Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Illustrator Azra Hirji creates colorful, food-filled scenes inspired by family, culture, migration, and shared meals, turning personal memories and global influences into a distinctive visual style.
Sponsored
Buildkite, the pro-tool that hands you building blocks for CI orchestration and then gets out of the way. Enough with constant downtime. No more band-aids or burden. We provide programmable primitives you co-ordinate, running only what changes, across any cloud, any chip, your secrets, your infra. Control without compromise. Build faster, more securely, more reliably, with the CI engineered for frontier scale, used by the teams setting the pace.
Cloudflare prototyped Cache Transcoding, a system that compresses eligible HTML, JSON, CSS, and JavaScript with Zstandard before storing and transferring them between cache tiers. Initial tests shrank these assets to roughly one-third of their original size while adding only a few percent of CPU overhead, potentially creating petabytes of effective cache capacity without deploying more storage.
Shortening individual tool outputs can make an AI coding agent more expensive when missing context forces it to reopen files, rerun commands, or take extra turns. GitHub reduced Copilot costs without materially hurting quality by selectively compressing repetitive logs, preserving source-like output, removing unused formatting such as line-number prefixes, shortening prompts carefully, and eliminating unnecessary retrieval steps.
Cline replaced its VS Code extension's roughly 76,000-line monolithic agent core with a shared SDK while protecting workflows used by more than 11 million developers. Since the VS Code Marketplace lacks gradual rollouts, the team bundled legacy and new engines together behind a feature flag with automatic fallback, then ran a month-long A/B rollout that reduced tasks hitting three consecutive agent mistakes from 6.34% to 0.62%.
Most known Val Town Pro referrals now come from AI, forcing the company to think about Answer Engine Optimization as traditional search traffic declines. Its early strategy is to measure model referrals and citations, selectively allow useful crawlers, and publish genuinely helpful material for both humans and agents without filling the web with AI-generated content.
AI is exposing the limits of mandatory code review by generating more code than humans can realistically inspect, turning senior engineers into a bottleneck. Teams should move design, knowledge sharing, testing, and architectural feedback earlier through pairing, collaborative design, automation, and fitness functions, then reserve human code review for exceptional changes involving security boundaries, major architectural decisions, large blast radii, or low confidence.
Sponsored
What happens when you give your agent an access token and it gets hijacked? Relay [11] keeps the credential at WorkOS, attaches it only when needed, and releases it only to allowlisted hosts. A hijacked agent session becomes a live process you can kill. You can't leak
Gemini 3.8 Flash improves coding, long-running agent tasks, and specialized reasoning while retaining 3.7 Flash's introductory price of $0.75 per million input tokens and $3.75 per million output tokens. Google also introduced 3.8 Flash Cyber for trusted defenders, reporting a success rate above 70% on an internal vulnerability-discovery benchmark spanning 20 programming languages and near-frontier performance on automated patching.
Muse Spark 1.3 is Meta's updated model for long-running agentic and coding workflows. It better tracks prior work, handles messy or conflicting context, asks for clarification when needed, and uses roughly 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2 in Meta's internal coding comparisons.
Uppy 6.0 rewrites its AWS S3 plugin around three explicit signing modes and adds direct support for S3-compatible services like Cloudflare R2, MinIO, and DigitalOcean Spaces. It also folds four frequently duplicated packages into Uppy Core, improves multipart-upload reliability and recovery, and tightens supply-chain protections.
Trellner queried two Perplexity models across 380 software categories and found that 59.8% of their 7,534 citations came from domains outside the 100,000 most-visited websites. Three apparently connected sites alone had published 215,128 machine-generated recommendation pages, highlighting how AI retrieval systems can ground purchase recommendations in content designed primarily for machines.
Wasmi 2.0 runs about 2.2 times faster than Wasmi 1.0 across its benchmark suite while preserving its strong startup performance. The Rust-based WebAssembly interpreter gains four instruction-dispatch modes, a fixed 64-bit cell representation, smaller binary options, stable fuel metering, and support for WebAssembly's deterministic profile.
Sponsored
Artificial Analysis and Baseten on Terminal-Bench 3.0 going 4.6% to 28.3%, and what it means for your workflows.
Claude Fable 5.1 agent swarms autonomously researched, modeled, and quality-checked browser-native 3D reconstructions of real places, shipping each world as a plain Three.js application.
Most users can satisfy the majority of their needs with models costing a fraction of the frontier options.
Cursor's Self-Hosted Machines let cloud agents execute tools against internal services, custom hardware, and private repositories on dynamically scheduled infrastructure you manage, while planning and inference remain in Cursor's cloud.
Tomasz Tunguz found that AI did not reduce his median 136 line-level edits per article, but it raised his writing's quality floor.
Sponsored
Train, test, and deploy the only configurable AI agent system built to ensure policy compliance in financial services: Fin [5]. Fin subjects every answer to rigorous, multi-stage validation so it can resolve complex queries with precision. As Lee Burkhill, PM at MonyGroup, says [6]: _"I've never seen Fin hallucinate—not once. That's a huge deal. We constantly test it, throw new questions at it, analyze variations and it holds up. That level of accuracy in an AI Agent is rare."_ Learn more about Fin [7], the AI Agent making perfect customer experiences possible in 45+ languages.
Kalshi signed a multi-year deal with the USTA to become the exclusive prediction market partner of the US Open. Branding, live market data integrations, fan experiences, and hospitality will begin with the 2026 tournament. The partnership also includes an integrity framework with the USTA and International Tennis Integrity Agency, including data sharing and restrictions on markets tied to umpire decisions, injuries, and code violations.
Hyperliquid Labs is reportedly in advanced talks with Kraken parent Payward to offer certain Hyperliquid-linked perpetual futures to US traders through Bitnomial, the CFTC-regulated derivatives platform Payward agreed to acquire earlier this year. The potential structure would give Hyperliquid a regulated path into the US market, but the talks remain unconfirmed and subject to regulatory approval, while speculation around a possible HIP-3 implementation has not been verified.
A2A transactions are projected to surpass 5.8 trillion by 2028, a 160% increase from 2024, making them increasingly attractive to fraudsters. Visa's A2A Protect uses AI and sophisticated transfer learning to give banks immediate access to critical global risk insights on transactions, without waiting months for models to develop intelligence from a bank's own transaction data.
Ben Rothenberg argues that the USTA's new Kalshi partnership accelerates an unhealthy fusion of tennis and gambling, pointing to massive trading volumes on even low-level matches and growing harassment of players by bettors. He also highlights the double standard between tournaments taking betting sponsorships while players face strict gambling rules, arguing that commercial incentives are outrunning concerns about integrity and player welfare
Stablecoins moved $33 trillion in 2025. Only about $390 billion of stablecoin activity that year went to real-economy payments for goods and services, roughly 1% of the total. The industry has largely solved how to create and move digital dollars. What it has not solved is what happens next: when someone wants to use that balance to pay a supplier, send money home, or buy something without converting back into fiat first.
Airwallex launched T:0 out of private beta, making its AI-native accounting platform available to startups across the US. The platform automates bookkeeping and revenue recognition, keeps financials continuously updated, and provides real-time visibility into metrics such as cash, burn, and runway while supporting growing complexity across entities, revenue models, and geographies.
Block is making its proprietary Cash App Score available to outside lenders for the first time through Nova Credit's Cash Flow Intelligence Platform, using real-time Cash App activity such as spending, saving, repayments, deposits, and peer-to-peer transactions to supplement traditional credit data. Block says the score approves 38% more Cash App Borrow customers at the same loss rate as traditional scores and could help lenders expand approvals across credit cards, auto loans, personal lending, device financing, and tenant screening without requiring consumers to link new accounts.
A consortium of 21 financial institutions including Goldman Sachs, Bank of America, Citi, and Deutsche Bank plans to form a company this year and launch a dollar-backed stablecoin in the first half of 2027, with euro and other G7 currency tokens potentially following. The initiative would put major banks in direct competition with stablecoin incumbents like Tether and with rival banking consortium Qivalis, though existing bank-issued stablecoins have so far seen limited adoption.
Morgan Stanley, Citi, and Goldman Sachs are pressing major law firms to share the efficiency gains from AI through lower fees, competitive bidding, fixed-fee arrangements, and other alternatives to traditional hourly billing. The shift threatens Big Law's leverage model, which depends on large teams of junior lawyers billing long hours, and could accelerate smaller associate classes, more in-house legal work, and a broader move toward pricing based on outcomes rather than time.
The one-click checkout startup is seeking a financial lifeline after its valuation collapsed 97% from $11 billion to $300 million, with Ryan Breslow personally committing $5 million. Existing investors who decline to participate risk losing much of their equity, suggesting the company may be running short on options despite claims it is nearing profitability. Breslow is betting that a 60-person, AI-powered team and Bolt's new financial super app can return the company to growth.
The Monetary Authority of Singapore will invest US$173 million over three years to accelerate the development and adoption of technologies including AI, blockchain, quantum computing, and digital assets. The program will also fund shared infrastructure, attract global financial institutions, and support at least 1,000 fintech internships as Singapore builds on an ecosystem of more than 1,800 firms.
Money movement platform TabaPay is set to buy a Denver-based bank after securing $155 million in strategic growth financing led by FTV Capital.
Orange has become the first merchant in France to offer subscription bill payment with Wero.
Félix, a chat-based platform that utilises WhatsApp to enable Latino immigrants in the US to send money abroad, is planning to add a host of other financial products after raising $87 million in a Series C equity round.
US unsecured personal loan balances has reached a record $281 billion, up 9.6% year over year.
Sponsored
Deploying AI agents without fixing your underlying telemetry data leads to lackluster results. Google Cloud commissioned an Omdia study of 300 security professionals, and broke down why data readiness is the #1 hurdle to agentic AI in the SOC: * THE DATA GAP: Only 48% of enterprise telemetry data is considered "AI-ready." * REARCHITECTURE NEEDED: 82% of respondents report that moderate or complete data redesign is required to support multi-agent workflows. * PRACTITIONER DIVIDE: Security engineers and architects are 2x more likely than SOC managers to call for a full telemetry overhaul.
Attackers hijacked Softaculous' 162.55.80.0/24 prefix and redirected Virtualizor update traffic to malicious servers. Hetzner's RPKI settings accepted /24 routes, while a forged AS path matched its ROA, so the attackers also obtained TLS certificates during the diversion. Virtualizor did not cryptographically verify update packages, and Softaculous cannot identify every affected server and urges customers to investigate.
Gambling Goblin installed malicious Apache modules on Brazilian government and education servers. The modules reverse-proxy visitors to fake Google Play, Microsoft Store, and Amazon pages promoting betting sites. Operators deployed DownPro, AlphaAgent, oRAT, 3snake, SSH brute-forcing, and reconnaissance tools. Administrators lack published module hashes, filenames, or paths for direct Apache checks.
A security researcher discovered two independent flaws in Unitree G1 humanoid robots that could enable RCE on the devices. The first vulnerability allows an attacker within BLE range to obtain an AES-128 key without requiring the device to pair. The researcher found that the Unitree cloud API would decrypt the key without verifying ownership or serial numbers, allowing the researcher to access protected BLE operations, eventually allowing the attacker to exploit a buffer overflow to achieve RCE. The second path involves a path-traversal vulnerability in the robot's AI service, which enables an attacker to upload a file to the bashrunner's whitelist directory, enabling RCE.
The npm worm lineage has evolved to a point where standard supply chain provenance mechanisms can no longer guarantee safety. During the August 2026 ChainDrop campaign, threat actors compromised over 400 packages in under four hours by hijacking a maintainer's legitimate signed release pipeline. The attackers utilized authorized infrastructure, so every poisoned package version successfully carried a perfectly valid SLSA attestation. To further evade traditional domain blocklists, the campaign hosted its command and control operations within an Ethereum smart contract. This attack builds upon the previous Shai-Hulud worms, which have recently pivoted to targeting authentication tokens for AI development tools like Claude, Codex, Cursor, and Gemini. Security researchers warn that these identical mechanics could easily compromise infrastructure-as-code pipelines, proving that current attestation models verify origin rather than actual code legitimacy. To defend against these pipeline hijackings, organizations must enforce strict version pinning, utilize curated registries, implement OIDC-scoped credentials, and apply aggressive network egress restrictions.
Since 2024, BREEZE COMET has targeted Brazilian banks, payment firms, retailers, exchanges, and fintech providers to make fraudulent Pix, STR, and Boleto transfers. Initial access included password spraying, fake IT-support calls, booby-trapped tax files, exposed JBoss servers, and rogue devices on retail networks. The group steals mTLS credentials, API keys, CI/CD secrets, and cloud tokens, then moves through SMB and RDP. Its toolkit includes REALBREEZE, COBALTSPIN, LIGHTPAINT, MILDFROST, KICKPLATE, and BOATBEAM. Mandiant recorded two transaction waves within 24-48 hours after access to core financial applications.
Claude-BugHunter packages 83 auto-loading Claude Code skills distilled from 681 disclosed HackerOne reports, so the agent loads per-class detection patterns, payload tables, and chain templates by topic rather than by name.
Anthropic will roll out Enterprise Frontier Safeguards in phases this fall. Customer activity logs can remain in customer-controlled AWS, Azure, or Google Cloud storage, using customer-managed keys and access policies. Automated systems inspect rolling traffic windows for offensive cyber capability, biological misuse, and stolen credentials. Flags go to customer teams. Anthropic staff does not review them.
OpenAI classifies Astra as Critical for cybersecurity capability. Tests found zero-days, browser sandbox escapes, and root privilege-escalation chains. OpenAI delayed release work after the Hugging Face incident. Astra refused 91.5% of cyber-jailbreak requests, versus 59% for GPT-5.6 Sol. Advanced cyber access starts with selected testers, then Daybreak Blue.
CrowdStrike reports that 88% of observed vulnerabilities were actively exploited within 48 hours of a public proof-of-concept release between January and June. Advanced threat groups like VAULT PANDA and GENESIS PANDA weaponized flaws in under 24 hours, while UMBRAL BISON targeted Linux privilege escalation vulnerability CVE-2026-31431 roughly 20 hours after disclosure. The report also highlights a rapid shift in alternative attack vectors, noting a fifteenfold increase in device code phishing alongside a massive concentration of software registry threats targeting npm.
A dark web service named Nexus recently advertised over 153 million scanned US and Canadian driver's licenses alongside millions of other identification documents on a Russian cybercrime forum. By correlating GMT timestamps embedded in the leaked image filenames with real-world travel records, researchers traced the infrared and ultraviolet scans back to a potential breach at the identity verification vendor idscan.net. Following the discovery of this continuous exfiltration campaign affecting major corporate clients, the FBI opened an official inquiry, and the Nexus site immediately went offline.
CVE-2026-0768, a 9.8-severity remote-code-execution flaw in Langflow, is under active exploitation.