Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Bluesky Protocol Services organizes all the documentation developers need within the ecosystem, clarifies the service contracts around Bluesky-provided infrastructure, replaces the old docs site, and gives the team a clean way to ship future releases.
Cami Clark keeps a low profile, but is a key adviser to Anthropic chief Dario Amodei.
Sponsored
Security teams are adapting AI at unprecedented pace, but attackers are not lagging behind: 78% of orgs saw AI-enabled attacks, and 95% of leaders believe adversaries are already using AI. The 2026 SANS AI Survey Insights report [5] breaks down where AI is paying off, where it's lulling teams into a false sense of security, and why "formal AI governance" often means less than leaders think. For a deep dive into the survey results, [6]catch the related webcast [6] [6]with Dave Shackleford, founder of Voodoo Security. Browse more SANS AI resources [7] for ways to build, break, and defend AI in production.
OpenAI previewed Ultrafast, a GPT-5.6 Sol mode capable of generating up to 750 output tokens per second and operating at up to 14x standard processing speed. The approach aimed to deliver real-time performance without switching to a smaller model.
Google is rolling out Gemini 3.7 Flash, a new version of its workhorse AI model. The release comes just three weeks after the release of Gemini 3.6 Flash, an unusually short turnaround attributed to developer feedback and algorithmic improvements. API prices have been temporarily cut in half. Gemini 3.7 Flash will cost $0.75 per million input tokens and $3.75 per million output tokens until the end of the year.
Investors anticipate Anthropic's IPO could exceed a $2 trillion valuation, potentially becoming the largest in history, fueled by rapid revenue growth and demand for its AI models. Despite regulatory hurdles and competition, the startup projects annualized revenue of $100-120 billion by the end of 2026. Investors remain optimistic, citing Anthropic's strong market position and performance compared to rivals like OpenAI and Google.
Much of the financing Anthropic needed for its infrastructure buildout was assembled before the company's revenue spiked. The buildout suggests that financing is unlikely to be the immediate limit on frontier computer growth. Institutional investors appear willing to lend capital against commitment to long-term payments, especially if more established organizations are willing to back up part of the risk.
Anthropic examined how individually benign behaviors in frontier AI agents could compound into systemic failures when many agents interact in shared environments. The research highlighted risks including confabulation, reward hacking, and unexpected dynamics emerging faster than human institutions can oversee them.
Recursive agent systems should be designed as dependency graphs because upstream mistakes can propagate across many workers while leaf errors stay localized. Reliability depends less on nesting depth than blast radius, requiring provenance, verification, and stronger controls at high-impact nodes.
Sponsored
AI agents are autonomous actors, operating at a speed and scale that challenges today's security frameworks. “_From Zero Trust to Agent Trust_,” from Teleport [15], explains why Zero Trust must extend to Agentic Trust Principles: Enforce Continuously, Bound Collective
Agent Plugins package skills and MCP dependencies into one portable, vendor-neutral folder that compatible clients can load. The standard reduces fragmented setup through shared manifests, portable paths, isolated failure boundaries, and client-specific extensions, while leaving authentication unresolved.
Agents are only as capable as the environments they work in. Cursor can now continuously prepare development environments in the background at no additional cost. Agents can now start in ready environments, resulting in up to 3x faster responses. Agents will keep using the last successful build, and work continues uninterrupted while developers debug in the background.
Mistral OCR 4.1 is a highly specialized vision-multimodal model engineered specifically to ingest, parse, and structure complex document formats. It excels at reading complex tabular layouts, understanding hierarchical document structures, and directly outputting clean, machine-readable JSON or Markdown. The release pressures cloud providers and proprietary players to re-evaluate their pricing tiers for multimodal inputs.
TLDR is hiring a GTM Engineer to join our Applied AI team and own our AI-native GTM stack. We're looking for someone comfortable building AI agents and working with HubSpot. Click here to learn more [20]!
Google Sheets canvas is a Gemini-powered feature that turns spreadsheet data into custom, interactive mini-apps inside Google Sheets. The new layer sits directly on top of a spreadsheet to give users a visual way to organize, edit, and navigate information without formulas, programming, or a separate app. The visual layer is not static and will update along with the spreadsheet underneath it. Sheets canvas is now available globally in English to Google AI Pro and Ultra subscribers.
Denise Dresser, OpenAI's Chief Revenue Officer, leaves the company just ahead of its anticipated IPO, marking the second executive departure this week. Dresser, who joined OpenAI from Salesforce less than a year ago, will be replaced by Dali Rajic from Wiz. Earlier this week, OpenAI executive Brad Lightcap also announced his departure to start a new venture.
Sponsored
Scribe Optimize captures how work happens — no surveys, no
A China-trained model could give Apple greater control over its AI offerings in the country.
Writer, which offers AI tools and agents for marketers, has launched a new flagship model called Palmyra X6 that provides deployment-ready capabilities at a much lower price.
Google is testing a new Agent management UI in AI Studio, adding a dedicated tab for managing Cloud Agents within Google Cloud projects, not in consumer-grade sandboxes.
KPMG US issued an unqualified "clean" audit opinion on Tether International's full 2025 financial statements, marking Tether's first complete independent audit since promising one in 2017. KPMG examined the full balance sheet, income statement, equity changes, and cash flows, physically counting every gold bar in reserves rather than relying on custodian reports. Audited statements showed reserves exceeded liabilities by $6.8 billion at year-end 2025. Tether previously only published periodic BDO Italia attestations, not full audits, and paid $18.5 million and $41 million to settle New York AG and CFTC actions in 2021 over reserve claims.
Baltimore filed consumer-protection lawsuits against Kalshi and Polymarket alleging their sports-event contracts are unlicensed gambling. The Kalshi suit also names distribution partners Coinbase, Robinhood, and Webull for offering Kalshi contracts through their own apps. The Polymarket suit separately alleges the platform blurs prediction-market and sportsbook lines through an internal market-making team that can trade against users. The city seeks penalties, restitution, and an injunction. Kalshi says it will fight the claims, calling itself a regulated, transparent marketplace.
Yukon Research is a platform where distributed teams of humans and AI agents compete on verifiable science and engineering benchmarks, with submissions scored by deterministic sandbox evaluations before merging into shared repositories. In its first two months, participants on the ECDSA.fail challenge surpassed Google Quantum AI's quantum circuit benchmark by 50.3% within 8 hours of launch, while other challenges produced a 9.5x throughput improvement for Lighter's exchange prover and a 3.5x increase in post-quantum Ethereum scaling. Researchers publish a challenge by installing the Yukon GitHub App and defining a verifier and baseline metric, after which independent solvers submit against the gate using their own models, harnesses, and prompts. Active challenges span ZK proof acceleration (SNARK.fast), Apple silicon model inference (MLX.fast), and open computer science problems, with partners including Stanford, UC Berkeley, Princeton, Ethereum, Poolside, and Lighter Exchange.
Coinbase deployed a new matching engine on Deribit capable of processing over 100,000 orders per second with sub-millisecond latency, bringing the options exchange onto the same core execution infrastructure as Coinbase International Exchange. The engine introduces brief delays for aggressive orders on select instruments, giving liquidity providers additional time to reprice quotes during fast-moving markets. Coinbase plans to migrate all perpetual futures contracts to the engine before year-end, consolidating its derivatives stack across venues.
Figure posted 95% adjusted net revenue growth and 55% adjusted EBITDA margin in Q2. Figure functions less like a lender and more like a marketplace earning infrastructure-style fees. Its Figure Connect product (launched June 2024) now handles 65% of consumer loan marketplace volume across a 489-partner network of mortgage banks, depositories, and fintechs, and partner originators do 2.6x more HELOC volume than their pre-Figure baseline.
Crypto payment cards processed $759 million in July, up roughly 2.5x year over year from under $1 million a month in late 2023, per a16z crypto and Paymentscan. This guide lays out five architecture decisions for teams building stablecoin card programs. Cards require coordinating an issuing bank or BIN sponsor, card network, compliance/KYC providers, and wallet infrastructure rather than a single API integration. Teams must choose between prefunded balances, just-in-time funding pulled from wallets or DeFi vaults, or secured credit (which in the US can require collateral locked at least 14 days). Compliance obligations under frameworks like Regulation E shape who can get a card, what onboarding looks like, and where it can be used. Programmable wallets allow authorization rules to be enforced at both card and onchain levels, useful for role-based business spending or AI agent budgets, and cards should be built on account infrastructure that can extend into funding, payments, and yield rather than as a standalone product.
Pump.fun's launchpad market share dropped from 79.4% in late June to 26.7% in mid-July as Robinhood Chain competitors launched, then recovered to 51.7% by early August. Absolute fee revenue grew 30% between the two 30-day windows, from $24.45M to $31.83M, though the broader launchpad category expanded 77% to $75.39M, meaning Pump.fun underperformed category growth on a relative basis. On Robinhood Chain, Pons generated $19.80M in fees within 30 days to rank second across all launchpads. NOXA peaked at $2.33M in daily fees before disabling new launches due to bot activity, while Uniswap Pools' zero-fee model drew only $806K total despite launching August 5. The volume surge has not translated into memecoin appreciation: the sector sits at $25.15B, 83% below its December 2024 peak of $150.6B, with only 34% of 5,774 tracked tokens gaining value over 30 days and the median token down 99% from its all-time high.
After IBM CEO Arvind Krishna warned on CNBC that quantum computers could crack Bitcoin's cryptographic protections within three to four years, Jim Cramer announced he was selling his Bitcoin holdings and suggested Ethereum faces even greater exposure. Bitcoin rose approximately 1.6% the day of his announcement, consistent with his historical track record: his December 2022 sale near $16,800 preceded a 400%-plus gain over the following three years. The quantum risk framing has empirical grounding, as AI-assisted exploits including the Coldcard vulnerability have produced over $100 million in combined losses across recent incidents. Protocols including Zcash are responding by deploying AI to audit their own cryptographic implementations for weaknesses before adversaries can exploit them.
Bitcoin miner transaction fee revenue has fallen to 0.69% of total income, a 10-year low.
ERC-6551 (Token Bound Accounts) assigns each NFT its own smart contract wallet, enabling NFTs to hold tokens, other NFTs, and interact with DeFi protocols.
Investors undervalue DeFi apps because they're actually targeting the $500 trillion global asset market rather than the $2 trillion crypto market, and current fee generation barely scratches their real pricing power.
Standard Chartered analyst Geoff Kendrick projects Chainlink reaching $200 by 2030, about 25x its current price near $8.25, via year-by-year steps through $13 (2026), $41 (2027), $82 (2028), and $133 (2029).
Instagram has introduced a redesigned wordmark for the first time since 2016, giving its name a sharper, more modern look while keeping references to the original style. The new cursive-inspired lettering has received mixed reactions, with some praising the fresher design and others joking that the stylized “r” makes it look like “Instagzam.” Instagram has not said whether the wordmark update will be followed by broader visual changes, such as a new app icon.
Google has unveiled its newest Pixel lineup, including Pixel 11 phones, Pixel Watch 5, and its first-ever Pixel Tag finder device. The Pixel 11, 11 Pro, 11 Pro XL, and Pro Fold are built for Gemini Intelligence, with greater durability and upgraded cameras. Pixel Watch 5 adds proactive Gemini features, strength coaching, improved GPS tracking, and Health Guardian tools like breathing emergency detection.
A new leak suggests Apple's upcoming foldable iPhone Ultra may place its two front-facing cameras in the corners of the displays rather than the traditional centered position, with one camera on the outer screen and another on the inner screen. If accurate, this would mark a significant departure from Apple's usual iPhone design and could require users to adjust how they take selfies.
Figma's Data Science team ran a randomized controlled trial with 100 participants to measure how much time Figma Make actually saves on design tasks. Design work was found to be 20% faster and 16% easier overall, with product managers seeing the largest gains—23% faster and 37% easier—though designers benefited most on complex tasks while PMs gained more on simpler ones. Figma plans further RCTs, including one testing its new design agent's parallel-agent capabilities.
AI governance frameworks bring documentation, traceability, and accountability to AI features embedded in design tools. Building a framework involves defining scope and owners, classifying design data, standardizing vendor evaluation, and embedding controls into workflows and CI/CD processes. Penpot positions itself as a governance-friendly alternative through self-hosting, open file formats, and its MCP server, letting organizations control AI models and data flows on their own infrastructure.
Large language models are not neutral and often reflect or amplify biases related to language, geography, gender, culture, and socioeconomic status because of the data they are trained on and the feedback systems used to refine them. As a result, AI-generated personas, research summaries, copy, and recommendations can overrepresent dominant groups and present their perspectives as consensus while overlooking minority viewpoints. Designers should treat AI outputs as inputs for further thinking rather than objective conclusions, and actively review them for hidden assumptions, missing perspectives, and demographic bias.
ShootClip is a complete video editor for macOS — cut, grade, mix, and export with zero learning curve. AI speech-to-text turns every voice into spot-on captions, and the built-in MCP server lets Claude or any AI assistant edit right alongside you.
Remix lets product and design teams create live variants of the existing app, test them with users, and send the best version to engineering for review.
Agents generate more changes than teams can review. Argos makes every change obvious, then lets your team and your agents comment on the diffs, request updates, or approve before merge.
Nine common persona mistakes undermine product design, from over-relying on AI to using outdated, generic, or ignored personas. Personas should be research-based, focused, specific, updated per project, and actively shared across teams to be effective. Avoiding these pitfalls helps teams build products that better match customer needs, improving satisfaction and business outcomes.
PRDs still suit large teams aligning around a known target, but they assume the destination is already chosen, which doesn't fit early exploration. A lighter six-line alternative—Intent, Bet, Bar, Scope, Loop, Never—distills into one sentence short enough to hand directly to an AI agent. Unlike lengthy PRDs models can read but which lock in premature decisions, this compressed brief keeps both humans and agents flexible while testing risky assumptions.
"Slop" has expanded from a label for AI-generated filler into a broader cultural judgment about attention, trust, and creative authenticity.
Paris-based illustrator Julie Jonquet-Caunes transforms overlooked everyday objects into meticulous, nostalgic compositions—blending hyperrealist drawing, collage, and art-historical influences to reveal the stories, emotions, and beauty hidden in ordinary details.
Apple is facing criticism and regulatory scrutiny in Italy over a billboard showing a toddler using an iPhone.
Sponsored
Plug CompactifAI API into your coding agent and cut your bill drastically, without giving up top-tier AI quality. Now with a free first month for new sign-ups. The best API for output-heavy workloads: * GLM 5.2 AT $3.50 PER 1M OUTPUT TOKENS: same intelligence tier as Sonnet 5, at 65% lower cost. * DROP-IN FOR THE DEV TOOLS YOU ALREADY USE like n8n, LiteLLM and Cursor: no migration, no workflow change. * TRY OUR UNCENSORED VERSIONS of GLM 5.1 and Qwen 3.6 27B, ready to answer fully on any topic without giving up accuracy. * FIRST MONTH FREE for new sign-ups. No commitment, cancel anytime.
A range of AI models, including DeepSeek, Qwen, and Kimi, were tested for generating web pages, showing large variations in performance and credit costs. The best-performing models for a static coffee shop website were DeepSeek V4 Flash, which consumed only 2.4 credits on average, and GPT 5.6 Sol in low effort mode, which averaged 141 credits. In contrast, higher-tier models like Claude Opus 5 showed a lot of credit usage with results that didn't necessarily justify their cost.