and Teams, with text messaging planned. Their floating, cartoon-like avatars give autonomous software a more approachable identity, while OpenAI envisions specialist Dots working together with their own tools and permissions.
Replit has acquired Atta, an AI-powered business analysis workspace, bringing its data-analysis and charting technology directly into Replit's platform. Users can upload or connect a dataset and ask Replit Agent to investigate a question or generate a visualization without writing SQL. The move reflects a broader shift toward folding business analysis into the same environment used to build and operate software.
Formula E's new identity by MOX centers on a deliberately divisive logomark that combines an F and E at a 30-degree angle to convey speed and momentum. A bespoke type family based on AP Gavel expands the system with condensed italics for energy and a monospaced style for technical data, complemented by Inter for longer copy. Surge Purple, Signal Orange, and Electric Lime add a distinctive palette designed to give the championship a more expressive identity as it enters its GEN4 era.
Design isn't something that happens after fully understanding a problem. Making sketches and prototypes is itself a way to discover what the problem really is. Rough ideas expose gaps, tensions, and requirements that research, data, and discussion alone may never reveal. An excessive focus on outcomes can discourage designers from making anything, even though reaching a good outcome often requires producing and evaluating lots of imperfect work along the way.
AI creates opportunities to skip traditional tools and deliver the outcomes those tools were originally designed to help people achieve. An AI-powered newsroom called Exposit discovered that users overwhelmingly preferred asking for a topic and receiving a personalized, up-to-date report instead of navigating headlines, categories, and search interfaces. This “outcome first, tool second” approach suggests a useful product design question: whether a new AI experience needs another interface at all, or can simply produce what people came for.
Designers are rushing through their process to prove value through speed, but the real contest is winning stakeholder alignment, where they are both vulnerable and well positioned. The "Fastest Gun in the West" hijacks planning by anchoring early on an idea, using vague agile language, and pressuring designers to produce polished mockups that lock in unexamined assumptions while research is dismissed as slow. Since designers can't outrun someone shooting from the hip, they should use low-fidelity tools like scenario storyboards and PRFAQs to engage stakeholders on mental models and outcomes.
Initium gives your AI agent the style, strategy, blocks, and skills to generate a site that looks designed, not generated.
A curated library of real product onboarding examples from Notion, Slack, Stripe, Cal, and more. Patterns, components, and the activation playbooks that move retention.
Free online design battles in real time, with the same brief and kit for every entrant.
WCAG 3's September 2026 draft sets a single conformance level, called core requirements, built on WCAG 2.2's Level A and AA success criteria. Supplemental requirements, assertions, and recommended practices cover accessibility needs that can't be objectively measured or don't apply everywhere, encouraging organizations to go beyond the baseline. Tags let regulators tailor requirements to specific situations and let websites report progress toward and beyond conformance, with multiple reporting tiers built in.
Everypixel's blind evaluation of 863 pairwise judgments across 48 commercial scenes found no universal winner among 2026's leading AI image models. GPT Image 2 showed the strongest evidence for high-precision commercial generation and controlled photorealism, though its direct comparison against real photography wasn't statistically conclusive. Qwen Image 2 suits controlled editing, Nano Banana 2 fast iteration, Nano Banana Pro consistency, Grok Imagine creative interpretation, and FLUX.2 an explore-then-finish workflow.
Apple in Colour is an interactive archive from sheets.works that brings together 341 Apple product colors spanning nearly three decades, from the Bondi Blue iMac to recent iPhones, complete with release years and official color names.
Hong Kong-based illustrator Brian Yuen paints in Photoshop with a style drawn from the Old Masters and artists like Jaime Jones and Craig Mullins.
Westwing's 2026 advent calendar comes in an oversized house-shaped box illustrated by Munich-based artist Kera Till, with 24 doors hiding home decor, tableware, and self-care pieces.
Sponsored
Agents can generate code. Getting it right for your system is the hard part. You end up wasting time and tokens in the correction loops. More MCPs, rules, and bigger context windows give agents access to information, but not understanding. The teams pulling ahead have a context layer to give agents exactly what they need for the task at hand. Join us live on Oct 7 (FREE) [4]to see: * Where teams get stuck on the AI maturity curve and why common fixes fall short * How a context layer solves for quality, efficiency, and cost * Live demo: the same task with and without a context layer
Muse runs in a dedicated cloud VM where the agent operates inside an isolated runtime cell and never sees real credentials. A separate Sentinel controls connector actions and network egress, while host-side services enforce least privilege, approval scopes, and prompt injection defenses. The design assumes the agent can make mistakes or process hostile data, then limits the damage through layered boundaries and a public bug bounty.
This visual history traces text classification from bag-of-words and recurrent networks through transformers and Jev-style decision APIs. Jev is positioned between narrow specialist classifiers and general-purpose language models, offering broad zero-shot classification with lower latency and cost. The practical opportunity is to use calibrated decision models for routing, screening, evaluation, and other bounded choices inside agent systems.
A vibe-coded website became more distinctive when the work shifted from asking an agent for a polished result to building design taste and giving specific visual direction. The process shows how coding agents can make implementation cheap while increasing the value of references, critique, typography, and deliberate constraints.
Give Opus 5.5 a complete task, define what done means, and state when it should stop for input. The model already reasons before every reply, so prompts such as “think carefully” add delay without improving the result. For long coding runs, persistent task files and explicit rules about when to continue help the work survive context compaction and remain steerable.
Platform engineers often need to create their own roadmap by reading signals from systems, users, the organization, and the wider industry. Useful inputs include postmortem patterns, cost centers, user workarounds, migration holdouts, repeated leadership concerns, and ideas that have already matured elsewhere. The goal is to combine strong evidence with leading indicators instead of chasing only the loudest recent problem.
Agentic coding makes interfaces and features cheap to revise, but user data remains a one-way door that cannot be recreated after a destructive mistake. Schema design, migrations, backups, privacy, and retention therefore deserve slower decisions, explicit recovery plans, and careful validation even when an agent can generate the code in seconds.
Sponsored
Go deeper on training, inference, performance optimization, and scaling. Learn directly from the engineers and maintainers advancing PyTorch, vLLM, Ray, and the tools shaping production AI. Join us Oct 20–21 in San Jose and save 25% with code TLDR25.
GPT-6.1 Sol targets near-Astra performance on coding, computer use, and professional work at one-fifth of Astra's standard token prices. It improves on GPT-6 Sol by 6.4 percentage points on DeepSWE 1.1 and is priced at $2 per million input tokens and $10 per million output tokens. The model is available in the API, ChatGPT Work, and Codex, with an Ultrafast tier planned.
DevDay introduced dots, GPT-6.1 Sol, faster inference, cloud Codex environments, a refreshed CLI, code review, and Codex Security Cloud. The developer platform also gained a Decisions API for bounded classifications and an Agents API with computer use, multi-agent orchestration, tool search, and context compaction.