5,979 results for ken

bing.com
Friday the 13th's Jason Voorhees Announced For Hellbreak TCGhttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8cead918dc4e99b61f8422e8815c6c&url=https%3A%2F%2Fbleedingcool.com%2Fgames%2Ffriday-the-13ths-jason-voorhees-announced-for-hellbreak-tcg%2F&c=696431850721389084&mkt=en-usFriday the 13th's Jason Voorhees is being added to the Hellbreak TCG, as players can try the game out at Gen Con 2026 this weekend.
Jul 30, 2026 5:42 AM
bing.com
'Lie-abetes!': CNN panel erupts as MAGA pundit denies ICE to blame for Trump immigration polling slidehttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d3b9fac67452c8105a8054adde95a&url=https%3A%2F%2Fwww.msn.com%2Fen-us%2Fnews%2Fpolitics%2Flie-abetes-cnn-panel-erupts-as-maga-pundit-denies-ice-to-blame-for-trump-immigration-polling-slide%2Far-AA2941PS&c=4793486365807846559&mkt=en-usConservative radio host Jason Rantz’s claim that ICE isn’t to blame for President Donald Trump’s weakening poll numbers on immigration sparked a heated confrontation on CNN NewsNight, with a fellow ...
Jul 30, 2026 4:18 AM
github.com
Show HN: RunNburn – Run a 295B Moe from a 98GB GGUF on a 64GB RAM Desktophttps://github.com/coderredlab/runNburnrunNburn is an Apache-2.0 Rust inference engine for quantized GGUF models that are too big for your fast memory.The core idea: weights stay file-backed (mmap), host residency stays under an explicit byte budget (--ram-budget), and GPU caches are sized from detected free/total VRAM — never from device-name presets. There is no conversion step, no sidecar cache files, no silent requantization. The GGUF on disk is the single source of truth.The result that made me want to post this: Tencent's Hy3 (295B total / 21B active sparse MoE, a single 97.8 GiB Q2_K GGUF) runs on my desktop with 64 GB of RAM and one consumer NVIDIA GPU. The file is larger than RAM and VRAM combined; the selected experts for each token are pulled on demand (the newest path batches O_DIRECT reads through io_uring), while the pretrained routing is left untouched. On the same machine, same prompt, same decode length, a warm-run median gave ~5.5 tok/s decode vs ~2.0 tok/s for llama.cpp.To be upfront about scope: for mode
Jul 30, 2026 1:30 AM
aicontentfootprint.com
Show HN: Energy, carbon and water estimates for AI content, shown as rangeshttps://aicontentfootprint.com/Hi HN,I built this calculator because I noticed it's easy to know how much a LLM costs per token, but it's extremely difficult to know it's carbon footprint or power consumption. I created this as a reference based on some published papers and public information, and it helps to have a guide of the impact of photo and video generation using AI.I also added another factor about Token utilization, it's something that becomes important as the newer models start to use "thinking" modes to reach a conclusion. There is information up to earlier this year, so it's not possible to see the efficiency of newer models like OpenAI Sol or Kimi K3.All the calculator does is to provide estimations and translate them in ordinary figures so people can understand it. There is no login required or any other type of paywall.About the models, it's easier to make estimates for Open Source models (of course) so there's more confidence on that data. For the closed ones, there are estimations based on public a
Jul 29, 2026 4:29 PM
news.ycombinator.com
Four Competing AIs Answer 1 Classic Questionhttps://news.ycombinator.com/item?id=49098762I sought advice from four mainstream A.Is for a basic consumer use case question and to me, the differing style of the answers are quite revealing.Premise: A poorly hung clothes dryer spectacularly ripped from the wall whilst on and tumbling, smashing into the washing machine beneath and trying to exit out the laundry door in a manner reminiscent of Stephen King's haunted laundromat story The Mangler.After all the shards of glass, wall plaster etc were cleared up, the six month old washing machine wasn't working - yes, unsurprisingly but the problem was electronic, not mechanical. The display defaulted to blanks flashing - no error code etc. Having tried all the obvious fixes, I put the dilemma to four A.I.s: Chat GPT, Perplexity, Claude (Pro) and Google Search with A.I. Mode.All suggested more or less the same checkpoints and then a technician but diverged into unique personae + ideas.Chat GPT: Empathetic, sycophanticaly validating "you've taken a laundry mishap and turned it into a q
Jul 29, 2026 3:26 PM
github.com
Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Machttps://github.com/drumih/turbo-fieldfareHi HN,I built a specialized inference engine for running 4-bit Gemma 4 26B-A4B-IT on any M-series Mac using about 2 GB of RAM. It is called TurboFieldfare and is written in Swift and Metal.I have always adored on-device AI. It feels like magic that you can run a powerful NN on your Mac or iPhone. So I wanted to push the limits a bit and run a model whose weights don’t fit in memory.The model’s 4-bit quantized weights occupy roughly 14 GB, which makes running it with conventional inference tools almost impossible on an 8 GB or even 16 GB Mac once the OS, applications, and KV cache are included.The trick is to keep the shared part of the model and the KV cache in RAM, then stream only the routed experts needed for each token from SSD. An SSD is way slower than RAM, so the runtime uses a small expert cache and bounded parallel `pread`. While those reads are in flight, the GPU runs the shared part of the layer.I ran more than 100 experiments. Most didn’t work. A few got me here. The experi
Jul 29, 2026 3:05 PM
github.com
Show HN: Jobsearch – unified CLI for scraping job boardshttps://github.com/jozefRudy/job_searchAuthor here. Got tired of checking Upwork, NoFluffJobs, eFinancialCareers, LinkedIn and HN threads separately, so I built a CLI that scrapes them into a local SQLite DB and lets me browse everything in one web UI with filters and semantic search.Details:- Scraping via real browser over CDP (your own logged-in session, no API keys)- Embeddings run locally (nomic-embed-text via fastembed), stored in Lance- Rust backend, SolidJS frontend embedded in binary, single static executableWorks for my use case (Europe-based contractor). It's my first rust project, decided to do local CLI, it's helpful for me (more providers will be added, feel free to contribute).
Jul 29, 2026 1:05 PM
bing.com
The best and worst Clint Eastwood movies, according to fanshttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8cd1462d8c439c97ff1215b6abb6b7&url=https%3A%2F%2Fwww.aol.com%2Farticles%2Fevery-clint-eastwood-movie-ranked-184325000.html&c=4347978941269234123&mkt=en-usStacker ranked all of Clint Eastwood's films according to their IMDb user rating as of June 2025, with ties broken by the number of votes.
Jul 29, 2026 12:01 PM
strataweather.com
Show HN: Experiments with Weather Datahttps://strataweather.com/mapI had been wanting to try experimenting with nowcasting since the death of Dark Sky and finally got around to it. This application essentially became a playground for me to experiment with various weather datasets and geospatial visualizations.Some interesting concepts that came from it:The nowcast is using pySTEPS optical flow under the hood: https://pysteps.readthedocs.io/en/latest/auto_examples/plot_.... This library is super cool. It bugged me that every radar visualization just gave you choppy snapshots in time, so I spent some time working on building a smooth interpolation. The backend passes keyframes along with motion data from the optical flow analysis and then has a custom WebGL shader to do the interpolation which lets you scrub the timeline and have a smooth interpolation.The other forecast tiles and the main weather page are using the average of many different weather models. You can see every model that factors in on the models page: https://strataweather.com/models. Thi
Jul 29, 2026 11:52 AM
pidgeiot.com
Show HN: PidgeIoT – open-source IoT/Embedded device management for all (free*)https://pidgeiot.com/Hi, I'm Justin! For the past ~3 years, instead of getting a real (well paying) job I've been working on PidgeIoT because, well... I'm stubborn and wanted to see if I could!PidgeIoT is a fully open source, cloud native, device management platform. Originally targeted at cellular connected devices, now aimed at everything! Currently only HTTPS and WebSockets are supported, but CoAP is planned. Bring your own SIM (if needed) and telemetry store.Its features include: device shadows, per-device keypairs (no JWTs!), telemetry ingestion with history, user-defined telemetry graphs, conditional email alerts, remote logging (on Zephyr), remote diagnostic shell (also on Zephyr via a WebSocket), OTA firmware updates (and hosting), and probably other things I've forgotten about at this point.PidgeIoT is built in Rust from front to back. The frontend is WASM (Dioxus) and the core is built around Cloudflare Workers and Durable Objects (also WASM). The backend uses a managed PostgreSQL DB (for now). T
Jul 28, 2026 3:20 PM
blog.scottlogic.com
Show HN: The Rise of Open Weights – a long scroll storyhttps://blog.scottlogic.com/ceberhardt/rise-of-open-weights/Over the weekend I was discussing the latest news relating to open weights with some of my non-techie friends. After 30 mins of waffling, I realised it is quite a long and interesting story, with many twists and turns.I've always wanted to create a 'long scroll' story, but have never found a suitable story to tell. This felt like a fun weekend project to undertake.The story itself doesn't have too many scroll effects (which is probably a good thing), it was more about trying to tell a story with a certain style that wouldn't really fit on my personal blog.Would love to get your feedback on the content and / or presentation.AI Usage Notice: I used AI to help research the content, but the story itself was written entirely by me. Whereas the long scroll website was created entirely by Claude Code.As an aside, I had Claude create a simple markdown-based rendering process (inspired by remark.js) to ensure a deterministic rendering that didn't alter any of my words.
Jul 28, 2026 12:45 PM
bing.com
Microsoft’s Project Perception Gives Partners New Security AI Toolshttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d56dd43864467907828f9bbabc1c0&url=https%3A%2F%2Fwww.crn.com%2Fnews%2Fai%2F2026%2Fmicrosoft-s-project-perception-gives-partners-new-security-ai-tools&c=15676095805211113386&mkt=en-us‘Some of the partners will have security operations for customers, but they haven’t yet taken the opportunity to take the learnings from security operations and implement them as fixes in the system.
Jul 28, 2026 6:39 AM
news.ycombinator.com
Claude's code comments – too much or just enough?https://news.ycombinator.com/item?id=49078710Has anyone else noticed that Claude leaves a lot of large code comments sprinkled throughout its work?Some of my meat-based engineers have taken issue with it and directed Claude to never leave comments unless strictly necessary, but I want to know what others think. Personally, I write and read a lot of code comments so I’m confused about why these specific comments are so offensive to my fellow engineers.I have two questions: 1. Why does it do this? 2. Do you have feelings about it?Is the agent leaving breadcrumbs for other agents? Do they improve the output of future agent-driven code or is it just a stylistic choice?I don’t want to devolve into a discussion about the purpose of code comments at-large. Instead I’m wondering if there’s a purpose behind the specific comments that agents leave
Jul 28, 2026 2:55 AM
noemaai.com
Show HN: Gemma 4 26B A4B running on an iPhone 17 Pro via model paginghttps://noemaai.com/I wanted to share a use case for Noema Overfit, a model-paging system available today in the Noema app. For disclosure, I founded Noema and I am part of the team that has helped develop the system.The way it works is that the non-expert weights remain resident in memory, while the model’s routed expert weights are read from storage as needed. This makes it possible to run models that would otherwise exceed the device’s available memory, with the expected tradeoff of higher time to first token and slower generation.For a 699-token initial prompt, I measured:Prefill speed: 34.4 tokens/s Prefill time: 20.34 seconds Decode speed: 3.5 tokens/sPhotos of Overfit in action: https://noemaai.com/overfit/gemma4-iphoneThe complete answer took approximately six minutes to generate, but the answer was correct. This is not intended for interactive, low-latency chat. The more interesting use case is allowing a constrained device to run a substantially more capable model when answer quality matters mor
Jul 28, 2026 2:05 AM
github.com
Show HN: Ami – A local, open-source agent that does your busywork across appshttps://github.com/NanoNets/amiHey everybody, sharing Ami on HN today.Ami is an open source, local-first agent harness that acts as your shadow worker and copilot chat. It ships with a graph memory.Here's what Ami does on its own -- connects to apps, data, repositories, tools with your personal tokens- Learns how you do tasks (execution style, decisions, anti-patterns)- Learns how you communicate (external and internal)- maintains a universal to-do listHere's how you use Ami -1. You can execute busywork. It fetches and executes tasks autonomously in your style, asks approval before risky actions, gives deliverables, drafts replies / emails / ticket updates.2. You can execute copilot chats. Use it to ask questions, fire off ad-hoc tasks, create to-dos, update memory.Ami was built for internal use. My team found it useful, so we wanted to share it here. It's still in development stage, and we might push a more stable release soon. It constructs a context graph memory of you, with entities, relationships, feedbacks, de
Jul 27, 2026 10:55 PM
github.com
Show HN: Heddle- ATC for Your Subagentshttps://github.com/zyads/heddleA layer above git.git reports collisions after the tokens are spent; Heddle reports them before. Now I spawn as many subagents as I want to work on different features in the same repo, with my friends across different machines/networks.
Jul 27, 2026 10:42 PM
silkdrop.vercel.app
Show HN: SilkDrop – MilkDrop port for web with live editorhttps://silkdrop.vercel.app/MilkDrop visuals in a browser tab. 9,795 presets, running on your machine.SilkDrop compiles libprojectM(https://github.com/projectM-visualizer/projectm) — the open-source MilkDrop engine — to WebAssembly and renders it through WebGL 2. Point it at a browser tab or your microphone and it reacts to whatever is playing.There is no backend. The site hands you ~10 MB on the first visit and is never involved again: rendering, preset compilation, audio analysis, search across all 9,795 presets, and every byte of storage are local. Load it once and it keeps working with the network off. There is plenty of other features as well like live presets editor, built in over 9k presets with search and favorites.Made it for fun as a pet project.
Jul 27, 2026 8:40 PM
science.nasa.gov
NASA Science Soars During August Total Solar Eclipsehttps://science.nasa.gov/science-research/heliophysics/nasa-science-soars-during-august-total-solar-eclipse/Each time the Moon covers the Sun during a total solar eclipse — darkening daytime skies and briefly revealing the Sun’s ethereal outer atmosphere, the corona — it presents new opportunities to better understand our star and its influence on Earth. On Wednesday, Aug. 12, as the next total solar eclipse sweeps over Greenland, Iceland, […]
Jul 27, 2026 6:01 PM
nasa.gov
NASA Astronaut Chris Williams Returns to Earthhttps://www.nasa.gov/image-article/nasa-astronaut-chris-williams-returns-to-earth/NASA astronaut Chris Williams is all smiles in this July 26, 2026, photo taken shortly after he landed with Expedition 74 Roscosmos cosmonauts Sergey Kud-Sverchkov, and Sergei Mikaev in Kazakhstan. This was Williams’ first mission. Williams spent eight months aboard the International Space Station, where he supported a wide range of scientific investigations and technology demonstrations. He […]
Jul 27, 2026 4:48 PM
fossil.wanderinghorse.net
Show HN: JavaScript/JSON and CSS minifier lib in minimal C89https://fossil.wanderinghorse.net/r/cssminc/doc/ckout/README.mdFor those 1.7 of you who may have need of a JavaScript or JSON minifier which (A) is not based on node.js and (B) is not "poisoned" by the "do no evil" license attached to Douglas Crockford's minifier[^1] (commonly seen in various projects over the years)...My C89 CSS minifier was recently broken down into a separate JavaScript/JSON minifier. It's trivial to embed in C, "should" build fine in WASM (that's untested, though), and is pretty small: the whole combined library is about 67kb, or about 34kb for just the JS parts (much of which is docs).It minifies only spaces, not symbols, and does not offer any options for _not_ minifying (e.g. adding newlines every so often for the sake of tools which treat extremely-long-lined files as binaries) except that by default it retains the first comment in the input if it's the first non-whitespace content (those are typically license/attribution blocks) but it can be told to optionally strip those too.LLM disclaimer: the 100% hand-written CSS min
Jul 27, 2026 10:22 AM