Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Qwen Intelligence has launched three mobile agents for planning, cross-app execution, and rapid content creation, alongside open benchmarks for planning, real-device performance, and safety. Qwen reports top benchmark results and 90% end-to-end success for Mobile-Use.
Perplexity's SPACE platform tested VM isolation and network confinement using nine AI models, revealing no VM-host breaches across 108 trials. However, four models exploited network-policy vulnerabilities via DNS spoofing and IP-sharing, bypassing restrictions in 11 out of 54 partial-network trials. Post-remediation, none of the models succeeded in bypassing the updated security measures, highlighting the necessity for robust policy enforcement against shared infrastructure attacks, which were also found in eight of ten tested third-party platforms.
Robotics needs a universal post-training recipe similar to language models to achieve reliable autonomous deployment. Existing methods like imitation learning benefit from intuition rather than structured processes, and RL's instability hinders robotics. The development of EXPO-FT demonstrates a stable RL approach, emphasizing the need for standardized protocols, reward specifications, human feedback systems, and scalable processes to enhance robotics reliability.
Sponsored
Generic agents hunt for context as they work. Databricks Genie learns it automatically from knowledge already embedded across your data environment — so it can reason over live data, run new analyses and take action.
An AI agent's capability is largely set by its harness. Evolving the harness against a fixed evolve set is effective, but the harness memorizes the training tasks, and large in-distribution gains shrink or vanish out of distribution. RRSI keeps the harness edit space open and regularizes the search trajectory through it instead. Prompts, control flow, configuration, context management, tools, skills, memory, and sub-agents may all be modified. The constraints act on how the search moves, not on what the harness may contain.
Ember-1 is a new specialized model built on Kimi K3 that delivers Kimi K3's quality with 40% fewer tokens. It is now rolling out as a serving option alongside the base Kimi K3 model as a Research Preview release on Serverless. Fireworks Research is introducing research releases to give developers two-week serverless access to new research models. The models with the most demand will be made permanent.
tev1-4B-experimental is a Jev-like classifier fine-tuned on top of Qwen3.5 4B. It is available on Together serverless at $0.042 per million input tokens and $0 per million output tokens. Together AI has released the data recipe and a tutorial on how to fine-tune a custom model. Tev1 only cost $17 to train.
Forecasters dramatically underestimate AI progress on benchmarks. Predictions on AI adoption and diffusion often lean toward underestimation. There is not enough evidence to assess forecasters' track record on predicting macro-scale economic and societal impacts of AI.
Google says its planned Private AI Compute memory would let assistants recall context across devices without Google being able to read it. Data stays encrypted in cloud storage, keys stay on users' devices, and a protected enclave decrypts it only while answering a request.
Sponsored
Deploy a functional RAG pipeline on AWS: Elastic Cloud for retrieval, AWS Lambda for orchestration, fully provisioned in one Terraform
AI-era education should prioritize curiosity, project-based learning, and intellectual exploration over grades and credentials.
MentalHealthBench is an open benchmark created by OpenAI with more than 80 licensed mental health experts to evaluate AI responses across realistic mental health conversations.
Comfy Router, available on the Comfy Developer Platform, unifies access to various frontier media models with one API, allowing developers seamless model integration and provider flexibility for better pricing and availability.
Gemini is adding direct integrations with tools including Linear, Adobe, Webflow, Peloton, Experian, and SeatGeek.
BMO, CIBC, National Bank of Canada, RBC, Scotiabank, and TD jointly announced they're exploring a Canadian-dollar tokenized deposit system, with the first phase focused on moving tokenized deposits efficiently between institutions. The project follows Canada's OSFI clarifying that tokenized deposits are "not legally distinct from traditional deposits," and builds on March's Project Samara pilot testing blockchain-based government bond issuance.
The CFTC's Division of Market Oversight issued an advisory warning that prediction contracts betting on whether someone will "mention" specific words or attend an event carry heightened manipulation risk, and can only be listed in limited circumstances under the Commodity Exchange Act. The agency has already charged a former White House teleprompter operator and ex-Rep. George Santos over trades tied to advance knowledge of "mention markets" on Kalshi, and it's telling exchanges to assess whether the person controlling a contract's outcome faces independent obligations that deter manipulation.
Rain introduced real-time funding, letting cardholders connect a wallet and approve a spending limit once, with each purchase authorizing only the exact amount needed directly from the wallet instead of requiring users to pre-fund a separate collateral balance. The feature is in beta for card programs where Rain manages authorization, launching with USDC on Base and Arbitrum and USDT0 on Plasma.
Circle has launched a Bitcoin-backed USDC borrowing facility for Mint clients through Morpho's Arc market, wrapping deposited BTC as cirBTC to serve as collateral. The cirBTC/USDC market grew ~10x in five days, from $1.37 million on September 17 to $14.3 million in borrowed USDC against 287 cirBTC tokens. Galaxy and Keyrock supply the dollar liquidity, while Circle handles cirBTC issuance and wallet infrastructure, and Morpho governs loan terms. The structure gives institutional Bitcoin holders access to USDC without liquidating their BTC positions.
CME Group plans to list futures on Bitcoin Cash and Uniswap on October 19, pending regulatory approval, with standard and Micro contract sizes of 250/25 BCH and 10,000/1,000 UNI. The addition brings CME's single-asset crypto derivatives roster to 11 tokens, with Uniswap marking the first DEX governance token the exchange has listed. CME averaged 279,800 contracts per day in H1 2026, about $8.3 billion in daily notional, while altcoin-specific contracts generated more than $1 billion in notional for the year. The listings follow CME's June launch of a 24/7 crypto futures and options market, as institutional participants seek regulated venues for altcoin price risk management across a widening asset set.
The emergence of Flock as a practical post-quantum proof system for binary circuits removes Ethereum's requirement for circuit-friendly hashes like Poseidon, opening five protocol layers (consensus signatures, CL aggregation, state tree construction, EL signatures, and EL aggregation) to standard hash function candidates. Performance targets set the bar at proving 2^25 bytes per second with LeanVM and approximately 1 million hash calls per second, a throughput floor that filters the field. BLAKE3 leads on raw speed at 7 GB/s parallel and 1.1 cycles per byte on long messages but carries no indifferentiability proofs and a thin cryptanalytic record. SHA-3 holds 50+ published cryptanalytic papers and sponge-mode indifferentiability proofs but runs at only 0.2 GB/s natively. The research ranks SHA-3 and BLAKE2s as the lowest-risk selections, with SHA-2's length-extension property disqualifying it as a random-oracle substitute despite 45+ cryptanalytic papers.
Tether abandoned MiCA compliance in the EU over a rule that required 60% of reserves in commercial bank deposits. CEO Paolo Ardoino argued it would expose 400 million users to risk since EU deposit insurance caps out at €100,000 per account. Tether holds roughly 80% of its $183 billion in reserves in US Treasuries, and USDT was delisted from all EU exchanges on July 1, though Ardoino said Tether would reconsider "when MiCA becomes safer." The ECB and all 27 EU central banks have separately called to scrap the 60% bank-deposit rule.
Kalshi published a detailed rebuttal to wash trading allegations on its perpetuals platform, arguing that its flat-fee liquidity incentive programs mirror standard structures at CME, CBOE, and Nasdaq. The core defense turns on trade economics: analyzed trades showed aggressors captured roughly $98,000 in profit potential, a signal of genuine price discovery rather than coordinated self-dealing. Kalshi also cited 350,000+ lifetime perpetual traders, open interest doubling over the past 30 days, and a July 2026 fee holiday structure for Self-Clearing Members that caps rebates at exact fees paid per trade, blocking the net-negative outcomes that would characterize artificial volume.
BitMEX ended exchange operations as of 04:00 UTC on September 23.
Binance will migrate crypto assets from Funding Accounts to Spot Accounts in phases starting September 29 through January, after which Funding Accounts become "Stocks Accounts" dedicated solely to equities and stock options clearing.
Strategy acquired 950 BTC for $75.7M at an average of $79,670/coin between September 14-20, bringing total holdings to 846,000 BTC, a position exceeding 4% of bitcoin's 21M hard cap.
Cardano has been added to Coinbase's x402 SDK, making it the first non-EVM chain on the protocol and enabling apps and AI agents to settle API calls with ADA or Cardano Native Tokens over standard HTTP requests.
Sponsored
The official ClickHouse provider lets you create tables, load data and run SQL from your Airflow DAGs. It works with self-managed Airflow and Astronomer. Set up the provider [5], connect it to ClickHouse Cloud and try it with your own data.
CedarDB ported the original Doom's game logic and renderer almost entirely to SQL, running the game loop at its original 35 Hz and rendering at up to 60 FPS. The experiment shows that surprisingly complex real-time logic can work with SQL's set-based model, while database features like transactions, access control, and shared state make multiplayer particularly natural.
Apache Cassandra 5 loosens many of the constraints teams learned with 3.11, adding storage-attached indexing, unified compaction, native vector search, and built-in repair scheduling. Cassandra 6 goes further with ordered cluster metadata and cross-partition transactions, although partitioning and workload design remain fundamental.
Concurrence uses Databricks to run and govern clinical AI at an annualized scale of roughly 1.2 trillion input tokens, combining patient data, agent state, testing, evaluations, and access controls on one platform. Unity Gateway centralizes access to approved AI models and tools, helping the company track usage, enforce compliance, and safely scale both clinical and coding agents.
Horizon audited over 5,000 AI benchmark tasks across 20 datasets and confirmed 29 were broken, with problems including leaked answers, incomplete tests, gameable graders, and failing reference solutions. Many of these flaws actually inflated model scores, highlighting how benchmark results can overstate capabilities when the underlying tasks and graders are unreliable.
TypeSafe's Jev uses a purpose-built decision model for real-time AI guardrails, checking agent inputs, responses, and tool calls before they execute. In Arize's tests, it made the same decisions as GPT-5.4 nano while running 15-18x faster and costing 12-14x less per call.
Sponsored
Become an expert in data storytelling with the online Master of Translational Data Analytics from Ohio State's Translational Data Analytics Institute. This interdisciplinary program blends machine learning, user experience, data visualization, and design thinking to prepare you to uncover and present insights from data.
Jevflake brings TypeSafe's Jev decision model into Snowflake through SQL functions and dbt, letting teams classify, score, and ask structured questions about their data with probabilistic answers. It can cache results in dbt models and add tests or human-review queues, making AI classifications reusable parts of normal analytics workflows.
MotherDuck's new prompt_jev function brings Jev directly into SQL, letting users classify and score text at analytics scale without training a model or parsing LLM responses. In its 100,000-row benchmark, Jev achieved 89% accuracy in 40 seconds for $0.50, outperforming the tested LLMs on speed, cost, and accuracy.
dbt v2's Rust-based Fusion engine now includes a built-in DuckDB adapter, removing the need for a separate adapter installation. It adds DuckLake and Iceberg REST catalog support, writes queryable project metadata as Parquet, and performs SQL analysis that finds type or column problems before warehouse execution. The migration path starts by testing the opt-in v2 parser against existing projects.
Apache Fluss 1.0 introduces a stateless Rust HTTP Gateway for metadata, table operations, and batch writes, alongside native Python access and deeper Flink, Spark, Hudi, and Paimon integration. It adds row and log TTLs, predicate pushdown, coordinator high availability, resource protection, and cleanup features, moving the project from incubator-era capabilities toward production real-time lakehouse workloads.
High-throughput transaction systems benefit from co-design rather than copying a connection-pool pattern from a general-purpose database. TigerBeetle models linked transfers atomically, aggregates balances without scans, batches operations automatically while a request is in flight, and prefers time-ordered identifiers for fast idempotency checks. Its benchmark illustrates why batching and ordered IDs can matter more than adding clients.
Fixed-rate sampling often discards the rare traces needed in an incident, while ordinary tail sampling loses statistical weights. An adaptive tail sampler for the OpenTelemetry Collector buffers traces, applies ordered rules, and attaches W3C probability thresholds so downstream counts can be reweighted. It adapts by route and service, bounds memory under load, and is progressing upstream toward alpha.
Polars now exposes a query profiler for local open-source workloads.
Common Crawl has released open web graphs covering 245.8 million hosts and 133.2 million domains.
Adobe is bringing its redesigned Premiere video-editing app to Android, nearly a year after its iPhone debut, replacing Premiere Rush, which loses support September 30. Layouts adapt across devices, pairing footage and a multitrack timeline side by side on foldables while still running on candy-bar phones like the Pixel. The app is free with no login required, though Firefly-powered features like Generative Fill require paid generative credits once the monthly allotment runs out.
macOS 27 Golden Gate introduces a customizable Liquid Glass appearance slider that lets users adjust the system's transparency from highly clear to more opaque, changing the look of the Dock, widgets, sidebars, toolbars, and apps like Music and Podcasts with a single system-wide setting. The update gives users more control over the Mac's visual style while preserving the modern Liquid Glass design language.
Meta introduced Ray-Ban Meta Audio, camera-free AI glasses that support music, calls, speech translation, and its Muse assistant, offering an alternative to smart glasses that can record people without their consent. Removing the cameras makes the frames slimmer and lighter at 43 grams, with adjustable temple tips, swappable nose pads, and a battery life of up to 12 hours. Starting at $349, the glasses come in Clubmaster and Burbank styles with prescription options.
As more interface work turns AI-generated, research and design output shifts from human-facing deliverables toward curated context that guides AI, dubbed UX-context design. DESIGN.md, which Google Labs open-sourced in April 2026 from its Stitch tool, encodes a product's visual identity for AI tools in exact values and plain-language guidance. A hypothetical broader file, UX.md, would add research findings, interaction standards, a glossary, and user and world models as a continuously updated source.