📦 tencent/contextpilot

ContextPilot: The AI That Cleans Its Own Desk

Make your AI agent smarter by teaching it to clean up its own memory. ContextPilot solves the problem of agents getting overwhelmed by too much information. Instead of letting a model drown in endless conversation logs, this tool trains it to actively organize, summarize, and forget useless details. It uses a clever reinforcement learning trick to figure out exactly which pieces of info matter, so the agent keeps a compact, focused workspace. This leads to better answers on long, complex tasks without needing massive amounts of raw data. It is a smart way to build agents that stay sharp, even when the conversation gets really long.

🆔 @hackernewsgithubprojects

📦 amap-ml/looparena

LoopArena: Testing AI Managers

LoopArena is the benchmark that finally measures which AI models make the best project managers. It tests a model acting as a controller guiding a separate coding agent through complex tasks, revealing who can actually steer the work. The twist is that the best model still succeeds on less than twenty-five percent of full tasks, proving that guiding AI is far harder than letting it code alone. If you care about AI reliability, this gap is exactly where the next breakthroughs will happen.

🆔 @hackernewsgithubprojects

📦 kingbootoshi/cartographer

How Cartographer Maps Your Codebase

Cartographer turns any codebase into a local, navigable map that AI agents can actually read. Instead of letting an LLM guess its way through your files, this tool spins up parallel subagents to analyze your code in parallel, then stitches their findings into a single, clear architecture document. It handles the heavy lifting of understanding dependencies and data flows, so you don't have to manually trace every file. The result is a living map of your project that updates as you make changes, giving your coding assistants precise context instead of vague guesses.

🆔 @hackernewsgithubprojects

📦 tetreesex/tetreesagent_ex

Tetrees Agent EX: Train AI Without a GPU?

Tetrees Agent EX is the integration kit that lets you run trained AI intelligence without owning a single expensive graphics card. Most developers assume training or running complex models requires heavy local hardware, but this tool flips that script by handling all the heavy lifting on hosted servers. You can search for specific AI packs, test them out, and even build your own custom intelligence that connects to your local tools. It handles the buying, building, and publishing process right from your command line. The best part is that your private data and credentials stay on your machine while the computation happens remotely.

📰 https://news.ycombinator.com/item?id=49512825

🆔 @hackernewsgithubprojects

📦 anysearch-team/anysearch-dsh

AnySearch for DeepSeek Harness

Give your AI agent live eyes on the web with AnySearch. This tool plugs into DeepSeek Harness to instantly add real-time web search and clean content retrieval, letting your model pull fresh facts right when it needs them. The coolest part is the batch search capability, which lets the agent run up to five different queries at the same time so one slow result doesn't hold up the whole process. You also get vertical search for specific areas like code or finance, all without needing any API key to get started. It is the easiest way to make your local AI agent actually useful for real-world research.

🆔 @hackernewsgithubprojects

📦 google-gemini/jot-gemini-transcribe-macos

Jot: Smart Dictation for macOS

Your Mac can now listen to you change your mind mid-sentence and fix it for you. A developer created Jot, a free dictation app that turns your speech into clean, punctuated text directly at your cursor. The real surprise is how it handles corrections. If you say you want to meet at one o'clock, then immediately backtrack and say two o'clock, Jot writes only the final decision. It strips out the hesitation and gives you a polished sentence without you having to edit a single word. It runs locally on your machine, keeping your audio private while using a specialized model to understand context.

🆔 @hackernewsgithubprojects

📦 ga1i13o/foundyou

FoundYou: One Model, Two Superpowers

FoundYou is the unified computer vision model that lets you find and trace a single object across massive datasets. You show the system just one picture of a specific item, like a blue mug, and it instantly locates every other instance in a library of one hundred million photos. It can even highlight exactly where that object sits in new images. This tool is incredibly fast and lightweight, running on a fraction of the hardware previous systems required. It bridges the gap between spotting something and cutting it out from the background. If you are building search engines or editing photos, this is the efficient engine you have been missing.

🆔 @hackernewsgithubprojects

📦 amap-ml/dreamx-creator

DreamX Creator Makes Video Sound

DreamX Creator takes a single image and a text prompt, then generates both video and sound at the same time. It is not just adding a soundtrack to a silent clip; the model builds the two streams together so the audio actually matches the on-screen action. This creates a realistic experience where what you hear is driven directly by what you see. The project is open source, meaning you can study the method or use the code to build your own synchronized media. If you want to create video that feels alive, this is the place to start.

🆔 @hackernewsgithubprojects

📦 yifanzhang-pro/fast-weight-attention

Falcon: How AI Models Learn On The Fly

Falcon changes how neural networks handle memory by treating every new piece of information as a quick, on the spot learning opportunity. Instead of trying to remember everything in a massive list, these models compress endless streams of data into a tiny, fixed size summary that updates constantly. This is called fast weight attention, and it turns the standard attention mechanism into a kind of live learning rule. You get a system that keeps adapting to what it sees right now without needing to rewind and retrain on old data. It is a clever way to make machines that learn continuously without getting bloated or forgetting what matters.

🆔 @hackernewsgithubprojects

📦 zjunlp/autoscirub

AutoSciRub: The Grader That Fixes AI Research

AutoSciRub is the research plugin that stops guessing and starts grading. Instead of letting an A I agent wander blindly through a scientific task, it first builds a specific checklist of what a good answer actually looks like. It checks the literature and the available data to define exactly which evidence proves success. Then, it uses that list to catch every gap in the draft, guiding the agent to fix only what is missing. It is the difference between hoping for a good result and verifying one. If you are building autonomous research tools, this is the missing link.

🆔 @hackernewsgithubprojects