RoboNeuron middleware layer connecting Vision-Language-Action models and LLM agents to robot middleware, standardizing tool API integration for embodied AI.
EvalBlocks modular framework for efficient evaluation of foundation models in medical imaging, reducing manual experiment tracking workflows.
Survey of meta-learning and meta-reinforcement learning methods enabling rapid task adaptation with minimal data, tracing DeepMind's adaptive agent research.
OPERA data pruning framework for efficient dense retriever adaptation, balancing quality-coverage tradeoff in domain-specific finetuning.
AI agents autonomously perform high energy physics analysis pipeline stages including event selection, background estimation, and statistical inference using LLMs.
LLM Router uses internal prefill activations for query-specific model selection, outperforming semantic routing by capturing model-specific failures.
CarbonEdge framework for carbon-aware deep learning inference on edge devices, optimizing for environmental impact alongside latency and throughput.
Agent2 open-source runtime for production AI agents with schema-to-API capabilities, auth, and provider routing.
Database optimizers become critical infrastructure when AI agents autonomously generate SQL queries.
3D semantic atlas of 188 constitutions using embeddings and UMAP for conceptual law search.
Hybro interoperability layer enables local and remote AI agents to coordinate in shared networks.
Trinity-Large-Thinking 398B sparse MoE model with chain-of-thought reasoning and agentic RL.
Trinity Large Thinking open-source reasoning model from Arcee AI optimized for agentic tasks.
Tutorial using Model Context Protocol server to turn NAS into self-hosted AI assistant.
Analysis of AI adoption risks: organizations mistaking temporary model limitations for safety assurance.
Analysis of two cache bugs in Claude Code API causing 10-20x token inflation and rapid rate limit exhaustion. Includes technical details and community-documented workarounds.
Study showing LLM-generated passwords are fundamentally insecure due to token prediction design. Examines risks in coding agents using LLMs for password generation.
Orbit: structured Python control layer for computer use agents, mixes model costs, extracts typed data, steers mid-task.
MetaLLM: security testing framework for AI/ML systems with 61 modules covering prompt attacks, RAG poisoning, agentic exploits.
Benchmarking 8 Bedrock models on RAG pipeline; cheapest Claude model outperforms competitors. Cost-quality analysis.
Analysis of Claude AI's ability to deobfuscate minified JavaScript code, discussing implications of AI-assisted code recovery.
Machine translations of Georges Perec's lipogram novel under 'no-e' constraint using models.
GlassFish Docker image maintenance effort. Open source infrastructure project with limited AI/ML relevance.
HN discussion about ChatGPT Containers feature for data analysis and Python execution. Incomplete post seeking use case examples.
Reverse-engineered Figma's WebSocket binary protocol using Claude to extract design scenegraph without API.
Claude-built skill analyzes plan denial patterns in Claude Code to iteratively improve autonomous planning behavior.
Open source Android AI assistant running LLaMA, DeepSeek, Qwen, Gemma locally without internet for privacy.
HN discussion thread on Japanese prompt injection attacks in LLM applications. No content provided.
HN discussion title on when to fine-tune image models. No content provided.
MAME emulation project migrating C codebase to Rust using AI refactoring for a complex 30-year preservation effort.
Open source markdown-native web server serving HTML to humans and markdown to agents. Zero build step, multi-runtime.
Essay on technical blogging practices and challenges in the age of AI-generated content.
Research from Fabraix on AI agent security incidents in 2026, runtime guardrails, and adversarial testing.
CAUM analyzes 80K agent sessions to detect loops and stagnation via behavioral patterns without reading prompts. AUC=0.814.
Tamp: token compression proxy for coding agents (Claude, Aider, Cline, Cursor). 52.6% token cost reduction.
Methodology for integrating agentic AI into production development using human-led TDD to maintain codebase control.
RAG-MCP server for offline MDN Web Docs with LanceDB vector store. 50k+ dataset on HuggingFace.
Video visualization of 51 real OpenClaw AI engineering tasks in 2D dungeon format. Minimal description provided.
Open source supply chain security analysis. Focuses on secret exfiltration attacks and GitHub capabilities.
Open-source protein language model pipeline using CodonRoBERTa-large-v2 trained across 25 species for $165, with structure prediction and sequence design capabilities.
Code typing trainer using real code snippets from repositories. Developer tool with niche application.
Lecture comparing software engineering transition to AI with 18th century structural design separation from craft construction.
Multi-model coordination patterns showing how Claude, Codex, Gemini compensate for each other's reasoning blindspots in agentic systems.
Open-source Rust/Bevy ECS simulation engine modeling tumor evolution and therapeutic resistance from first principles, clinically calibrated to real units.
MCP tool providing adversarial quality review for LLM agent outputs, integrating with Claude and any MCP client for AI pipeline evaluation.
Research on how agentic AI shapes human learning incentives and information ecosystem evolution through dynamic models of learning and decision-making.
Tamp.dev is a token compression proxy for coding agents that reduces input tokens by 52.6%. Works with Claude Code, Aider, Cursor, and OpenAI-compatible agents without code changes.
Brief reference to building CLI for AI agents and humans interaction in under 10 minutes. Minimal details provided.
Subjective comparison of Claude Opus 4.6 vs GPT 5.4 for coding work, focusing on working style differences rather than raw capability.
Open-source Go-based UI test automation runner for Android, iOS, Web, React Native, Flutter with single binary, no JVM or paywalls.