 |
|
|
|||||
Over the past five years, advances in AI models data processing and reasoning capabilities have driven enterprise and industrial developers to pursue larger models and more ambitious benchmarks. Now, with agentic AI emerging as the successor to generative AI, demand for smarter, more nuanced agents is growing. Yet too often smart AI is measured by model size or the volume of its training data. Data analytics and artificial intelligence company Databricks argues that todays AI arms race misses a crucial point: In production, what matters most is not what a model knows, but how it performs when stakeholders rely on it. Jonathan Frankle, chief AI scientist at Databricks, emphasizes that real-world trust and return on investment come from how AI models behave in production, not from how much information they contain. Unlike traditional software, AI models generate probabilistic outputs rather than deterministic ones. The only thing you can measure about an AI system is how it behaves. You cant look inside it. Theres no equivalent to source code, Frankle tells Fast Company. He contends that while public benchmarks are useful for gauging general capability, enterprises often over-index on them. What matters far more, he says, is rigorous evaluation on business-specific data to measure quality, refine outputs, and guide reinforcement learning strategies. Today, people often deploy agents by writing a prompt, trying a couple of inputs, checking their vibes, and deploying. We would never do that in softwareand we shouldnt do it in AI, either, he says. Frankle explains that for AI agents, evaluations replace many traditional engineering artifacts, i.e., the discussion, the design document, the unit tests, and the integration tests. Theres no equivalent to a code review because theres no code behind an agent, and prompts arent code. That, he argues, is precisely why evaluations matter and should be the foundation of responsible AI deployment. The shift from focusing on belief to emphasizing behavior is the foundation of two major innovations by Databricks this year: Test-Time Adaptive Optimization (TAO) and Agent Bricks. Together, these technologies seek to make behavioral evaluation the first step in enterprise AI, rather than an afterthought. AI behavior matters more than raw knowledge Traditional AI evaluation often relies on benchmark scores and labeled datasets derived from academic exercises. While those metrics have value, they rarely reflect the contextual, domain-specific decisions businesses face. In production, agents may need to generate structured query language (SQL) in a companys proprietary dialect, accurately interpret regulatory documents, or extract highly specific fields from messy, unstructured data. Naveen Rao, vice president of AI at Databricks, says these are fundamentally behavioral challenges, requiring iterative feedback, domain-aware scoring, and continuous tuning, not simply more baseline knowledge. Generic knowledge might be useful to consumers, but not necessarily to enterprises. Enterprises need differentiation; they must leverage their assets to compete effectively, he tells Fast Company. Interaction and feedback are critical to understanding what is important to a user group and when to present it. Whats more, there are certain ways information needs to be formatted depending on the context. All of this requires bespoke tuning, either in the form of context engineering or actually modifying the weights of the neural network. In either case, he says, a robust reinforcement learning harness is essential, paired with a user interface to capture feedback effectively. That is the promise of TAO, the Databricks research teams model fine-tuning method: improving performance using inputs enterprises already generate, and scaling quality through compute power rather than costly data labeling and annotation. While most companies treat evaluation as an afterthought at the end of the pipeline, Databricks makes it central to the process. TAO uses test-time compute to generate multiple responses, scores them with automated or custom judges, and feeds those scores into reinforcement learning updates to fine-tune the base model. The result is a tuned model that delivers the same inference cost as the originalwith heavy compute applied only once during tuning, not on every query. The hard part is getting AI models to do well at your specific task, using the knowledge and data you have, within your cost and speed envelope. Thats the shift from general intelligence to data intelligence, Frankle says. TAO can help tune inexpensive, open-source models to be surprisingly powerful using a type of data weve found to be common in the enterprise. According to a Databricks blog, TAO improved open-source Llama variants, with tuned models scoring significantly higher on enterprise benchmarks such as FinanceBench, DB Enterprise Arena, and BIRD-SQL. The company claims the method brought Llama models within range of proprietary systems like GPT-4o and o3-mini on tasks such as document Q&A and SQL generation, while keeping inference costs low. In a broader multitask run using 175,000 prompts, TAO boosted Llama 3.3 70B performance by about 2.4 points and Llama 3.1 70B by roughly 4.0 points, narrowing the gap with contemporary large models. To complement its model fine-tuning technique, Databricks has introduced Agent Bricks, an agentic AI-powered feature within its Data Intelligence Platform. It enables enterprises to customize AI agents with their own data, adjust neural network weights, and build custom judges to enforce domain-specific rules. The product aims to automate much of agent development: Teams define an agents purpose and connect data sources, and Agent Bricks generates evaluation datasets, creates judges, and tests optimization methods. Customers can choose to optimize for maximum quality or lower cost, enabling faster iteration with human oversight and fewer manual tweaks. Databricks latest research techniques, including TAO and Agent Learning from Human Feedback (ALHF), power Agent Bricks. Some use cases call for proprietary models, and when thats the case, it connects them securely to your enterprise data and applies techniques like retrieval and structured output to maximize quality. But in many scenarios, a fine-tuned open model may outperform at a lower cost, Rao says.He adds that Agent Bricks is designed so domain expertsregardless of coding abilitycan actively shape and improve AI agents. Subject matter experts can review agent responses with simple thumbs-up or thumbs-down feedback, while technical users can analyze results in depth and provide detailed guidance. This ensures that AI agents reflect enterprise goals, domain knowledge, and evolving expectations, Rao says, noting that early customers saw rapid gains. AstraZeneca processed more than 400,000 clinical trial documents and extracted structured data in less than an hour with Agent Bricks. Likewise, the feature enabled Flo Health to double its medical-accuracy metric compared with commercial large language models while maintaining strict privacy and safety. Their approach blends Flos specialized health expertise and data with Agent Bricks, which leverages synthetic data and tailored evaluation to deliver reliable, cost-effective AI health support at scaleuniquely positioning us to advance womens health, Rao explains. From benchmarks to business data The shift toward behavior-first evaluation is pragmatic but not a cure-all. Skeptics warn that automated evaluations and tuning can just as easily reinforce bias, lock in flawed outputs, or allow performance to drift unnoticed. In some domains we truly have automatic verification that we can trust, like theorem proving in formal systems. In other domains, human judgment is still crucial, says Phillip Isola, associate professor and principal investigator at MITs Computer Science & Artificial Intelligence Laboratory. If we use an AI as the critic for self-improvement, and if the AI is wrong, the system could go off the rails. Isola points out that while self-improving AI systems are generating excitement, they also carry heightened safety and security risks. They are less constrained, lacking direct supervision, and can develop strategies that might be unexpected and have negative side effects, he says, also warning that companies may game benchmarks by overfitting to them. The key is to keep updating evaluations every year so were always testing models on new problems they havent already memorized. Databricks acknowledges the risks. Frankle stresses the difference between bypassing human labeling and bypassing human oversight, noting that TAO is simply a fine-tuning technique fed by data enterprises already have. In sensitive applications, he says, safeguards remain essential and no agent should be deployed without rigorous performance evaluation. Other experts note that greater efficiency doesnt automatically improve AI model alignment, and theres no clear way to measure AI model alignment currently. For a well-defined task where an agent takes action, you could add human feedback, but for a more creative or open-ended task, is it clear how to improve alignment? Mechanistic interpretability isnt strong enough yet, says Matt Zeiler, CEO of Clarifai. Zeiler argues that the industrys reliance on a mix of general and specific benchmarks needs to evolve. While these tests condense many complex factors into a few simple numbers, models with similar scores dont always feel equally good in use. That feeling isnt captured in todays benchmarks, but either well figure out how to measure it, or well just accept it as a subjective aspect of human preference; some people will simply like some models more than others, he says. If the results from Databricks hold, enterprises may rethink their AI strategy, prioritizing feedback loops, evaluation pipelines, and governance over sheer model size or massive labeled datasets, and treating AI as a system that evolves with use rather than a onetime product. We believe the future of AI lies not in bigger models, but in adaptive, agentic systems that learn and reason over enterprise data, Rao says. This is where infrastructure and intelligence blur: You need orchestration, data connectivity, evaluation, and optimization working together.
Category:
E-Commerce
Want more housing market stories from Lance Lamberts ResiClub in your inbox? Subscribe to the ResiClub newsletter. To identify which metro areas might offer buyers the most opportunity right now, ResiClub analyzed the share of homes on Zillow with a price cut. Of course, just because a home listed for sale gets a price cut doesnt guarantee that comps in that area are falling. After all, there will always be some sellers who overshoot their market. Even during the height of the Pandemic Housing Boom in June 2021, 11.4% of homes for sale that month saw a price cut. The way to interpret this data is to compare it over time. If the percentage of homes with a price cut in a given area decreases noticeably (particularly, beyond the normal seasonal swing), it suggests the market has tightened and sellers have gained relative leverage. Conversely, if the percentage increases noticeably (again, beyond the normal seasonal swing), it suggests the market has cooled and buyers have gained relative leverage. The national share of U.S. homes with a price cut is further confirmation that the housing market has been shifting, relatively speaking, toward buyers ever since mortgage rates spiked and the Pandemic Housing Boom fizzled out: June 2018 > 18.1% June 2019 > 20.1% June 2020 > 14.8% June 2021 > 11.4% June 2022 > 17.2% June 2023 > 19.1% June 2024 > 23.5% June 2025 > 25.6% Click here to view a searchable/sortable table with data for more than 900 metropolitan and micro-area housing markets. If youre a ResiClub readerespecially a ResiClub PRO memberneither the softening nor the bifurcation should surprise you. As ResiClub has well documented, many housing markets in the Northeast and Midwest have thus far had a milder and slower softening, while many areas in pandemic boom areas in the Mountain West and Sunbelt have seen a faster and greater softening in the post-boom market. To help you better view the story, weve created a map showing the share of homes that saw a price cut in June for every year since 2018. Below is June 2025 !function(){"use strict";window.addEventListener("message",function(a){if(void 0!==a.data["datawrapper-height"]){var e=document.querySelectorAll("iframe");for(var t in a.data["datawrapper-height"])for(var r,i=0;r=e[i];i++)if(r.contentWindow===a.source){var d=a.data["datawrapper-height"][t]+"px";r.style.height=d}}})}(); Below is June 2024 Below is June 2023 Below is June 2022 Below is June 2021 Below is June 2020 Below is June 2019 Below is June 2018 According to Zillow: Share of listings with a price cut: The number of unique properties with a list price at the end of the month thats less than the list price at the beginning of the month, divided by the number of unique properties with an active listing at some point during the month.
Category:
E-Commerce
For hardcore retro-tech fans and Steve Jobs groupies, a treasure trove of vintage Apple devices, ultraexclusive memorabilia, and forgotten tech has just been collected into one websiteand it’s all for sale. The collection of items, titled Steve Jobs and the Apple Revolution, is currently being sold by RR Auction, and will remain live until August 21. Its one of the companys 12 annual speciality auctions, which focus on specific subjects like space exploration, the Olympics, and animation. “Created over a decade ago, this signature auction tells Apple’s full arcfrom garage-built Apple-1 to world-changing innovations, says Bobby Livingston, executive vice president at the Boston-based auction house. We source directly from early engineers, employees, and elite collectors, often bringing items to market for the first time. It’s Apple’s history told through the objects that made it possible. From Apple-1 to the iPod RR Auctions Apple-centric collection comes as interest in retro tech is taking off, especially among younger generations. Recently Sega, Commodore Corp., and Fujifilm have all released new products that buy into the demand for vintage-tech aesthetics. Retro gaming is also riding high on a Gen Z-fueled resurgence. [Screenshot: RR Auction] RR Auctions Apple collection, though, is composed of true vintage items, some of which are one of a kind. A particular highlight is the fully functional Apple-1 computerthe first Apple device ever builtsigned by cocreator Steve Wozniak and early Apple employee Daniel Kottke. According to an analysis by eBay, only about 200 Apple-1 devices were ever built, with just 82 believed to still exist. In 2022, eBay sold an Apple-1 for $340,100 at auction. Livingston points to one specific check in the collection, signed on March 28, 1976, as another standout object: “Check No. 6written four days before Apple’s founding, signed ‘steven jobs,’ listing all three cofoundersreads like Apple’s birth certificate, he says. [Screenshot: RR Auction] Other items of note include a rare Lisa computer, released in January 1983, with its custom Twiggy floppy drives intact; a prototype iPod with a red logic board; a factory-sealed 4GB iPhone; and an assortment of vintage Apple-branded merch. Together, they track Apple’s evolution from startup to giant, Livingston says. [Screenshot: RR Auction] So far, Livingston adds, interest in the collection has been “extraordinary,” ranging from veteran collectors to first-time bidders. With more than a week left for incoming bids, the Apple-1 computer has already surpassed the 100,000 threshold, while many other items have top bids in the tens of thousands. These aren’t just nostalgic artifacts; they’re cultural touchstones, Livingston says. The strongest interest comes from seasoned tech collectors and younger successful entrepreneurs who see these as physical chapters of a story still shaping the world.
Category:
E-Commerce
For something as simple as setting a timer, the built-in apps on our computers can be awfully fiddly. Usually you have to open a Clock app first, then navigate to a separate tab for timers. After that you have to hit another button to create the timer, and only then can you finally set the time. You might even have to wade through a messy list of all the previous timers youve created. Fortunately, theres a faster way when time is of the essence. Even better, its full of powerful features that dont detract from its up-front simplicity and delightfulness. This tip originally appeared in the free Cool Tools newsletter from The Intelligence. Get the next issue in your inbox and get ready to discover all sorts of awesome tech treasures! A plain but powerful desktop timer The next time you find yourself with a timer-needing task, you can skip past all the cruft of your built-in Clock app by just opening your favorite web browser and visiting the appropriately named E.ggtimer. E.ggtimer is a free web-based app for setting quick timers with simple keyboard commands. Setting a timer takes just a few seconds. To set up a basic timer, try typing 5 minutes or 5m into the box on the e.ggtimer landing page. This also works with other units of time, so you can type 3h48m15s for a timer that runs for three hours, 48 minutes, and 15 seconds. When time expires, the site will play a tone andwith your permissiondeliver a push notification to your device. E.ggtimer’s interface is both easy to use and packed with interesting extras. To speed things up further, just include the time directly in URL. For instance, entering e.ggtimer.com/5m into your address bar will bypass the setup page and set a five-minute timer immediately. You can even bookmark your most-used timers for faster access in the future. I was chatting with my fellow Cool Tools writer JR Raphael about this, and he asked a valid question: Why use E.ggtimer instead of, say, Google Searchs built-in timer tool? To this I offer a few answers: E.ggtimer supports more time formats. In addition to the hour-minute-second format I mentioned earlier, E.ggtimer lets you input absolute times such as 4:56 pm or August 10 2025 3pm. It can also count down to holidays such as Mothers Day or New Years. E.ggtimer works offline. If youre using Chrome, click the little download icon in the top-right of the address bar to install E.ggtimers Progressive Web App. Now you can launch a freestanding version of the site that works without an internet connection. E.ggtimer supports the Pomodoro method. Beyond just individual timers, you can also set up sequences of consecutive timers. Try typing 25m/5m/25m for a pair of 25-minute work sessions broken up by a five-minute break. E.ggtimer is more fun. Beyond the basic black text on white background, the site offers a bunch of themes, including a digital clock, dot matrix, andmy personal favoriteone that looks like the Windows Blue Screen of Death. E.ggtimer’s amusing take on the classic Blue Screen of Death. This only scratches the surface of whats possible. Check out the sites Help and Settings page for even more possibilities, such as adding labels to your timers and tweaking things like the default alarm sound. One quick caveat: While E.ggtimer technically works in any browser on any device, Ive found that its alerts dont come through on mobile devices unless your screen is on and the site is open, so youre better off using it on desktop browsers only. Too bad, because the built-in Clock apps on our phones are just as fiddly as their desktop counterparts. E.ggtimer is entirely web-based, though you can download it as a Progressive Web App if you would like. It is free to use and doesnt include any ads (unless you select the Ugly theme which has fake ads on the page). E.ggtimer doesnt ask for any personal information to use the service. Treat yourself to all sorts of geeky goodies like this with the free Cool Tools newsletterstarting with an instant introduction to an incredible audio app thatll tune up your days in truly delightful ways.
Category:
E-Commerce
In the filtered water space, there is one company that has dominated brand awareness for decades. Water pitchers and filtration devices from Brita can be found in so many millions of homes and offices around the world that the term market saturation is more than just a pun. But there’s another water filtration company that, despite lower kitchen visibility, is actually a bigger player in the clean-water game. Culligan, founded in 1936 as a water softening and filtration service company, became known for its white-glove service. [Photo: Culligan] Often installed in basements or storage closets, Culligan’s equipment was as utilitarian as a water heater or furnace. Once the system was installed in a home or office, its users hardly gave it another thought, or look. “It was the technician that was actually working with the product,” says Kathy Chi Thurber, Culligan’s new global president of consumer products. “The products didn’t have to be beautiful, but the technicians had to be able to talk about our history, our capabilities, our research, and innovation.” Now, as a private 15,000-person company that pulled more than $3 billion in revenue in 2023, Culligan is embarking on a total brand and strategy overhaul. And aggressively so. Within the past five years, Culligan has acquired 362 companies in the clean-water industry, from local water purifiers to filter companies to component manufacturers. It’s positioning itself as a dominant player in a world where water safety and water scarcity are of increasing concern. [Photo: Culligan] Out of the basement and into your kitchen One priority is to start competing more directly in the consumer space, bringing its equipment out of the basement and into the hands of water drinkers everywhere. “We’d never really given an eye to the consumer, and that has 100% changed,” says Chris Quatrochi, chief product and technology officer at Culligan International. To venture into the Brita-dominated consumer market, Culligan turned to the industrial design firm Ammunition Group. Known best for its work designing Beats by Dre headphones and products for companies like Polaroid, Square, and Lyft, Ammunition was tasked with helping Culligan develop products that appeal to regular consumers. It also updated the brand to tell those consumers that Culligan is not the box-in-the-basement brand they may have known in the past. “Our portfolio has not been the greatest from a, I would say, beauty perspective,” Quatrochi says. “If you really want to show that you are leading edge from a water-quality perspective, you have to have a product that demonstrates that.” Ammunition started by applying its deep product design background to creating a water filtration pitcher that embodies this new company focus. Building on its 2020 acquisition of the water filter maker ZeroWater, Culligans ZeroWater Technology line of three handheld pitchers and two countertop dispensers is the companys first foray into the consumer space. [Photo: Culligan] Designing a better water pitcher Ammunition’s design focused primarily on the ways people actually use filtered water pitchers. “One of the constraints is putting it in your refrigerator,” says industrial designer Robert Brunner, Ammunition’s founder. Research into the market showed that more than 70% of water pitcher users, particularly those in the U.S. and Western Europe, keep their pitchers in the refrigerator, often in the door of the appliance. At the same time, most of the pitchers on the market don’t actually fit into a fridge door all that well. Their rectangular shape and bulging handle tend to take up a lot of space, and need more room around them to be moved in and out. [Photo: Culligan] Ammunition rethoght that form factor to better fit inside the refrigerator door, using a rounded square shape for the pitcher that allows it to fit more like a carton of milk. The pitcher also has an innovative open-ended handle that cuts down on its overall bulk and allows more stuff to fit in the refrigerator door’s shelves alongside it, while also being more ergonomically comfortable to carry and hold. “Figuring out how to have that single connection point for that handle so it’d be mechanically robust and reliableit was actually a fair amount of engineering effort to make sure that could work when it’s getting filled up with water,” Brunner says. “The handle is extremely important, because when this thing is full, it’s quite heavy, and you have to be able to manipulate it, carry it, pour it. We wanted to maintain this simplicity.” The design team also thought about the spout shape and the challenge of pouring water for people with dexterity and mobility issues. That led to considerations about one of the key parts of using a water pitcher: refilling it. Ammunition designed a sliding lid that makes holding the pitcher under the tap and refilling it easier. The lids circular shape became a recurring theme in the design of the pitcher line, as well as the broader work Ammunition is doing across Culligan’s other product and service categories. “The circular element is really the most natural shape to route water from one place to another, pipes being the most obvious example,” says Christopher Kuh, vice president of Ammunition’s industrial design studio. “It’s really an important and core element.” Another differentiating factor is the built-in water-quality meter. Measuring total dissolved solids (TDS) at the scale of parts per million, the digital meter slots into the pitchers and the countertop dispensers to give users a clear readout of how well the filter is functioningand when it’s time to replace it. “The TDS meter actually is going to start to read a value above zero at some point in time, which gives you a clear indication of the end of filter life,” Kuh says. In a clever turn, the meter can be removed from the pitcher or dispenser to dip into, say, a glass of water direct from the tap to see just how much the filtration system is doing. [Photo: Culligan] A bigger rebrand moment These design moves were informed by deep user research Culligan has conducted over the past three years. Thurber says Ammunition was game for putting its design prototypes in the hands of users from the very early stages and taking their feedback to inform new iterations of the designs before landing on a final product that looks and feels different from what’s already out there. “We all know who the major competitor is that has, like, 60% to 70% market share,” Thurber says. “It would be very hard to break through if we were not serious about what we wanted to do, and if we were not game-changing in our design and our functionality.” But this doesn’t mean Culligan is abandoning the more utilitarian water products that have kept it in business for nearly a century. Instead, Ammunition’s design approach for the pitcher is being extended throughout Culligan’s product offerings, including the industrial-scale water softeners and filtration systems that still live in basements and utility closets, as well as the company’s large and growing business in office water coolers. Some of those redesigned products will be coming online in the next year.
Category:
E-Commerce
Sites : [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] next »