Like a good pair of Basquiat Crocs, there are innumerable bad ways to license an artist’s work. So when Airstream looked to partner up on a project with the Frank Lloyd Wright Foundation, the aluminum-clad trailer brand could have just printed one of the architect’s famous patterns on a limited run of its vehicles and called it a day. It probably would have even sold well. But that is decidedly what Bob Wheeler, Airstream’s president and CEO, did not want to do.  “We said, ‘All right, let’s make sure that everything has a purpose and a function—that way it’s not just a pastiche, or some kind of lame attempt to mimic something,’” Wheeler recalls. “We didn’t want it to seem overdone or kitschy.” Instead, the brand embarked on a multiyear collaboration with the experts at Wright’s Taliesin West home and studio in Scottsdale, Arizona, and today the two are rolling out the 28-foot Airstream Frank Lloyd Wright Usonian Limited Edition Travel Trailer. With just 200 numbered vehicles that retail for $184,900 on offer, you—like me—might not be able to afford one at the moment, but they just might also restore your faith in the art of the artist collab at large.  [Photo: Airstream] BETTER LATE THAN NEVER Wheeler has a passion for midcentury design (as you might expect of Airstream’s CEO), so it tracks that he’d be a natural fan of Wright’s organic architecture. “Honestly, this has been a dream of mine for the last 20 years, which is about as long as I’ve been president of Airstream,” he says. “Why are Wright’s designs so celebrated today? It’s because they’re timeless. I think there are values there that incentivize someone to buy an Airstream that overlap in some meaningful ways.” Though Wright and Airstream founder Wally Byam were active at the same time and likely shared some of the same design fan base, there’s no record of them ever meeting. But a collaboration between the two ultimately proved inevitable when Wheeler reached out to Wright’s foundation in 2022. Foundation historian Sally Russell says her team wasn’t initially sure how robust a joint project could be. They eventually toured the Airstream factory in Ohio where the trailers are handmade using 3,000 rivets over the course of 350 hours, and saw how much customization was truly possible. Then she realized that it could be a great showcase of Wright’s work.  Beyond an Airstream’s signature aluminum exterior, Wheeler says the trailer is essentially a blank canvas. “And that’s where we can really flex some design muscle and allow others to do so.”  Russell says the foundation first explored whether to make the trailer feel like an adaptation of a specific Frank Lloyd Wright home. “The answer to that was no,” she says. “We didn’t want to try to re-create the Rosenbaum House and shove it into the size of a trailer. It didn’t make sense, because Frank Lloyd Wright certainly designed for each of his individual projects—he created something new, something that expressed the individual forms of the project, the needs of the client. So there was a great awareness of wanting to continue that legacy through the work that we did on the trailer.” The two teams ultimately homed in on the concept of Usonian design, a style that aimed to democratize design via small, affordable homes with a focus on efficient floor plans, functionality, and modularity.  In other words: an ideal fit for an Airstream. [Photo: Airstream] COLLAPSIBLE CHAIRS AND CLERESTORY WINDOWS When you approach the trailer, the connection to Wright is immediate on the custom front door featuring the Gordon leaf pattern, which the architect commissioned his apprentice Eugene Masselink to design in 1956. It’s a tip of the hat to nature, presumably an Airstreamer’s destination, and can be found subtly throughout the trailer in elements like sconces and cabinet pulls—but not too much, per the design mission at the outset. (“At one point we had a lot more of that Gordon leaf in there,” Wheeler notes. “We dialed that way back.”) With the push of a button, the bench seating converts into a king-size bed—one of Wheeler’s favorite elements. It is the largest bed in any Airstream, and is a first for the company, he says.  [Photo: Airstream] Another convertible element, in line with that focus on modularity, is the living space at the front of the trailer. Here, a dining table, desk, and seating inspired by the slant-back chairs that Wright used throughout his career collapse into a wall cabinet. Wheeler says Airstream used to deploy clever features like this in the midcentury era, before modern preferences trended toward built-in furniture. “So in some ways, this is a bit of a flashback to an earlier design in the ’50s, which is appropriate.” The teams also honored Wright’s focus on natural light, relocating Airstream’s usual overhead storage in favor of clerestory windows, which are prominent in Usonian homes. Meanwhile, the overall color palette comes from a 1955 Wright-curated Martin-Senour paint line. Russell says the team selected it for its harmonious blend with the natural settings where the trailer is likely headed, featuring ocher, red, and turquoise.  Ultimately, “It’s like a Frank Lloyd Wright home, where you walk into it, and it’s a completely different experience from any other building,” Russell says. “I hope that he would be very happy to see that design legacy continue, because he certainly did that with his own fellowship and the apprentices that he worked with.” [Photo: Airstream] USONIAN LIFE Starting today, the limited-edition, numbered trailers will be available for order at Airstream dealerships. Wheeler says the company was originally going to release just 100 of them, but got so much positive feedback from dealers and others that they doubled the run.  On the whole, the collaboration comes in the wake of a boom time for Airstream, which is owned by Thor Industries. Airstream experienced a surge during the pandemic, resulting in a 22% jump in sales in 2021 as people embraced remote work or realigned their relationship to the world.  “We’ve come back to earth now, and now we’re much more tied to actual market retail rates, which is what we know,” Wheeler says. In its third-quarter financials, Thor reported $2.89 billion in revenue (up 3.3% from previous year). While the company declined to provide Airstream-specific numbers, its overall North American towable RV division is up 9.1% from the same period in 2024. But there’s a problem afoot: The current administration’s tariffs, which Wheeler says made settling on the price for the Frank Lloyd Wright collaboration tricky. He adds that the company is struggling with shortages caused by the disruption in the supply chain, and high interest rates are also a problem.  [Photo: Airstream] “Look, we’re 94 years old,” he says. “We’ve been through more of these cycles than we can count, so we’re fine, and we’ll continue to trade on authenticity, quality, great service and support, a great dealer network, and a brand that really has become part of the fabric of the U.S. traveling adventure.”

Whether Apple calls it iOS 19 or iOS 26, these are some of the things we hope the tech giant brings to the iPhone next.

If you're familiar with the basics of Linux and ready for a new approach, NixOS merges thoughtful design with rich customization.

Google loads its Pixel phones with the most essential apps, but there are some obscure gems worth adding, too.

Wu-Tang Clan has a new game. At Summer Game Fest 2025, Brass Lion Entertainment has introduced its debut game, Wu-Tang: Rise of the Deceiver. In it, you'll have to fight alongside the group's members to defeat the invading forces of the Decei

We expect to see new designs, new branding, and more at Apple's WWDC 2025.

The administration has ended free income tax filing, but that didn't stop the IRS from open-sourcing its frontend Direct File tax program.

One of the key use cases for generative AI involves answering questions over private datasets, with retrieval-augmented generation (RAG) as the go-to framework. As new RAG techniques emerge, there’s a growing need to benchmark their performance across diverse datasets and metrics.  To meet this need, we’re introducing BenchmarkQED, a new suite of tools that automates RAG benchmarking at scale, available on GitHub (opens in new tab). It includes components for query generation, evaluation, and dataset preparation, each designed to support rigorous, reproducible testing.   BenchmarkQED complements the RAG methods in our open-source GraphRAG library, enabling users to run a GraphRAG-style evaluation across models, metrics, and datasets. GraphRAG uses a large language model (LLM) to generate and summarize entity-based knowledge graphs, producing more comprehensive and diverse answers than standard RAG for large-scale tasks.  In this post, we walk through the core components of BenchmarkQED that contribute to the overall benchmarking process. We also share some of the latest benchmark results comparing our LazyGraphRAG system to competing methods, including a vector-based RAG with a 1M-token context window, where the leading LazyGraphRAG configuration showed significant win rates across all combinations of quality metrics and query classes. In the paper, we distinguish between local queries, where answers are found in a small number of text regions, and sometimes even a single region, and global queries, which require reasoning over large portions of or even the entire dataset.  Conventional vector-based RAG excels at local queries because the regions containing the answer to the query resemble the query itself and can be retrieved as the nearest neighbor in the vector space of text embeddings. However, it struggles with global questions, such as, “What are the main themes of the dataset?” which require understanding dataset qualities not explicitly stated in the text.   AutoQ: Automated query synthesis This limitation motivated the development of GraphRAG a system designed to answer global queries. GraphRAG’s evaluation requirements subsequently led to the creation of AutoQ, a method for synthesizing these global queries for any dataset. AutoQ extends this approach by generating synthetic queries across the spectrum of queries, from local to global. It defines four distinct classes based on the source and scope of the query (Figure 1, top) forming a logical progression along the spectrum (Figure 1, bottom). Figure 1. Construction of a 2×2 design space for synthetic query generation with AutoQ, showing how the four resulting query classes map onto the local-global query spectrum.  AutoQ can be configured to generate any number and distribution of synthetic queries along these classes, enabling consistent benchmarking across datasets without requiring user customization. Figure 2 shows the synthesis process and sample queries from each class, using an AP News dataset. Figure 2. Synthesis process and example query for each of the four AutoQ query classes.  About Microsoft Research Advancing science and technology to benefit humanity View our story Opens in a new tab AutoE: Automated evaluation framework  Our evaluation of GraphRAG focused on analyzing key qualities of answers to global questions. The following qualities were used for the current evaluation: Comprehensiveness: Does the answer address all relevant aspects of the question?  Diversity: Does it present varied perspectives or insights?  Empowerment: Does it help the reader understand and make informed judgments?  Relevance: Does it address what the question is specifically asking?   The AutoE component scales evaluation of these qualities using the LLM-as-a-Judge method. It presents pairs of answers to an LLM, along with the query and target metric, in counterbalanced order. The model determines whether the first answer wins, loses, or ties with the second. Over a set of queries, whether from AutoQ or elsewhere, this produces win rates between competing methods. When ground truth is available, AutoE can also score answers on correctness, completeness, and related metrics. An illustrative evaluation is shown in Figure 3. Using a dataset of 1,397 AP News articles on health and healthcare, AutoQ generated 50 queries per class (200 total). AutoE then compared LazyGraphRAG to a competing RAG method, running six trials per query across four metrics, using GPT-4.1 as a judge. These trial-level results were aggregated using metric-based win rates, where each trial is scored 1 for a win, 0.5 for a tie, and 0 for a loss, and then averaged to calculate the overall win rate for each RAG method. Figure 3. Win rates of four LazyGraphRAG (LGR) configurations across methods, broken down by the AutoQ query class and averaged across AutoE’s four metrics: comprehensiveness, diversity, empowerment, and relevance. LazyGraphRAG outperforms comparison conditions where the bar is above 50%. The four LazyGraphRAG conditions (LGR_b200_c200, LGR_b50_c200, LGR_b50_c600, LGR_b200_c200_mini) differ by query budget (b50, b200) and chunk size (c200, c600). All used GPT-4o mini for relevance tests and GPT-4o for query expansion (to five subqueries) and answer generation, except for LGR_b200_c200_mini, which used GPT-4o mini throughout. Comparison systems were GraphRAG (Local, Global, and Drift Search), Vector RAG with 8k- and 120k-token windows, and three published methods: LightRAG (opens in new tab), RAPTOR (opens in new tab), and TREX (opens in new tab). All methods were limited to the same 8k tokens for answer generation. GraphRAG Global Search used level 2 of the community hierarchy. LazyGraphRAG outperformed every comparison condition using the same generative model (GPT-4o), winning all 96 comparisons, with all but one reaching statistical significance. The best overall performance came from the larger budget, smaller chunk size configuration (LGR_b200_c200). For DataLocal queries, the smaller budget (LGR_b50_c200) performed slightly better, likely because fewer chunks were relevant. For ActivityLocal queries, the larger chunk size (LGR_b50_c600) had a slight edge, likely because longer chunks provide a more coherent context. Competing methods performed relatively better on the query classes for which they were designed: GraphRAG Global for global queries, Vector RAG for local queries, and GraphRAG Drift Search, which combines both strategies, posed the strongest challenge overall. Increasing Vector RAG’s context window from 8k to 120k tokens did not improve its performance compared to LazyGraphRAG. This raised the question of how LazyGraphRAG would perform against Vector RAG with 1-million token context window containing most of the dataset. Figure 4 shows the follow-up experiment comparing LazyGraphRAG to Vector RAG using GPT-4.1 that enabled this comparison. Even against the 1M-token window, LazyGraphRAG achieved higher win rates across all comparisons, failing to reach significance only for the relevance of answers to DataLocal queries. These queries tend to benefit most from Vector RAG’s ranking of directly relevant chunks, making it hard for LazyGraphRAG to generate answers that have greater relevance to the query, even though these answers may be dramatically more comprehensive, diverse, and empowering overall. Figure 4. Win rates of LazyGraphRAG (LGR) over Vector RAG across different context window sizes, broken down by the four AutoQ query classes and four AutoE metrics: comprehensiveness, diversity, empowerment, and relevance. Bars above 50% indicate that LazyGraphRAG outperformed the comparison condition.  AutoD: Automated data sampling and summarization Text datasets have an underlying topical structure, but the depth, breadth, and connectivity of that structure can vary widely. This variability makes it difficult to evaluate RAG systems consistently, as results may reflect the idiosyncrasies of the dataset rather than the system’s general capabilities. The AutoD component addresses this by sampling datasets to meet a target specification, defined by the number of topic clusters (breadth) and the number of samples per cluster (depth). This creates consistency across datasets, enabling more meaningful comparisons, as structurally aligned datasets lead to comparable AutoQ queries, which in turn support consistent AutoE evaluations. AutoD also includes tools for summarizing input or output datasets in a way that reflects their topical coverage. These summaries play an important role in the AutoQ query synthesis process, but they can also be used more broadly, such as in prompts where context space is limited. Since the release of the GraphRAG paper, we’ve received many requests to share the dataset of the Behind the Tech (opens in new tab) podcast transcripts we used in our evaluation. An updated version of this dataset is now available in the BenchmarkQED repository (opens in new tab), alongside the AP News dataset containing 1,397 health-related articles, licensed for open release.   We hope these datasets, together with the BenchmarkQED tools (opens in new tab), help accelerate benchmark-driven development of RAG systems and AI question-answering. We invite the community to try them on GitHub (opens in new tab).  Opens in a new tab

미스트랄 코드는 보안과 컴플라이언스를 중시하는 대기업 환경을 대상으로 한 AI 코딩 어시스턴트로, AI 모델, IDE 내 어시스턴트, 로컬 배포 옵션, 엔터프라이즈용 툴링을 하나의 패키지로 통합한 것이 특징이다. 기반 기술은 오픈소스 AI 코딩 도구인 ‘컨티뉴(Continue)’를 활용했다. 현재 젯브레인IDE와 VS코드용 비공개 베타 버전으로 이 기술을 체험할 수 있으며, 정식 출시는 곧 이뤄질 예정이

Ready to try something new? I've installed almost every browser over the years, and these are my go-to favorites.