SCROLL NEWS / DISCOVERY
Search headlines
Find the stories shaping the conversation.
Results for " Hub " 6,879 found 🔀 AI-powered Shuffle
Colorado small business owner struggles with record diesel prices and energy inflation: "Prices are exorbitant"
With diesel prices nearly doubling, businesses that rely on diesel are hurting.
Political ad attacking CD8 candidate Manny Rutinel gets a Reality Check
Suspects smash car into business in Broomfield, steal vapes worth $5,000
Amid Ed Sheeran controversy, Philadelphia DJ Patty Jackson says it's "a new era" for music, politics
Baltimore businesses hoping to win big with Ravens and Orioles in town for back-to-back home games
Xavier Taylor's father shares health update 4 months after freak baseball injury: "He hears us"
Baltimore businesses hoping to win big with Ravens and Orioles in town for back-to-back home games
Thousands of Ravens and Orioles fans are expected to soar into downtown Baltimore, which means a big boost for local businesses.
NCDOA holds series of forum for minority small business owners after elimination of HUB in budget
Following its elimination in the new state budget, minority businesses will no longer have the aid of the Office for Historically Underutilized Businesses, which helped provide assistance with state contracts for more than 25 years.
NCDOA holds series of forum for minority small business owners after elimination of HUB in budget
Following its elimination in the new state budget, minority businesses will no longer have the aid of the Office for Historically Underutilized Businesses, which helped provide assistance with state contracts for more than 25 years.
Colorado health officials warn of possible measles exposure in Denver, Steamboat Springs
Officials from the Colorado Department of Public Health and Environment issued an alert about a possible measles exposure after a confirmed case with an unvaccinated out-of-state visitor earlier this week.
Book excerpt: "Go and Do Likewise" by Andy Beshear
The Democratic Governor of Kentucky writes about how the language of faith, which he believes has been co-opted by some politicians, can be used to bridge the nation's partisan divide.
Gov. Pritzker Launches Illinois Reproductive Health Care Hub
New website centralizes resources for reproductive care in Illinois, connecting site visitors to medical providers, affordable care options, and helpful information
ResearchHub: An Answer to “Dysfunction” in Traditional Science Giving?
The platform, cofounded by cryptocurrency billionaire and philanthropy critic Brian Armstrong, aims to eliminate counterproductive incentives that scientists face in traditional grant seeking.
Arlington officer kills man armed with knife during mental health call, police say
Engrams Embedding Entendre: Codesign for Efficient DRAM/SSD Offloading
Engram extends standard token embeddings with learned multi-token lookups. Recurring local patterns retrieve vectors directly, reducing the need to reconstruct them through attention and feed-forward layers. With Engram model architecture optimization, it allows for lower HBM capacity to be needed for models at the same quality. This does not mean there won’t be an insane demand for HBM but it just means that model architecture will continue to innovate around constraints. This model architecture design is naturally codesigned for parameter offloading: each token accesses a few embedding rows whose addresses depend on token IDs, not hidden states. The runtime can prefetch those rows from host DRAM while earlier layers compute, keeping the table outside HBM without transferring entire weight matrices. Our Memory model contains our latest estimates of quarter by quarter HBM, DRAM, & NAND supply and demand. Offloading frees HBM for model weights and KV cache, potentially supporting larger batches or more concurrent sessions. When DRAM becomes the next constraint, NVMe offers another tier. Recommendation systems already cache frequently or recently accessed embedding rows in faster memory while backing colder rows with SSDs. After NVIDIA roadmap had to change due to massively despec’ing Rubin Ultra from 1024GB to now ~200GB of HBM per chip, model architecture optimizations like emgram maybe helpful. Our DeepSeek-V4.1-Flash configuration uses roughly 189 GiB of memory for Engram. We replace it with a memory-mapped (mmap) file and measure serving performance with offload to SSD. Later on in our report, we will show our Engram offloading experiments along with the official InferenceX agentic inference serving results on Engram models like DeepSeekv4.1 Flash across all 6 NVIDIA GPU SKUs along with MI355X. Unsurprisingly, the CUDA Moat is still mogging MI355X on the ultra popular DeepSeekV4.1 Flash model. We also show how even on high capacity HBM SKUs, offloading emgrams to DRAM could result in even better performance for most of the pareto than keeping the emgram in HBM. Our benchmark has been widely reproduced, validated and/or supported by almost every major buyer of compute from Google Cloud to Microsoft Azure to Oracle, to Meta and many more. Furthermore, it has the support of the ML community including from vLLM, LMCache, SGLang, PyTorch, Huggingface and the support of major labs like OpenAI, MiniMax, ZAI, Qwen, Moonshot Kimi, etc. Star the InferenceX GitHub repository if you find the open-source benchmark and data useful!. InferenceX is the only inference benchmark in the world to have TPUv7, Jalapeño, Nvidia Rubin NVL72, AMD, and soon, SambaNova and Trainium. Due to how realistic AgentX scenario is to real world agentic inference workloads, AMD has committed to collaborating on MI455X UALoE72 too. DeepSeek did not release the original paper’s two trained Engram models. We replicated its setup on fineweb-edu using the released code and training hyperparameters, at an estimated 6E18 FLOPs per run. We observed the same U-shape scaling: Engram improved performance over pure MoE baselines. We also reproduced DeepSeek’s results, where earlier-layer representations with Engram resembled those of later layers. Like the original Engram paper, it is possible to probe Engram’s gate scores to see what n-grams DeepSeek-V4.1-Flash makes the most use of. Our gate scan finds names, code fragments, relational phrasing, and boilerplate. These examples prioritize interesting-ness over gate strength. One unexpected result was Wright : Ace Attorney. These examples suggest learned memory optimizes the training objective, not a judgment of which facts deserve storage. Licenses, bibliography fragments, API scaffolding, and website furniture can provide prediction shortcuts, so the value of additional Engram capacity may depend on what survives data preparation. This does not show that table capacity is “wasted”: the evaluation-corpus scan establishes neither training exposure nor the capacity occupied by each category. For offloading, strong gates do not identify cache-hot rows. Low gates do not automatically save reads either: computing the gate requires the retrieved key and undoes the performance gain of a fused kernel. Skipping reads would require a separate usefulness predictor before retrieval. In the original paper’s inference-time ablation, factual-knowledge benchmarks retained just 29–44% of their original performance, while reading comprehension retained 81–93%. This is due to the training–inference mismatch. The resulting degradation therefore measures this trained model’s dependence on Engram, not the performance difference between models trained with and without it. In our ablations, suppressing Engram worsens token likelihood across all evaluated domains, especially encyclopedia text and several code corpora. Surprisingly, GSM8K accuracy stays within measured run-to-run variation and removing Engram has no effect. Engram is not a detachable dictionary beside an unchanged MoE. Removing it changes downstream features and expert selection. We tested whether rerouting hurts or compensates by holding tokens fixed in a teacher-forced experiment on CRUXEval, a code-reasoning benchmark of small Python functions where the model predicts a function’s output from its code and an input, and scoring the reference answer. Removing Engram raised answer loss from 0.2848 to 0.3093 bits/token. Forcing the ablated model to use the original Engram-on expert choices made it worse still, at 0.3375 bits/token. Rerouting partially compensates for the missing memory. Memory features and expert selection work together, rather than following a clean “memory stores facts; experts reason” division. On the same CRUXeval, removing Engram during either phase reduced accuracy and increased generated tokens; removing it throughout produced the largest changes. Keeping Engram for prefill leads to more correct answers than keeping for decode likely due to semantically richer KV cache transferred to decode workers, allowing it to mitigate some of the performance loss. Engram’s table is large, but each lookup is small. DeepSeek-V4.1-Flash requests 24 rows at each of two Engram layers, about 12.4 KiB per processed token position across the model, or 3.1 KiB per GPU when split across four GPUs. Currently as of Day 7 since Model Release, MI355X is still 2-4x worse performance per dollar compared to B200 even when normalized by Mi355X’s lower TCO. Our full total cost of ownership breakdown comes from our AI Cloud TCO Model along with monthly market surveys of over 100+ gpu clouds & gpu cloud customers. On the Day 0 release of DeepSeekv4.1 Flash, NVIDIA vLLM works out of the box with zero issues across all 6 SKUs: H100, H200, B200, B300, GB200, GB300! This was thanks to the amazing work by the NVIDIA & Interact teams! In comparison, AMD vLLM did not work on day 0 for DeepSeekv4.1 Flash.Source: SemiAnalysis InferenceX AMD’s vLLM documentation points to using vllm/vllm-openai-rocm:deepseekv41-flash-0909, but from hour 0 of the model release to hour 23, AMD has not publicly released the image. AMD claims, “SPEED IS THE MOAT,” yet it has still not released it by the 23rd hour. We wish that, going forward, the AMD team has a better process for hour 0 model releases. Eventually when they did publicly release for “day 0” image support, performance-wise, it is currently up to 14.8x worse perf per dollar than H200 and up to 42x worse perf per dollar than B200/B300. The power of the CUDA MOAT is NVIDIA’s collaboration with its massive 6 million-developer community ecosystem including most of the vLLM & SGLang & Tokenspeed maintainers which means that CUDA is optimized on day 0. Overall, AMD did make significant improvements but the performance per dollar is still currently 2-4x worse than B200. AgentX Engram DRAM Offloading Improving Performance The HBM and DRAM offload use the same GPU kernel to select and dequantize rows. With HBM, it reads GPU memory; with Unified Virtual Addressing (UVA), it reads pinned host memory directly. Both support full decode graphs. Moving the table into HBM accelerates only the sparse lookup, leaving decoder computation and communication unchanged, resulting in little overall benefit while consuming memory otherwise available to KV cache. Another benefit of Engram offloaded to DRAM which means you can reduce the communication overhead by using less HBM GPUs per replica. For example, when enabling Engram offloading on B300, we are able to switch from TP4 to now TP2 which improves the pareto curve by up to 1.6x. When iso-model quality, less HBM is required as the engrams could be offloaded to host DRAM. Thus HBM bandwidth matters way more than HBM capacity. For inference workloads where memory bandwidth matters the most, 4-hi HBM provides the best $/bandwidth and therefore lowest cost per token. If China continues to make more and more revolutionary model architecture innovations, soon it could potentially 0Hi HBM stacks. [ ](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi) [ Long Live the Short King: Why 4-hi HBM Wins ](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi) Myron Xie, Bryan Shan, and 3 others · Sep 13 [ Read full story ](https://newsletter.semianalysis.com/p/long-live-the-short-king-why-4-hi) Moreover, on B300 and week-0 stack, moving the Engram table back to HBM did not improve results and stay within run-to-run variance. This is a result of the work optimizing DRAM offload, such as async, and overlap. SSD Offloading On B200, we created a unoptimized vLLM fork and stored the Engram tables in memory-mapped files on local SSDs. File backing lets the OS reclaim table pages when other applications need RAM. Pages already cached in memory can be served without reading the SSD again. Furthermore, note that we were unable to turn on GDS The unoptimized SSD implementation changes how rows reach the GPU. It copies row IDs to the CPU, deduplicates them, gathers the requested rows into pinned buffers, copies those rows back to the GPU and dequantizes them. This work runs between segments of the GPU execution graph. Native UVA performs row selection and dequantization directly on the GPU, avoiding the CPU round trip. A warm filesystem cache removes physical SSD reads, but leaves the coordination, row gathering and transfers. This is why a file already cached in RAM can still perform worse than a pinned DRAM table. The comparison measures the whole serving path; it does not separate the time spent on each of these operations. B200 DRAM dominates both measured SSD serving curves in total tokens per dollar and P90 interactivity. Near 125 tokens/s/user, DRAM delivers 121 million total tokens per dollar versus 52 million for SSD. For production serving, SSD offloading is likely not worth the tradeoff. On the B200 configurations we measured, SSD offloading loses on both measures: every observed SSD point has a DRAM alternative that delivers higher P90 interactivity and more total tokens per dollar. The only points where Cheaper storage does not automatically produce a cheaper inference service. Moving Engram to SSD leaves the same four expensive GPUs and the rest of the server in place. Reclaiming RAM only creates an economic benefit if it enables a cheaper server configuration or additional useful capacity. The current unoptimized path provides neither benefit, and the filesystem cache still consumes RAM when table pages are resident. Next we will look at the mechanisms and specific implementation of ngrams from DeepSeekv4.1 Flash, LongCat, Qwen3.8 Flash Next.
NASA’s Hubble Spots an Out-of-Sync Galaxy
Though the spiral galaxy in this new image from the NASA/ESA Hubble Space Telescope seems serene, it hides a chaotic secret. This galaxy is NGC
Burglary suspects smash car into Colorado business, steal vapes and merchandise worth $5,000
Reality TV star Porsha Williams reveals she had to undergo emergency partial hysterectomy: "It all happened so fast"
Porsha Williams revealed to "CBS Mornings" co-anchor Gayle King that she had to undergo an emergency partial hysterectomy this year after doctors discovered 21 uterine fibroids
Burglary suspects smash car into Broomfield business, steal vapes worth $5,000
Weekend Gardener: 2026 Texas A&M Department of Horticultural Sciences Field Day
Community members are invited to learn more about the research being done in the Texas A&M Department of Horticultural Sciences Department.
The Clicks Communicator Ships in December With More RAM and New Mobile Service
A new video shows off the BlackBerry-like phone's features, including the speech-to-text Prompt Key, Message Hub and new Clicks Mobile wireless service.
Billions at stake as Texas reshapes business contract program
Texas has dramatically overhauled its Historically Underutilized Business (HUB) program in the past year, shifting eligibility toward service-disabled veterans and leaving thousands of minority- an…
Farmer who voted for Trump says "I can't stay in business" amid rising costs
Maryland Gov. Wes Moore seeks federal aid for businesses impacted by Empire Towers emergency
Possible measles exposure reported at Children's Hospital of Philadelphia, health officials say
Health officials in Philadelphia are warning about possible measles exposures in the CHOP NICU.
This plus-size clothing business faced a $90,000 bill because of tariffs. Then the community came together to help
Dr. James Q. Simmons tackles everyday medical concerns in “What the Health”
Trump ordered DOJ not to appeal ruling that loosened limits on "gangster weapons," sources say
President Trump ordered the Justice Department not to appeal a court ruling that loosened restrictions on firearms known as "gangster weapons," sources familiar with the matter told CBS News.
When considering mental health, it's important to be aware of side effects of psychedelics
Gov. Moore requests federal assistance for Empire Towers businesses
Twin Cities gas prices surge 15 cents overnight as business owners feel the pinch
1 in 3 Americans with health insurance have medical debt, survey finds
E-bike, scooter injuries among children are a public health crisis, doctors on Long Island say
NSF invests in translating low-dimensional semiconductor technologies to establish U.S. leadership in microelectronics
The U.S. National Science Foundation today announced an initiative to translate low-dimensional semiconductor technologies from laboratory demonstrations into platforms for U.S. manufacturers to…
Clio hires former Michigan Supreme Court chief justice as GM of its judiciary business
Clio hires former Michigan Supreme Court chief justice as GM of its judiciary business
RWJBarnabas Health, Rutgers and AWS launch innovation hub
The new center gives students access to mentors from the New Jersey health system and Rutgers Robert Wood Johnson Medical School, powered by Amazon's cloud and artificial intelligence technologies.
Examining North Carolina’s Growing Latino Electorate
The analysis, which draws primarily on 2024 data available through the LPPI’s Latino Data Hub, finds that Latino eligible voters are a young and growing segment of the state’s electorate.
1 in 3 Americans with health insurance have medical debt, survey finds
Chubb Announces Leadership Appointments to Digital Business Organization
/PRNewswire/ -- Following Sean Ringsted's appointment as Chief Scientist and Executive Vice President, Chubb Group, Chubb Limited (NYSE: CB) today announced...
Sponsored Content: Better Business Bureau, CT – ignite: A New Hub for Business Growth, Collaboration & Innovation
NEW HAVEN, Conn. (WTNH) — The Better Business Bureau of Connecticut is expanding its mission beyond business reviews and consumer resources with the launch of ignite sparked by BBB, a 6,000-square-…
When DNA sounds a lasting alarm: cGAS-STING links aging to brain inflammation
Neurodegenerative diseases, cerebrovascular disorders and other age-related neurological conditions are major causes of long-term disability. A review published in SCIENCE CHINA Life Sciences by researchers from Huashan Hospital, Fudan University, examines how damaged nuclear DNA and mitochondrial DNA can become immune danger signals. The authors highlight cGAS-STING as a key molecular hub linking cellular damage, inflammaging and neuroimmune dysfunction, and discuss opportunities for more precise therapeutic intervention.
West Suburban Medical Center, Weiss Memorial now under Insight Health Systems
Two shuttered Chicago-area hospitals are under new control on Thursday morning.
West Suburban Medical Center, Weiss Memorial now under Insight Health Systems
Doral business owners say increased ICE activity is hurting business
Google Home just became a power user’s paradise
Google Home MCP opens the smart home to every AI agent; Android 17 QPR1 and One UI 9 land the same week; Trash: the 39 toll booth for your light bulbs.
Aurora Health Care 'health and wellness hub' going into closed school building in Milwaukee
A new health center is set to open next year inside a closed school building in Milwaukee.
America has more millionaires than you think. Here's how most of them got rich.
Doral mayor says city can't stop ICE operations as businesses report 70% drop in sales
Oakland small business owner looks forward to downtown revitalization effort
Oakland small business owner optimistic for downtown revitalization effort
Dreamforce conference brings another boost to downtown San Francisco businesses
Maryland health officials warn of possible measles exposure at 3 Ocean City locations
The Maryland Department of Health is warning about possible measles exposure at three locations in Ocean City earlier this month.
Luminis Health sued over alleged failure to protect patient data after cyberattack
A patient of Anne Arundel Medical Center alleges in a lawsuit that the hospital's parent company, Luminis Health, negligently failed to protect their medical information following a cybersecurity incident earlier this month, according to our media partner, The Baltimore Banner.
Celebrating Sacramento's 916 Day and looking into its impact on local business
Hubbard Homestead Park Sports Court Opens
Seattle Parks and Recreation (SPR) is happy to announce the new sports court at Hubbard Homestead Park is complete! We are celebrating with the community and our partners, One Roof Foundation, the philanthropic arm of the Seattle Kraken and Climate Pledge Arena, on Thursday, September 17 from 3 to 5 pm. Please join us at […]
Florida Democratic governor hopeful David Jolly says politics are at a "1974 post-Watergate moment"
Packers sign 320-pound defensive tackle to the practice squad
The Packers are taking a look at bolstering depth along the defensive line. According to Bill Huber of Packers on SI, the Packers have signed former Indianapolis Colts defensive tackle Eric Johnson to the practice squad. Packers Sign DL Eric Johnson Johnson, 6-foot-4 and 320 pounds, has appeared in 51 career games while producing 41 [...]
Senate Democrats meet with Abdul El-Sayed, rallying around Michigan nominee
Hubbard Hall announces Fall 2026 season
CAMBRIDGE, NY — Hubbard Hall, the renowned arts center in Washington County, anchored by its historic 1878 Opera House, announces the Fall 2026 season will open with A.R. Gurney’s Off-Broadway
House votes unanimously to hold Epstein business associate Leon Black in contempt of Congress
NSF announces 3 additional topics as part of the NSF X-Labs initiative to pursue generational breakthrough science and technology efforts
NSF X-Labs invites proposals from teams working in AI for physical systems and announces two forthcoming topics...
Introducing ‘865 Neyland’: Tennessee athletics names new entertainment district
A name has emerged for Neyland Stadium’s new entertainment district. The multi-million dollar mixed-use gathering hub at the University of Tennessee will be named “865 Neyland,” based on Knoxville’s area code — which also spells out V-O-L on the telephone keypad — and the stadium it lies beside. The development’s future address will also be...
SEC Signs Off on Finra Outside Business Reforms
Finra’s proposed rule change eliminates a reporting requirement for lower-risk side jobs and ends broker-dealer supervision of advisors’ work at unaffiliated RIAs.
BTR nears transition to an independent airport authority
The Baton Rouge Metropolitan Airport is targeting the end of October to complete its transition from city-parish governance to a stand-alone airport authority. That’s according to Louis Hubbard I… Already an INSIDER? Sign in. Already an INSIDER? Sign in
Lincoln College property to become business hub, leased for just $1
"This is an incredible opportunity for organizations to grow in a community that has so much to offer," said Uriah Kilgallin, CEO of LEAD, in a news release.
Colorado health officials warn that person with measles visited several locations
Health Foundation of South Florida Invests $225,000 to Establish Overtown Youth Center as a Community Health Access Hub
/PRNewswire/ -- The Health Foundation of South Florida today announced a $225,000 investment over three years to support the Overtown Youth Center (OYC) as it...
Georgia health care workforce shortage puts pressure on patients
Olympic gold medalist Aly Raisman on how she learned to advocate for her own health
In the latest episode of "Healthful," Olympic gold medalist Aly Raisman discusses what she's learned about speaking up for herself and taking care of her body.
Starbucks Explores Majority Stake Sale in Japan Business at Potential $3 Billion Valuation
Starbucks (NASDAQ:SBUX) is considering selling a majority stake in its Japanese operations in a transaction that could value the business at approximately $3 billion, according to two people familiar with the matter. The company has invited several financial advisers to present options for the business and is open to relinquishing majority ownership, the sources told Reuters.
Rutgers, RWJBarnabas Health launch innovation hub featuring Amazon Web Services cloud technology
Rutgers University and RWJBarnabas Health announced the launch of the Rutgers/RWJBH Innovation Hub, powered by Amazon Web Services.
Colorado science teacher hopes his Tyrannosaurus rex trackway discovery fuels ferocious passion for learning
A research associate with the Denver Museum of Nature and Science who teaches in Northglenn recently became the first person to ever identify a Tyrannosaurus rex trackway.
Students sent home after health inspector closes school cafeteria
Presidio Trust to host free wellness events in partnership with Sutter Health
Presidio Trust partners with Sutter Health to host free wellness events
Gov. Healey calls for new investments in postpartum healthcare after Lindsay Clancy trial
Scammers target Hennepin County fund meant to aid businesses impacted by ICE surge
Arc System Works are working on a brand new fighting game property as of a few months ago but it kind of slipped under the radar
Today, Automaton Media reported that for the fiscal year which ended May 2026, Arc System Works has returned to being profitable after posting a net loss in the year prior.This in itself is of course good news for the ...
Rutgers and RWJBarnabas Health to Launch Innovation Hub, Powered by Amazon Web Services
Photo: Executives from Amazon Web Services, Rutgers Robert Wood Johnson Medical School and RWJBarnabas Health gather at the AWS Imagine Summit in Chicago after announcing the launch of the Rutgers/RWJBarnabas Health Innovation Hub, powered by AWS, one of the nation’s first technology innovation hubs that brings together a medical school, academic health system and cloud[...]
RUTGERS AND RWJBARNABAS HEALTH TO LAUNCH INNOVATION HUB, POWERED BY AMAZON WEB SERVICES
/PRNewswire/ -- Rutgers University and RWJBarnabas Health today announced the launch of the Rutgers/RWJBarnabas Health Innovation Hub, powered by Amazon Web...
Griffin Museum of Science and Industry to redevelop historic main entrance with new Welcome Hall
The Griffin Museum of Science and Industry plans to construct a new Welcome Hall for its primary entrance facing 57th Drive and Chicago's Hyde Park neighborhood.
Detroit nonprofits team up to create family services hubs
Starfish Family Services and Brilliant Detroit set up shop next door to each other to build a pipeline of services in Detroit neighborhoods.
Georgia health officials urge vaccinations as state grapples with higher number of measles cases
Georgia has confirmed 16 cases of measles so far this year, most of which are in young children.
From Wirehouse Advisor to Business Owner to Legacy Builder
Kimberly Parker-Hansen of VersiFi uses LPL to offer a broader range of solutions and help clients look at their financial future differently.