26,648 results for men

news.ycombinator.com
Ask HN: Should I Combine Global Knowledge, Internet Search, and User RAGhttps://news.ycombinator.com/item?id=49059969I'm building a SaaS platform in Sri Lanka that handles documents and other sensitive data.Each user can upload their own documents and information, and the platform uses RAG to answer questions based on that user's data. That part makes sense to me.My main concern is what happens when the user hasn't uploaded enough information. I still want the LLM to provide accurate answers using reliable information from the internet (or from a curated knowledge base), with proper citations.These are the two architectures I'm considering:Option 1:Base LLM (OpenAI/Anthropic via Azure AI Foundry or Amazon Bedrock) ↓ Platform RAG (global knowledge base managed by us) ↓ User-specific RAG In this approach, we maintain a global knowledge base that we (the platform admins) curate and update. Every user can access this shared knowledge, while their own uploaded documents are searched through their personal RAG.Option 2:Open-source LLM ↓ Fine-tuned on Sri Lankan/domain-specific data ↓ User-specific RAG Here
Jul 26, 2026 4:55 PM
kypaku.github.io
Show HN: Scanner for cellular automata rule spacehttps://kypaku.github.io/pages/ca2026/Many years ago I was interested in cellular automata, inspired by Stephen Wolfram's work. I dug into elementary CA and totalistic 3-state rules. But when I went to more states it was getting boring. Almost all of them are dead or chaotic or something that everybody has already seen. But somewhere inside there could be another 110, and nobody looked.Now that we have AI coding, I decided I can build something in that area, and an app for scanning rules was the obvious thing. I implemented an explorer with a bunch of filters: gliders, chaos, large structures, and custom search. The last one works backwards — you draw a pattern and a rule passes only if this pattern never occurs. I could not define "interesting", but it was easy to say what disqualifies a rule. It also renders diagrams as big PNG.Demo: https://kypaku.github.io/pages/ca2026/ Source: https://github.com/Kypaku/ca2026
Jul 26, 2026 4:39 PM
github.com
Show HN: I built a transparent terminal wrapper for unobtrusive AIhttps://github.com/emosenkis/terminaiI wanted a way to use my existing AI subscriptions as a terminal assistant to analyze shell output or help me with a complicated awk incantation from time to time without getting in my way the rest of the time. I couldn't find anything existing that does this do I built it! I used pvolok's excellent mprocs as a starting point and leaned heavily on Claude and Codex for coding all but the gnarliest parts while providing extensive guidance on how it should be implemented.I know that coding agents can run shell commands already but that puts the AI in the driver's seat. Terminai let's me be the driver, with codex or claude (or any other agent) sitting quietly in the backseat until I want its help.The hardest part was making it truly transparent. Most TUIs that embed a terminal break the terminal emulator's native scrollback and frequently also break copying to the clipboard. I worked hard to make sure that that both of these work exactly as they would without terminai. Along the way I had
Jul 26, 2026 3:05 PM
getsnack.app
Show HN: Exercise snack reminders for iOS/Android, no server, one-time purchasehttps://getsnack.appI built Snack App for my wife. She works from home, and every night, she would feel stiff and creaky after sitting all day, so she decided that she had to be less sedentary. Her idea was to break up her day with "exercise snacks" - formally defined as "isolated ≤1-min bouts of vigorous exercise performed periodically throughout the day" (https://pubmed.ncbi.nlm.nih.gov/34669625/) but really any kind of movement (a couple flights of stairs, a 5-minute walk, or a set of air squats) is better than nothing.So I wrote her an app that lets her set a goal and pick a reminder schedule. The app sends her notifications; she logs her snacks in the app; and every day she meets her goal extends her streak.Speaking of streaks, I forgot to do my Wordle one day and, to my horror, broke a 171-day streak, so I made Snack App's streaks flexible. You can pause your streak anytime, and it'll resume when you start logging snacks again. There's also a built-in grace period, just in case you forget. The whole
Jul 26, 2026 1:56 PM
gnoseed.com
Show HN: I built Gnoseed – free flahscards to learn K8s, DevOps, AI and morehttps://gnoseed.comHi everybody,I created my personal weekend project Gnoseed to learn more about Kubernetes with spaced repetition. Very soon it evolved into a small library of various topics about modern tech stack - AI/LLMs, DevOps, Security, and anything I found interesting. Currently it has 47 decks and ~4000 questions.A bit more about the process: I create decks from the open documentations or free papers I found here or there. The decks are built with a custom made pipeline of skills to extract the data, select questions, creating distractors and verification. I'm curating the entire process.To keep cards up-to-date I have a weekly automated routine to check release notes of selected topics and curating new cards.From the beginning I also realized just the random algorithm to show cards is not very efficient, so I implemented two spaced-repetition algorithms SM-2 and FSRS.Would you like to check it and give me some feedback? It's free, no sign-up needed
Jul 26, 2026 1:54 PM
news.ycombinator.com
Ask HN: Something fishy is going on with Google feed on my Androidhttps://news.ycombinator.com/item?id=49057483Hello folks. I want to know if this is something only happening to me.I noticed on the past few days something weird going on with Google feed on Android (when you swipe left from home main screen)I started receiving deceiving news with some "shocking" headlines like: "Contagious cancer found in fresh water fish". Android says it was published by "Forbes" but the icon is weird. That caught my attention and I used share button to check the URL and for my suprise the target link is "discover.aigc/?id=.........." (I removed the number just in case)I used some search engines and I found no relevant related domain. Pinging the domain also returns nothing.i tried accessing it from an isolated environment and the domain seems really offline now.I'm sure I saw other suspicious links, some had related titles some did not.It does look like a clickbait based campaign for something "fishy" (pun intended).I should have taken notes of the other news and domains which i crossed paths with but I did n
Jul 26, 2026 12:32 PM
github.com
Show HN: Lemmings in HTML in Canvashttps://github.com/VorticonCmdr/lemmingsFor fun I decided I wanted to reimplement Lemmings using HTML-in-cavas and a DOM-based Entity-Component-System. I made heavy use of Claude Code (using only Sonnet 5 High) and I used everything already published on this topic. I used already existing HTML ports (like DHTML Lemmings, LemmingsJS and Lemmings.ts) to kickstart development.Even though I was able to add some new stuff like EGA, CGA and Tandy as well as Adlib, Tandy and PC Speaker that wasn't documented (or I didn't find it). I, or rather Sonnet, was able to do so because I hooked it up to a Ghidra MCP and a DOSBox MCP.At some point feature creep set in so I am guessing there are still a few things not working as intended. I'd love to hear all about it.
Jul 26, 2026 12:23 PM
philomatics.com
Show HN: Organic cell simulation with infinite zoom (100k cells, raw WebGL)https://philomatics.com/cell-game-source/Made this little cell simulation with an infinite zoom mechanic in raw WebGL. Renders up to 100k cells on my laptop (M1 MBA).It's a single HTML file, not minified, without dependencies, so you can just inspect the source code. It's just raw JS and WebGL shaders. No assets, everything is procedurally generated.For each cell, I draw a single quad using instanced rendering. To create a nice organic wobble, I use SDF (Signed Distance Fields) to draw a circle (or another base shape), which is then distorted by adjusting the radius using multiple overlapping sine waves. I explain this visually here. [1]To get the cells to stretch dynamically based on their velocity, I use a technique called Domain Warping. Basically in the fragment shader, the fragmentPosition is adjusted inwards/outwards depending on the current velocity, which squashes/stretches the entire coordinate system of the quad in either direction. Gives it a very organic look. I explain this in more detail here. [2]Physics are als
Jul 26, 2026 12:04 PM
news.ycombinator.com
ASK HN: Why has technology become so unreliable?https://news.ycombinator.com/item?id=49056900Even though I'm a software engineer with long experience with serious projects, I'm become so disillusioned with the whole thing nowadays.Forget the ridiculous AI hype, forget about the lack of ethics and a moral compass; What's getting to me right now is the unreliability of it all. It seems like everything is held together with duct tape, and the tape is falling off.Just in the last few months, all these have happened:- My Macbook Air M1 has been acting up, refusing to charge in a stable manner. The audio jack is dead because apparently it is linked to the charging circuit. The screen sometimes shows a tint on a whim. Now it is dead and not charging at all. Granted, it has otherwise been a fine machine so far, and I've knocked it over a few times, but still. Even my old trusty 2014 Macbook Pro is still going strong, and I've abused that one much more.- I also can't use my old Macbook Pro well, because the replacement Transcend 256GB SSD I put in it around 2020 is now failing, causing
Jul 26, 2026 11:13 AM
termic.dev
Show HN: Termic – desktop app for running CLI coding agents (Claude Code, Codex)https://termic.dev/So, https://termic.dev open source desktop app for managing claude code / codex / agt / etc coding sessions. main checkout or worktrees.Here's the story: a couple of months ago Anthropic announced pricing changes over the programatic use of Claude code (Agent SDK + including -p), i suspected tools like Conductor will stop working.Started building a clone myself, which uses the CLI binaries in PTY terminals as first class citizens, not custom agent sdk. This is the most bulletproof way of always being up to date with latest improvements of the cli agents + and never have this pricing issue again for 3rd party tools, and you get the flexibility of using any CLI agent.That's how https://termic.dev was launched.Also has some pain points with conductor which i adressed in termic: launch in main checkout (doesnt force you into a worktree), sandboxing of agents (filesystem + network), multi-repo projects and agents, and bring-your-own agent CLI.Stack is simple, and i;'ve been obsessing over p
Jul 26, 2026 9:36 AM
github.com
Show HN: ScreenFocus – keyboard focus follows the pointer across Mac displayshttps://github.com/therohitdas/ScreenFocusI have a multi-monitor setup and spend a lot of time typing in Codex and other AI agents. When I move to another monitor to run a shortcut, focus often stays in Codex or another text app, so the shortcut runs in the wrong place.I asked ChatGPT if there was a macOS setting or free app for this. It couldn't find one that did exactly what I wanted, so I asked it to prototype one. It built a working version in one go.I then used Codex to add sensible defaults, focus protection, an optional display border, packaging, documentation, an MPL 2.0 license, a logo, and screenshots.ScreenFocus is the result: keyboard focus follows the pointer across displays while global shortcuts keep working.This was my first native Mac app, and I used GPT-5.6 Sol with High reasoning. If you check out the README, help docs, and other Markdown files, you might be surprised by how complete it feels.I also added Launch at Login, full-border and corner-marker styles, custom colors, automatic pausing when only one mo
Jul 26, 2026 7:07 AM
news.ycombinator.com
Ask HN: What's expected from an engineering manager or director now?https://news.ycombinator.com/item?id=49051762I just joined a 30-person AI startup in the Bay Area as Director of Engineering.Two non-coding founders (CEO and CTO) and a flat team. We have 2 PMs, 20 engineers. Every engineer is senior or staff level. They use Claude Code and AI agents etc quite well.What is my job in an environment like this? According to the CTO, I'm to improve the velocity and productivity while also balancing quality.Also, are there any Bay Area meetups or peer groups where engineering leaders are actually talking about this specific setup?
Jul 25, 2026 9:26 PM
news.ycombinator.com
Ask HN: HotPin – lossless 120B MoE inference on 24GB RAM (CPU, 50 loc)https://news.ycombinator.com/item?id=49050356I'm a mechatronics designer with a background in control systems, robotics, PCB design, and embedded hardware. I design physical systems: motors, sensors, microcontrollers, and real-time control loops.I applied this design thinking to LLM memory management – and it worked.HotPin is a set of patches for llama.cpp that runs 30B–120B Mixture of Experts (MoE) models on far less RAM than their disk footprint, with bit-identical (lossless) output.Tested on an AMD Ryzen AI 9 HX 370 (Zen5, AVX512), 23.6GB LPDDR5X, NVMe >1GB/s, CPU-only.Results: | Model | Disk | Min RAM | Savings | tok/s | |-------|------|---------|---------|-------| | gpt-oss:120b | 58.5GB | 19.1GB | -67% | 3.84 | | qwen3:30b-a3b | 18.0GB | 10.4GB | -42% | 19.7 | | gemma4:26b-a4b | 16.2GB | 10.6GB | -35% | 11.5 | | GLM-4.7-Flash | 19.0GB | 13.3GB | -30% | 12.4 |Output is SHA-256 bit-identical to full-RAM runs. Verified.How it works (~50 lines of C++ in llama.cpp): 1. Profile MoE expert routing frequencies. 2. mmap the entire m
Jul 25, 2026 6:41 PM
spacepolicyonline.com
Successful Flight for Starship IFT-13https://spacepolicyonline.com/news/successful-flight-for-starship-ift-13/The 13th Integrated Flight Test of SpaceX’s Starship system, IFT-13, was mostly a success. One week after the launch was scrubbed at the last moment because of engine failures, and […]
Jul 25, 2026 5:36 PM
news.ycombinator.com
The Cost of Seamlessnesshttps://news.ycombinator.com/item?id=49048352The Hidden Cost of Frictionless Technology Taotuner — 2026I recently asked an AI assistant for restaurant recommendations. It listed five places, complete with ratings, distance, and popular dishes. I picked one, went, had a fine meal. Later, when a friend asked what I'd eaten, I couldn't remember the name of the restaurant.The experience wasn't bad. It was frictionless. And it left almost no trace.We've built tools that eliminate the gap between want and satisfaction, question and answer, curiosity and resolution. We call it seamlessness. We've assumed it's progress. But the gap wasn't inefficiency — it was where memory, judgment, and understanding used to form.What if removing it is making us less capable?The Pattern This isn't just about restaurant recommendations. The same dynamic shows up across AI, therapy, cities, and education.Take AI assistants. The industry evaluates them on speed, accuracy, and satisfaction. Those metrics capture whether the tool works. They don't capture wh
Jul 25, 2026 3:24 PM
github.com
Show HN: Cubic Doggo 06R: 12-DOF 4-Legged Robot with IMUhttps://github.com/SphericalCowww/CubicDoggo_06RThis repo is an upgrade to my previous project, Cubic Doggo: https://github.com/SphericalCowww/CubicDoggoIt has since implemented an IMU and several mechanical reinforcements. The next step will be simulation and reinforcement learning.
Jul 25, 2026 2:48 PM
cygnus.run
Show HN: Cygnus – A fast, lightweight self-hostable serverless runtime and PaaShttps://cygnus.runI built Cygnus because of a long standing frustration with the compromises needed to be made when choosing a deployment option for web applications.The ecosystem is fragmented into a few distinct camps, each sacrificing user experience or runtime compatibility to balance isolation, startup latency, and their own profit margins.Docker: Heavier and slower because it has to supervise more than web apps. Paying overhead you don't need. MicroVM's: Good isolation, but huge maintenance surface area and substantial overhead. Great for untrusted code, overkill for your own apps. Workerd: Tries to dance around hardware isolation by enforcing an in-process V8 isolate model. Neutered runtime to prevent arbitrary syscall execution. Personally I still find the tradeoff worthwhile for their edge footprint, but I still find myself complaining about it from time to time. Vercel/Managed serverless: Delivers a great developer experience, uses microVM's which is good for compatibility but has a business m
Jul 25, 2026 12:20 PM
algohol.net
Show HN: I made an algorithmic arena game with physics in 3Dhttps://algohol.net/Hey,I finally dare to raise the curtains to my 3D physics-modeled arena game, where players' algorithms fight and perform tasks.The arena treats algorithms as a black box, meaning that the implementation of algorithms is in the free hands of the players via the Algohol API. Algohol is roughly in the same context as Screeps or Battlesnake, but brings its own dimension to the genre.At the moment the only world is energy economy based SciFI, but there will be Terra (traditional kinetic combat) and Space. I'm most excited about Space because of its multidimensional control and strategic challenges, and I will likely add it next.In addition to open public arenas, competitions takes place in blind and sealed arena modes. The sealed mode is the actual competitive standard and is coming later. Players also have private freely buildable arenas where they can play and develop their algorithms by themselves and with friends. At first, everyone has the opportunity to create two algorithms, but the
Jul 25, 2026 12:18 PM
bing.com
Costco Just Settled a Major Class Action Lawsuit—and You Could Get More Than $500 Backhttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8cf1deb6bf4566945c699c18ad2320&url=https%3A%2F%2Fwww.yahoo.com%2Fnews%2Fus%2Farticles%2Fcostco-just-settled-major-class-180000268.html&c=16677459591476326955&mkt=en-usAdd Yahoo as a preferred source to see more of our stories on Google. Costco has reached a proposed $14 million settlement over allegations of misleading marketing emails sent to Washington residents.
Jul 25, 2026 10:59 AM