Real problems people complain about online, pulled every morning and scored out of 100. Build, validate, or skip. How scoring works
Hey HN,Henry from Cactus here!We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2.The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges 300-700 on sub-$200 phones such as the Samsung A-Series.On the tool call and mobile device use benchmarks, Needle 2 trades wins with closest small models like LFM2.5 230M and Apple Foundation Model, at 5x to 70x smaller, both at f16 vs Needle 2 at 2bit. Needle is based on Simple Attention Networks from our paper (https://arxiv.org/abs/2607.18363).Edge AI has lately meant Macs and PCs, but that is just 1.5 billion of over 21 billion connected IoT devices in the world today, and in emerging markets most phones ship under $200, no NPU, cheap GPUs. These include budget phones, Raspberry Pis, microcontrollers, wearables, small robots like Reachy Mini, and connected home devices.A conventional transformer of Needle's width and depth spends 164 MFLOPs per token, and even one squeezed down to Needle's parameter count spends 87, Needle spends 70. Even on a high-end phone, an always-on assistant lives inside a power budget; every MFLOP is milliwatt-hours, and Needle spends 7x to 85x fewer of them per token than the smallest performant LLMs. More about the architecture in the link.When we structure intelligence for consumer devices as functions with typed parameters, the only hard part is mapping a messy sentence onto them; which function, with which values. Our research found that when framed that way, the problem needs no world knowledge and no open-ended prose
Hacker News1mo agoToolAI
Every single day of my life since I started programming back in 2020 I open my projects through the terminal/ghostty. "cd path" + "code .". Until today I don't even know how to open vscode first then open a project from there.I usually have 3-4 vscode windows open at the same time, and I like them in separate desktops (for mac users). i gotta admit, i use 8 desktops, 3-4 for vscode, 1 for arc, 1 for docker, 1 for postman etc.So my daily routine when I turned on my mac was: (1) open ghostty; (2) cd-path+code.; (3) drag the vscode window to a desired desktop; (4) get back to the ghostty desktop; (5) repeat until all is open.i was unbothered for 6y. Until i realized how inefficient i was being: (1) opening ghostty just to run one-off commands to open another app felt wrong, like why do I need a terminal if it's not a process I need to watch; (2) the need to go back to the ghostty desktop every time i wanted to open a new project and drag it to a new desktop and switch back... it's funny how we simply accept a habit and don't bother to fix iti wanted a terminal that persisted across desktops AND that was always visible so i could access it anytime AND that wasn't big because i would only run one-off commands. so I created starboard.where could I fit a terminal that was always visible and accessible? i looked at the macOS dock, i looked at the empty space to its right, and I said "that's it".starboard is a floating terminal window that's literally glued to the dock, filling the blank space to the right, tracking its height and position live. it's visible from any desktop, a real companion to the dock, and it's a persistent shell.to close the story on my routine, I now can happily jump straight to the desktop i want my project to be in and open it there through Starboard. i rarely open ghostyy, unless i really need it.here's where feedback is massively important. i built this for the way
Hacker News1mo agoToolAI
Hacker News1mo agoToolAI
Hacker News1mo agoToolAI
Hey HN, we’re Will & Johnny from ProvenMetal (https://provenmetal.com). You send us design files or specs and we give you assembled boards domestically in days.The US produced 30% of PCBs globally in 2000, now they produce 4%. Chinese manufacturers have completely dominated this space at 55% of global production.Now, the need for a domestic PCB supply chain is higher than ever before, yet the infrastructure has been dissolving over the last 2 decades. What is left is mostly small family run manufacturers (CMs) that have been operating in largely the same, labor intensive way since the early 2000’s.When you place an order through a CM it typically takes several days to receive a quote and complete design for manufacture review, and then you have to source all of the components (the hardest part) and bare boards yourself, then wait anywhere from a few days to several weeks for the assembly and testing of those boards.We started off assembling circuit boards out of a garage with prosumer grade equipment (NeoDen YY1, Glenbrook X-ray, solder paste stencils, and manual rework stations). We believed that by owning the manufacturing process, we could turn all the front of house automations inwards. But here’s the thing… manufacturing circuit boards with prosumer grade equipment out of a garage takes a huge amount of time. Suddenly we were spending 90% of our time assembling circuit boards instead of growing the business. We were completely capacity constrained, with fancy software automations that are not the binding constraint at low volumes.Then we got our heads out of the weeds, took a step back, and realized that we were trying to solve the wrong bottlenecks.These manufacturers are good at manufacturing and terrible at the front of house (quoting, DFM review, and part procurement). When you look at the full process, you realize that assembly is not the bottleneck. So we stopped trying to solve the problem that we can’t solve at this stage.We measured the b
Hacker News1mo agoToolAI
Hi HN, today I’m revealing Altar II, an ultra-low profile (4.75mm) mechanical keyboard designed to replace your Magic Keyboard. It’s the keyboard I wish Apple made.Altar II has fully mechanical switches, plus a magnetically detachable dial, haptic feedback, a red backlight, and a native macOS companion app.I’m a heavy Mac user and I feel that Apple hasn’t given its keyboards the attention they deserve, in terms of hardware, software, and features, so I set out to try and change that.It was a massive undertaking and it has taken almost 2 years to get to this stage. One of the main obstacles was fitting a reasonably sized battery into the chassis, and aggressive power profiling without messing with the user experience. Another issue was we couldn’t fit any components on the underside of the PCB because there was literally no space to do so (and there are a lot of components).Around 4 years ago I posted my first keyboard to HN (https://news.ycombinator.com/item?id=32507172). Its response blew me away and ultimately led to me quitting my job and making and selling keyboards full time. I’ve taken some of the feedback from that original post and incorporated it into Altar II (e.g. the detachable dial).I’ll be around to answer any questions. Please let me know what you think, I’d love to get some feedback.
Hacker News1mo agoToolAI
Hi HN, this is Shailendra and Karan here. We are building a fast and safe way for coding agents to debug issues live in production.When prod breaks, it lets Cursor, Claude, and others drop virtual breakpoints or probes safely in your running code, and extract the exact variable values that logs don’t have.All this saves time and effort for engineers who’d otherwise dig through logs and traces or redeploy with console.logs or print statements until they find the root cause.Here is the link to the video that explains this: https://www.youtube.com/watch?v=ivV7I--ta5cAgents write most of our code now. This shrinks the useful context engineers need to debug AI written code, a problem not helped by the limited telemetry added in the same code by AI.So when something breaks in prod, the first instinct for an engineer is to open logs or throw them to your agents. But if the line you are looking for is not there, agents will start guessing the root cause on non-existent data, forcing you to add a log, and redeploy.This analysis-inference loop of agents with existing data does not come cheap, burning a lot of tokens. And the add log, redeploy cycle is so slow and painful that it makes engineers hate on-call.Our approach lets agents capture telemetry on-demand at the exact moment and point of failure, killing the log-redeploy cycle and getting the most accurate RCA while burning fewer tokens.The obvious problem is making it work on a running service. You can't pause a live service the way you'd pause a debugger on your laptop. Getting the value out of a running process safely, without pausing a thread or slowing the host is the challenge.We are making this happen.Before this I ran engineering at a 100 member team. Then Karan and I spent three years on HyperTest which was a testing tool.At HyperTest, we turned production traffic into integration tests using OpenTelemetry. That was production instrumentation too. The hard parts of pulling real runtime st
Hacker News1mo agoToolAI
Hacker News1mo agoToolAI
Hi folks,I've noticed that a lot of people here seem exhausted by the amount of AI news on the front page. I shared hcker.news here a year ago and it has since gained a ton of filtering features, including a dedicated AI filter, so I figured I should tell more people about it.The filter works in three passes: 1. Known AI-related keywords and domains are filtered automatically. 2. An agent reviews the remaining articles and removes those it identifies as AI-related. 3. I make the call on uncertain cases. You can also exclude GitHub repos posted to HN that show signs of AI authorship. The filter checks: - Commit messages for agent attribution - Contributors for known coding agents - Repository files for agent instructions and configs No filter will be perfect, but this is as good as it gets. lmk what you think. Cheers.More links:Original Show HN: https://news.ycombinator.com/item?id=44075353If you only want AI stories: https://hcker.news/?view=frontpage&ai=includeRSS: https://hcker.news/feeds/atom?period=day&ai=excludeA github-only non-AI feed: https://hcker.news/?ai=exclude&include_domains=github.com
Hacker News1mo agoToolAI
We are Rui and Michael and we’re building EdotEnv (https://edotenv.com): self-improving RL environments from Quant Trading workflows.With all the benchmaxxing around, evals saturate and become meaningless for model comparison. Useful benchmarks should increase in difficulty as models advance. Back in our Quant jobs, Michael and I saw that the market has exactly this property: markets became more efficient as people profited from trading inefficiencies, making new profitable strategies harder to find and old ones decay over time.This makes markets an ideal, continuously evolving benchmark for LLM training. The hard part is to turn professional quant workflows into reliable training envs, as this is a very niche expertise.In our environments, we give LLMs a quant trading workflow and evaluate their performance on out-of-sample data: build predictive features/ models, design a portfolio, backtest strategies, adapt continuously to market regimes. Each step is a task with different self-built tools. For example, a predictive feature building task gives the agent cleaned market data of time period [0,T] to research ideas, a backtesting tool to test created features at time t on [0, t], an execution tool to trade strategies with the new features on [t+1, T] and a final evaluation. Our reward isolates the agent's feature building skills and yet benefits from market properties.From running SOTA models in our environments, we see that i) they seem to struggle with iterating deeply on research ideas, preferring broad shallow searches; ii) higher reasoning does not seem to increase performance and iii) agents do not understand trading, e.g. when losing money they stop trading instead of trading smarter. Check out our blogs for more details! https://edotenv.com/?tab=blogQuant workflows are essentially applied ML research, long-horizon planning and continual learning. Through our envs, we teach these transferable research skills, rather than tas
Hacker News1mo agoToolAI
Hi HN, we’re Bence and Ryan, founders of Hoplite (https://hoplite.sh). Hoplite lets you deploy coding agents in the cloud, with a suite of tools that makes it incredibly easy to QA features. During onboarding, we port over your local setup - sessions, memories, MCP servers, and get your projects ready to run in the cloud.Here’s a demo: https://youtu.be/bnyktZ_9pjEWe got here after pivoting away from the idea we applied to YC with; AI for retail investing. It ultimately wasn’t a product that we ourselves would use, nor served a customer base that we felt connected to. In reflecting on what we really wanted to do, we realised that we loved talking to founders and developers, and were really opinionated about the specific area of cloud agents. We tried out all the existing solutions, and didn’t find one that A) took good advantage of being in the cloud, and B) was performant and felt good to use.We’re building a product that we feel reflects what mainstream development will look like in 6-12 months. As models improve, developers will end up reviewing less and less code, and will instead focus on reviewing the product output. That means evaluating new user flows, visually verifying that new features look good, that the API works as expected, that the CLI works on Windows, etc. And doing it while running hundreds of agents concurrently.On the agent side, we’ve created a custom harness. We spent a lot of time deciding on whether we should use an off the shelf solution like Codex/Claude Code, but ultimately wanted the independence and freedom that came with building it in house. It also means that we can test out completely new features without relying on Anthropic and OpenAI to catch up.Everything is hosted on AWS, with the exception of: Temporal for durable workflows, Modal for sandboxes, and Planetscale for our database. Our infra decisions were driven by a strong belief that agents are becoming a tier 0 piece of infrastructure, and they ne
Hacker News1mo agoToolAI
Users of Grok Imagine face excessive moderation and blocks on fully clothed, non-explicit prompts, despite seeing unmoderated explicit content elsewhere on the platform.
X1mo agoToolAI
Hi HN! We’re Akilan and Miguel, the creators of MarbleOS.The inspiration for Marble comes from the GUI work at Xerox PARC, the 1984 Macintosh, and later NeXTSTEP, which became the foundation for Mac OS X. Before GUIs, interacting with a computer was limited to strange terminal commands:C:\> DIRC:\> COPY FILE.TXT A:You had to remember the command, syntax, paths, and parameters.The GUI made those capabilities visible. Instead of remembering commands, you could point at files, drag them, click buttons, and select actions from menus. It didn't necessarily make entirely new things possible; it just made existing capabilities much easier to understand and use. We feel like AI is still somewhere around this command-line stage.Even though the strict syntax has been replaced with natural language, the interaction can still be quite stiff and depend on heavily recall. Tools like Claude Cowork still look surprisingly terminal-like: /skill-name [param1] [param2]. The parameters are written in natural language, but the user still needs to know the capabilities that exist, and how to invoke it. Command-line flags and arguments have just been replaced with tools, skills, and context.ChatGPT works very well for asking questions, but we aren't convinced that it's the final interface for delegating work across multiple agents. A blank text box and a list of chat threads feels limiting. An agent's capabilities are mostly invisible. With Marble, we're exploring an interface that treats AI more like a workspace than another chat app. Each delegated task becomes a card. Multiple jobs can sit next to each other and run at once. Files, tools, and finished artifacts are visible at once. Before a task runs, Marble also shows which tools it expects to use. The basic idea is that the user should not have to hold the entire structure of the task in their head. And the result should be something directly usable, like a spreadsheet, PowerPoint, or other file, r
Hacker News1mo agoToolAI
for decades, i have been concerned that the internet was being built around non-fiction, so i built something to show how we could celebrate great fiction on the web.i started with an amazing library from the public domain, and i also added tools to add your own storiessuper interested to hear any feedback, and if you read anything good!
Hacker News1mo agoToolAI
Just launched my app, VimGolf.ai! As users complete the levels on the map, I introduce progressively more complicated concepts — it’s aimed at people who have never used Vim. Any feedback would be much appreciated!
Hacker News1mo agoToolAI
Hi 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 experiments are described in the GitHub repo.It currently generates 5–6 tok/s on an 8 GB M2 MacBook Air and 31–35 tok/s on an M5 MacBook Pro.I also added an experimental OpenAI-compatible local server. It supports streaming and tool calls, and reuses one prompt prefix from the KV cache.Try it! The Mac app is easy to install. On the first run, it will download 15 GB of weights from Hugging Face. The model is surprisingly capable.I would love any kind of feedback!
Hacker News1mo agoToolAI
I have been pushing up to 90 commits a day on a MacBook Air via 4-5 parallel agents. As you can imagine when all the agents try to build, test and run dev servers on an 8GB machine it is the fast lane to a force quit and restart. I also did not want to pay the CI minutes on 90 pushes a day.So I designed a local merge queue to have all commits land one at a time and fully tested. Hopefully this helps other folks with more modest machines. Appreciate any feedback.
Hacker News1mo agoToolAI
I use separate Claude Code accounts for work and personal projects. Having to log out and go through the login flow every time I switched accounts became annoying, so I built a small CLI to solve it.The commands are intentionally simple:claude account add myworkaccount claude account add mypersonalaccount claude account use myworkaccount claude account currentAfter switching, Claude Code works normally:claude claude "fix this bug in main.py" Repository: https://github.com/hamzarehmandeveloper/claude-accountGive it a try
Hacker News1mo agoToolAI
Hi HN, we're Marinos and Hudson, founders of Prized (https://prized.dev)! Prized lets non-engineer employees describe the internal tool they need and get a full-stack app, wired to their company’s data and deployed behind the company’s sign-in, without them ever juggling API keys or connectors.Here's a demo: https://www.youtube.com/watch?v=730MuYOfZTYThe way Prized provides security is by limiting what the agent can reach at the network layer and by keeping credentials out of the sandbox entirely. The sandbox never holds any keys or connector secrets, it only uses scoped session tokens that are stored as opaque placeholders. The real values are swapped into the request headers on our egress proxy. When production data is connected, the sandbox's network policy is set to deny by default so the only path out is via the proxy. Any call the agent makes to an external connector is reviewed by an LLM judge to prevent dangerous operations.Prized is meant for the internal workflows that start as notebooks or spreadsheets but never become real tools because engineering has more important things to work on. One customer’s data scientist pasted in his personal fraud-detection notebook with hardcoded thresholds and all. After a few prompts, it became a published risk console connected to the company’s data with those thresholds turned into UI controls. Earlier today, we got off a call with them and most of their company is using it.To do this, you need to give people the freedom to build without having unaudited access to company systems. We allow admins to scope data to specific users or teams and data access is recorded in an audit log. Each tool is built with its own Postgres schema and role, with queries running via an authenticated SQL gateway as that role.We think Prized sits between products like Lovable and Retool. Lovable makes it easy to generate and host software, but it isn’t designed around distribution with permissions. Reto
Hacker News1mo agoToolAI
Hey HN, I’m Shreyash from Feyn. We help companies build custom models from their data.Today, we’re releasing FeyNoBg, an automatic background removal model. Alongside it, we're open-sourcing NoBg, the Python library we built to train and run it.Try the model here: https://huggingface.co/spaces/feyninc/feynobg. Check out the library here: https://github.com/feyninc/nobgSome sample outputs:(1) Soccer Freekick: https://drive.google.com/file/d/1MZkAGLwbhNVOZ0Oi7XvpCfSEu9Q...(2) Hair in wind: https://drive.google.com/file/d/1Odc2m0XMVH9uZtvI_KjaRbXzhLL...(3) Bicycle with visible spokes: https://drive.google.com/file/d/1h99ahjfrtS1MFQJJgiKE2fuM3HZ...(4) Live Demo video: https://youtu.be/b1heHPvY8BMBackground removal separates an image's subject from its surrounding. We've all tried it at some point. Often it is to reuse the subject in a different artifact. Nowadays, it is common to make chat stickers out of it. It is one of the most common but under-appreciated uses of AI. It is also surprisingly complex. Models can be easily confused by camouflage, motion blur, or fine structures like hair.The task requires two skills. First, a model has to identify the foreground. Second, it has to trace the foreground’s boundary and estimate an opacity value for each pixel. Generally, these skills are taught with different datasets. That creates a failure point. A poor training mix can improve one skill at the expense of the other. We saw this in our controlled evaluation. A training run with just the MaskFactory dataset improved on the CAMO benchmark but regressed on DIS5K.For FeyNoBg, we took an interpretability-first approach to training. We first studied how BiRefNet’s stages contribute to finding the foreground and reconstructing its boundary. We discovered that the third stage of it's feature extractor holds a lot of informatio
Hacker News1mo agoToolAI