SCROLL NEWS / DISCOVERY
Search headlines
Find the stories shaping the conversation.
Results for " python " 26 found 🔀 AI-powered Shuffle
Interns In Florida Released A 16-Foot, 148-Pound Python Back Into The Everglades On Purpose. She Now Carries A Radio Tag For Drones To Follow - Futura-Sciences
In a state where around 100 contractors remove pythons all year round, letting one go looks like a mistake. It is how researchers find the others.
Teradata Brings Elastic Python Compute to AI Studio, Unifying Data, Governance, and AI
New distributed Python capability accelerates large-scale AI operations while keeping data and compute within the Teradata platform SAN DIEGO, Oct. 14, 2026 — Teradata today announced the general availability of Elastic Python Compute within AI Studio. The capability gives enterprise data science teams a managed, distributed Python environment for scaling AI compute and machine learning […]
'Monty Python' legend John Cleese doubles down after incidiary posts about Islam spark UK compaint
Adult female pythons cause the most trouble in Florida
Hunting them first can help curb the invasive species.
Small business owner turns invasive Everglades pythons into luxury handbags
"It's not their fault that they are here, but they are, and they have to go."
Fall plant sale, 'Monty Python,' 'Jersey Boys' and more events: Around Baton Rouge
Looking for something to do?
Run Positron on Amazon SageMaker AI for data science workflows
Positron, Posit's IDE for data science, now runs on Amazon SageMaker AI. This post shows how a data scientist explores an Amazon Athena table, validates features in R, trains an XGBoost model in Python, deploys a real-time SageMaker AI endpoint, and reports results with Quarto, all in one governed SageMaker Studio Space.
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.
From Python and Statistics to Agentic AI: 5 Data Science Certificate Programs in 2026
Python and statistics still sit at the center of data science, but the work surrounding them has expanded. Professionals now move from cleaning data and testing hypotheses into predictive modeling, recommendation systems, deep learning, Generative AI, and applications that can retrieve information or act through external tools. That progression matters when comparing programs. A curriculum […]
Text Watermarking in Python: Catch Whoever Copies Your Writing
AI companies quietly watermark billions of words a day. Here’s how to apply the same three families of techniques to your own writing—and what real experiments reveal about which watermarks survive copy-paste, editing, and paraphrasing.
Eastlake science students make district history
Three Eastlake High School students made Sweetwater Union High School District history this year when they became the first students from the district to qualify for the International Science and Engineering Fair, an international competition that brings together some of the top high school researchers from around the world. Ronald Wyatt, Giles Del Priore and Ihan Sung advanced to the competition after successful runs at regional and state science fairs. Sung went a step further, placing fourth in the chemistry category at ISEF. The students were recognized on Aug. 10 during a Sweetwater Union High School District board meeting at the district office in Chula Vista.For Wyatt and Del Priore, recent Eastlake graduates, the journey began with a shared concern about dementia and the challenges faced by families caring for loved ones with memory impairment. Their engineering project, “Design and Evaluation of a Behaviorally Grounded Caregiver-Governed AI Companion for Individuals with Memory Impairment,” earned the Sweepstakes Award at the Sweetwater District STEM Fair and first place in its division at the Greater San Diego Science and Engineering Fair. The project explored whether Artificial Intelligence could provide companionship and support while helping reduce stress for people with dementia and their caregivers. “Our inspiration came from personal experience. Both of us have seen family members and loved ones affected by dementia, and we’ve witnessed how overwhelming the caregiving process can be,” Wyatt said. Del Priore said the students also wanted to address the high cost of third-party health care services while creating something that could potentially help their own families as well as people elsewhere. The students developed an AI system designed with safeguards intended to prevent misinformation reinforcement, emotional escalation and unsafe responses. In 240 controlled trials involving memory, identity, agitation and neutral scenarios, they said the system consistently prevented harmful incorrect reinforcement while using context-aware silence, de-escalation and caregiver verification. The project required the students to combine Artificial Intelligence, psychology, neuroscience and human-computer interaction. For Del Priore, one of the biggest technical challenges was getting the AI companion to communicate verbally with a patient. Hardware limitations required the students to collaborate with multiple Python libraries for speech-to-text and text-to-speech functions while keeping the cost of the companion low. “For us, the biggest challenge was translating complex behavioral and cognitive care principles into a functioning AI system,” Wyatt said.Qualifying for ISEF was a milestone that student expected. The international competition proved to be another milestone. “Competing at ISEF was an incredible experience and unlike anything we’ve ever done before,” Del Priore said. “Being surrounded by some of the most talented young researchers from around the world was both inspiring and motivating.” He said the experience stood out because of the emphasis on collaboration, innovation and scientific discovery. “The most memorable part was seeing how people from completely different backgrounds and disciplines approached solving real-world problems with creativity and passion,” he said. The students also credited teachers and mentors who supported their work. Del Priore said his sophomore-year AP Physics teacher, Dr. Price, was particularly influential. “Through all of my personal and academic struggles, Dr. Price always believed in me,” Del Priore said. “All of his faith in me really molded me into the person I am today and helped create my love for science even through tough times.” Wyatt is continuing his education at UC Santa Cruz, where he plans to study biomolecular engineering and bioinformatics, with an interest in research involving Artificial Intelligence, biology and health care. Del Priore plans to double major in chemistry and accounting. Sung, a rising Eastlake senior, qualified for ISEF after winning first place in the Engineering: Energy, Materials & Transport category at the Greater San Diego Science and Engineering Fair and first place in chemistry at the California Science and Engineering Fair. His project, “Renewable Ammonia Electrochemical Synthesis by Glow Discharge with an Iron-Based Catalyst,” involved a device he designed and engineered to produce renewable ammonia, with potential applications in fertilizer production and energy storage. Sung became interested in the project after considering the relationship between food insecurity and fertilizer costs. His system uses plasma to break apart nitrogen molecules so they can interact with hydrogen and form ammonia. At ISEF, Sung competed against more than 1,700 students from 65 countries and earned fourth place in chemistry. “I am proud to become one of te first students in the district to qualify for ISEF because this milestone can inspire others in the district to attain similar goals,” Sung said.
Florida Python Challenge winner slayed 96 snakes
Veteran hunter Tom Rahill has battled the invasive species for over 10 years.
Python Is So Slow. Can Julia Solve the Two-Language Problem?
By some benchmarks, Julia code can run 10X to 1,000X faster than Python—but there’s a reason it’s not a very popular programming language.
Pythons' extreme biology may hold clues for treating human disease
These snakes can go for months without eating, grow and shrink the size of their hearts and jump start their metabolism on a dime.
Meet a scientist tracking invasive reptiles, and even their parasites
An insider's look at Florida’s war on invaders: the giant snakes, egg-eating predators and parasites spreading through the Everglades.
Boffin claims Microsoft's supposed quantum leap does not compute due to 'basic Python errors'
Nature paper argues researchers cherry-picked data. Redmond insists its work is sound
Can AI Write Your Code?
What a recent study on ChatGPT, Python, R, and Stata tells us about AI-assisted coding for causal inference
JDownloader site hacked to replace installers with Python RAT malware
The website for the popular JDownloader download manager was compromised earlier this week to distribute malicious Windows and Linux installers, with the Windows payload found deploying a Python-based remote access trojan.
Scientists sacrifice delicious opossums to fight Florida’s invasive pythons
$190 tracking collars lead scientists right to a snake's stomach.
How to Learn Python for Data Science Fast in 2026 (Without Wasting Time)
What I wish I did at the beginning of my journey
The NASA Artemis 2.0 Smartwatch Runs Python And Lets Kids Code Their Own Wearable
NASA's Artemis II lifted off from Kennedy Space Center on April 1, 2026, carrying four astronauts on humanity's first crewed lunar mission in over 50 years. Reid Wiseman, Victor Glover, Christina Koch, and Canada's Jeremy Hansen are currently aboard the Orion spacecraft, preparing for a lunar flyby that will take them farther from Earth than
A geneticist is studying pythons for clues about human health and longevity
A team of researchers believes that pythons may contain clues to help treat a range of human ailments — from heart disease to muscle atrophy, and more.
Causal Inference Is Eating Machine Learning
Your ML model predicts perfectly but recommends wrong actions. Learn the 5-question diagnostic, method comparison matrix, and Python workflow to fix it with causal inference.
Molecule in python blood could pave way for new obesity drugs, scientists say
Researchers find snake metabolite that suppresses appetite of obese mice ‘without some of side-effects’ of GLP-1 drugs
Longest snake ever measured is over 23.5 feet long
Nicknamed the ‘Baroness,' this python is longer than two great white sharks.
Public Health Informatics Specialist
📊 DATA MEETS PUBLIC HEALTH! 📊Northern Nevada Public Health is seeking a PUBLIC HEALTH INFORMATICS SPECIALIST to turn data into life-saving insights! 💡 $78,936/yr + excellent benefits & NV PERS retirement! 📍 Reno, NV | Perfect for you if you have: Bachelor’s in computer science/data sciences and 2+ years health informatics experience; OR an equivalent combination of training and experience ✅ Skills in SQL, R, Python, Power BI desired | Ready to start the new year using your tech skills to improve public health outcomes? This role combines cutting-edge data analysis with meaningful community impact! 🚀 You’ll analyze health data, support disease surveillance, and help shape public health decisions that protect our community! 🌟👉 Apply now: https://www.governmentjobs.com/careers/washoecounty ⏰ Application Deadline: January 30! #HealthInformatics #HealthStatistics #DataScience #PublicHealth #WashoeCounty #TechJobs #RenoJobs