📦 rpatrik96/hallmark

Hallmark: The Citation Checker

Hallmark is the citation checker that finally helps you spot fake references in machine learning papers. After a recent scandal exposed hundreds of invented citations slipping through peer review, this project offers a standardized way to test if tools can actually distinguish real from fake. It contains over two thousand annotated examples sorted by how tricky they are to detect. You can run your own detection methods against it or compare them against top AI models and database tools to see what really works. This gives researchers a clear, fair yardstick to measure accuracy instead of just guessing. It is the perfect reference for anyone tired of trusting citations blindly.

🆔 @hackernewsgithubprojects

📦 openlabs-so/openanalytics

OpenAnalytics: Privacy-First Web Analytics

OpenAnalytics gives you the rare ability to track revenue from your Stripe account alongside every page view, all without using cookies or storing IP addresses. The system protects privacy by turning visitor identities into daily rotating hashes that expire every night, so you never build long-term profiles. You can watch exactly who is online right now with city-level locations resolved locally on your own server, ensuring no data ever leaves your hardware. The tracker script is kept tiny to avoid slowing down websites, and the whole stack is designed for easy self-hosting.

📰 https://news.ycombinator.com/item?id=49302597

🆔 @hackernewsgithubprojects

📦 snowteam2023/cognivia

Cognivia AI Therapy

Cognivia lets you talk with an AI that actually understands cognitive behavioral therapy. It listens to your worries, spots your negative thought patterns, and gives you rational responses to help you feel better. Think of it as a supportive chatbot trained by real therapists to guide you through tough feelings instead of just giving generic advice. It is a genuine tool for mental health support that helps you reframe your thoughts with care. Go check it out and see if it helps.

🆔 @hackernewsgithubprojects

📦 hrtan/fastmix

FastMix: Auto-Tune Data for LLMs

FastMix lets you automatically find the perfect data mix for training large language models without wasting weeks on trial-and-error. Instead of manually guessing which sources matter most, this tool trains a small helper model that watches how each data type affects performance. It then adjusts the proportions on the fly, boosting useful content and dialing back noise in real time. You get two clear paths: one that optimizes for general validation accuracy and another that fine-tunes for specific tasks like math or science. The code uses efficient kernels to keep training fast, and you can track progress through standard logs.

🆔 @hackernewsgithubprojects

📦 xufeiqiong/gama-bench

gama-bench

Large language models judge the exact same bad behavior differently depending entirely on whether a man or a woman did it. A new benchmark called gama-bench tests this by feeding top artificial intelligence models over a thousand mirrored conflict scenarios, changing only the gender of the person at fault. The results are fascinating. When a male character steps out of line, the models tend to write harsher, more punitive, and blaming responses. But when a female character does the exact same thing, the models suddenly shift to a soft, therapeutic, and highly empathetic tone. It is a brilliant way to expose hidden double standards in AI, showing us that our...

🆔 @hackernewsgithubprojects

📦 am-will/gooey-pi

GooeyPi: One Desktop for All Your Coding Agents

GooeyPi is the desktop workspace that lets you run Pi, OMP, and Prime Agent in the same window without switching apps. It solves the headache of juggling different coding tools by giving each agent its own familiar interface while keeping everything in one place. You can switch between workspaces using a simple top-left button, and each agent keeps its own sessions and models safely separate. This means you can work on multiple projects or test different AI models simultaneously without ever closing a window. It is perfect if you want the power of specialized agents but hate the clutter of dozens of open applications.

🆔 @hackernewsgithubprojects

📦 arcademakersources/arcademaker

ArcadeMaker 2D Game Engine

ArcadeMaker is the 2D game engine that actually lets you build cross-platform games using your own custom programming language. Inspired by the classic GameMaker workflow, it bundles a complete visual workspace, a sprite designer, and a custom scripting language called Exp into one lightweight tool. Instead of wrestling with complex setups, you can design rooms, paint sprites, and write your game logic inside a single program. Since it uses MonoGame under the hood, you can compile your creations for desktop, mobile, and even consoles. Grab your coffee and see how fun building games from scratch can be.

📰 https://news.ycombinator.com/item?id=49258085

🆔 @hackernewsgithubprojects

📦 trae1oung/swe-touch

SWE-Touch: Testing Code Agents with Human Intervention

SWE-Touch lets you evaluate how well coding assistants handle real-world changes by letting humans modify code during tests. It provides a benchmark where researchers can see how agents adapt when users actually touch and alter the codebase, rather than working in isolation. This approach reveals whether AI tools can truly collaborate with developers by adjusting to new requirements and fixes on the fly. It solves the problem of static testing that ignores the messy reality of human collaboration. You get to watch agents demonstrate resilience and flexibility as they react to live edits.

🆔 @hackernewsgithubprojects

📦 pkucs-ltf/refcaptioner

RefCaptioner: Grounded Video Descriptions

RefCaptioner is the open-source model that turns video footage into precise, image-anchored stories. Instead of just describing what happens on screen, it looks at a set of reference photos and builds a description that explicitly ties specific visual details to those pictures. It cleverly ignores irrelevant reference images that do not match the video content, ensuring every mention is accurate. This makes it incredibly useful for checking if a caption actually matches the visuals. You can even train the model yourself to get even better results. It is a smart step forward for making video descriptions truly factual and grounded.

🆔 @hackernewsgithubprojects

📦 sonhm3029/freqortho-sr

FreqOrtho-SR

FreqOrtho-SR is a one-step diffusion framework for real-world image super-resolution that fixes the common problem of blurry details by using two specialized experts. It works by routing low-quality inputs to pixel-level experts based on frequency analysis, ensuring sharp restoration. Then it separates the visual fidelity from artistic details by projecting gradients onto independent spaces, so the model doesn't confuse structure with style. This orthogonal expert learning approach means you get a single forward pass that delivers both accurate pixels and rich perceptual quality without the usual artifacts.

🆔 @hackernewsgithubprojects

📦 jasper-yan/agentstream

AgentStream: Testing AI That Learns On The Fly

AgentStream is the research framework that finally tests how well self-evolving AI agents perform when tasks come in a continuous stream rather than in isolated, one-off tests. The repository bundles the Exgentic evaluation engine to orchestrate benchmarks like customer support simulations and software debugging into sequential or interleaved flows. This allows researchers to see if an AI actually improves from its own mistakes over time, or if it just memorizes a single problem. It is a crucial step toward understanding if these agents can handle the messy, real world.

🆔 @hackernewsgithubprojects

📦 kbyrski/flare

FlaRe

FlaRe is a mysterious project that hints at something called Floating Radiance Networks, though the code itself is barely there right now. The repository contains almost nothing beyond a single readme file and a promise that it is launching soon. It appears to be a placeholder or an early teaser for a future network-based tool, but since there is no actual source code or functional software to test, it does not yet let you do anything concrete. It is essentially a digital placeholder waiting for the real development to begin, leaving viewers curious about what might eventually emerge from this sparse collection of files.

🆔 @hackernewsgithubprojects

📦 amelieheinrich/agfx

Agfx Cross Platform Graphics Library

Build and ship beautiful high-performance games across multiple operating systems using a single, unified graphics library. Designed specifically to help indie developers target Direct Three D Twelve, Metal Four, and Vulkan without drowning in platform-specific code, agfx is a lightweight wrapper that keeps things incredibly simple. Instead of dealing with complex descriptor sets or root signatures, it uses a bindless first approach where you reach GPU resources using plain integer handles. You get immediate access to modern features like inline raytracing, mesh shaders, and GPU-driven rendering while writing your shaders just once.

📰 https://news.ycombinator.com/item?id=49304558

🆔 @hackernewsgithubprojects

📦 don-chad/ninfer-3090

NInfer 3090

NInfer 3090 is the local inference engine that finally runs massive twenty-seven billion parameter artificial intelligence models directly on a single consumer graphics card. By writing highly optimized native hardware code from scratch, this system bypasses heavy frameworks to squeeze every drop of performance out of a standard graphics processing unit. It introduces custom memory layouts and clever batching techniques to let you serve complex reasoning models on your own machine without relying on expensive cloud servers. If you want to host private, incredibly fast language models on standard hardware, this project shows just how far consumer hardware can be pushed.

🆔 @hackernewsgithubprojects

📦 darksidewalker/dasiwa-comfyui-workflows

Dasiwa ComfyUI Workflows

Dasiwa ComfyUI Workflows is the workflow library that keeps your AI video generation setup perfectly organized through changing models. If you use ComfyUI, you know how fast things break when new models drop. This collection acts as a living archive of working setups for heavy hitters like Wan 2.2 and Lightricks LTX. Instead of rebuilding your complex image and video pipelines from scratch every single time, you can grab these pre-configured snapshots. It is a massive timesaver that lets you skip the tedious troubleshooting and get straight to making cool art.

🆔 @hackernewsgithubprojects

📦 zfu006/ssg

SSG: Self-Guided Image Generation

SSG is the frozen diffusion model that teaches itself to make better images using its own mistakes. Most AI image generators need a heavy external helper to guide their creativity, but SG proves that a fixed model can actually look at its own half-finished samples and use them to self-correct. This repository provides the official code and pretrained weights for that Synthetic Self-Guidance technique on pixel-based diffusion models. You can clone the repo, set up the Python environment, and download ready-to-use checkpoints. It lets you generate high-quality pictures directly from the command line without needing complex training runs.

🆔 @hackernewsgithubprojects

📦 compactifai/full-chunked-kl-loss

Full Chunked KL Loss: Shrinking AI Memory Needs

The full chunked KL loss kernel that slashes video memory usage by processing large language models in bite-sized pieces. This project offers a practical benchmark showing how chunking the training process avoids storing massive intermediate data, which traditionally crashes your GPU when sequences get long. Instead of computing everything at once, it processes small segments forward and backward, keeping your memory footprint tiny even with huge context windows. It is a clever engineering trick that lets you train complex AI models on smaller hardware without sacrificing accuracy.

📰 https://news.ycombinator.com/item?id=49257181

🆔 @hackernewsgithubprojects

📦 longlongaaago/ultradiffedit

UltraDiffEdit: Edit Images Up to 8K Resolution

UltraDiffEdit lets you edit massive, ultra-high-resolution photos without needing to train a new model from scratch. It takes standard AI tools and scales them up to handle images up to eight thousand pixels wide. The secret is a smart multiscale process that breaks the huge picture into smaller pieces, edits them one by one, and then blends everything back together seamlessly. This approach avoids the blurry seams and weird artifacts that usually ruin high-resolution edits. You can swap objects, change colors, or add details using simple text prompts or shape guides.

🆔 @hackernewsgithubprojects

📦 zhaobo253-crypto/attrguide

AttrGuide Ultrasound Classifier

AttrGuide injects clinical attribute priors into ultrasound classifiers. Instead of relying solely on raw visual texture, it uses a parallel branch to match image features with medical concepts like shape and margin. This dual-branch approach improves accuracy and provides interpretable evidence for why a diagnosis was made. The project offers a lightweight, plug-and-play framework that guides deep learning models with human-readable semantic cues, making computer-aided diagnosis more robust and transparent for medical imaging tasks.

🆔 @hackernewsgithubprojects

📦 opengeos/anymap-ts

anymap-ts

Build interactive custom maps inside Python notebooks using your favorite frontend mapping tools. This tool lets you render geographic data using MapLibre, Leaflet, Cesium, or even three-dimensional point clouds with Potree, all written in TypeScript under the hood. It sets up a bidirectional bridge so your Python data and your visual map stay perfectly in sync as you zoom, pan, or click. You can draw shapes, manage complex layers, and instantly export the final creation to a standalone HTML file. It is the ultimate superpower for anyone who loves python data but wants gorgeous, responsive web maps.

🆔 @hackernewsgithubprojects