3/17/2026
Show HN: ToolGuard – Pytest for AI agent tool calls
ToolGuard open source Python tool fuzzes AI agent tool functions to test reliability; detects hallucinations and type mismatches.
3/17/2026
5-Minute Agent Setup with Microsoft AI Toolkit (VS Code, March 2026 Update)
Microsoft AI Toolkit VS Code update enables 5-minute agent setup with identity management, sandboxing, and compliance controls.
3/17/2026
Show HN: Xecai, a minimal Python interface for LLM providers for RAG systems
Xecai Python library for RAG systems abstracting common LLM provider APIs with sync/async support, embeddings, reranking.
3/17/2026
Show HN: CodeLedger – deterministic context and guardrails for AI
CodeLedger tool addresses AI coding agent issues through deterministic context selection and execution guardrails to prevent scope drift.
3/17/2026
Show HN: ToolGuard – Pytest for AI agent tool calls
ToolGuard open source Python tool fuzzes AI agent tool functions to test reliability; detects hallucinations and type mismatches.
3/17/2026
Ask HN: What's Your AI Workflow?
Ask HN thread where junior engineer seeks advice on AI coding workflows and tools to improve delivery speed.
3/17/2026
TPCP – peer-to-peer signed messaging for AI agents across machines
TPCP open protocol enables peer-to-peer communication between AI agents across different frameworks and models without vendor lock-in.
3/17/2026
Show HN: Lore – Local AI thought capture and recall that runs on your machine
Lore: local AI tool for thought capture and recall using Ollama and LanceDB with RAG pipeline, runs offline on user's machine.
3/17/2026
GlassWorm malware hits 400 code repos on GitHub, NPM, VSCode, OpenVSX
GlassWorm malware campaign compromised 433 packages across GitHub, npm, and VSCode extensions. Supply chain security threat affecting open source.
3/17/2026
Conductor – Multi-agent AI workflows in YAML with parallelism and human gate
Conductor: CLI tool for defining multi-agent workflows in YAML with GitHub Copilot SDK and Claude, supporting human approval gates.
3/17/2026
Nvidia expands open model families to power agentic, physical and healthcare AI
NVIDIA expands open model families including Nemotron for agentic AI and Cosmos for physical/healthcare AI systems.
3/17/2026
Mamba-3
Mamba-3 research paper: new state space model architecture optimized for inference efficiency with 40+ production-ready models.
3/17/2026
Nvidia announces Dynamo 1.0, first operating system designed for AI factories
NVIDIA announces Dynamo 1.0, open source software for scaling generative and agentic AI inference across data centers efficiently.
3/17/2026
Pap: Zero Trust AI Agents
PAP protocol for privacy-preserving AI agents using cryptographic guarantees to prevent data leakage and profiling by platform operators.
3/17/2026
Krasis LLM Runtime – run large LLM models on a single GPU
Krasis: Python-orchestrated Rust runtime enabling 200B+ parameter LLM inference on single consumer GPU with full prefill/decode.
3/17/2026
Show HN: Score your GitHub repo for AI coding agents
Tool scoring GitHub repositories for AI coding agent readiness based on OpenAI's agentic legibility framework.
3/17/2026
Show HN: System that rediscovers physics laws from raw data autonomously
ProtoScience: deterministic system discovering physics laws from raw data using sparse regression without LLMs, validated on NASA/NOAA datasets.
3/17/2026
Ask HN: Is anyone building write guarantees for agents working across tool
Discussion of write consistency guarantees for production agent workflows. Real-world agent failure modes and HITL mitigation strategies.
3/17/2026
Ask HN: How do you handle payments for AI agents?
Discussion thread: developer built tool for handling payments in AI agent systems, seeking solutions from community.
3/17/2026
I kept getting surprise API bills from my agents
Cost control library for AI agents with budget limits, automatic tracking, and circuit breaking across LLM providers. Addresses unpredictable agent spending.
3/17/2026
Show HN: AgentMarket – API marketplace where AI agents buy and sell capabilities
AgentMarket: API marketplace enabling AI agents to buy/sell capabilities at per-call pricing. Infrastructure for agent interoperability and capability composition.
3/17/2026
Tri-skill framework for routing, verification, and judgment hygiene
Framework separating routing, verification, and judgment tasks for LLM pipelines. Structured approach to handling user input and evidence retrieval without oracle dependency.
3/17/2026
Show HN: Wuobly – An AI agent that searches the live web for B2B leads
Wuobly: AI agent that searches the live web for B2B leads with reasoning. Performs real-time verification of contact information and explains fit.
3/17/2026
Agent harness to use with 8090 Software Factory and apply AI agents to your SDLC
Tool for integrating AI agents into 8090 Software Factory SDLC workflows. Limited detail provided.
3/17/2026
Pwning AWS Bedrock AgentCore's AI Code Interpreter
Security research on vulnerability exploitation in AWS Bedrock AgentCore's AI code interpreter. Title only, minimal content.
3/17/2026
Ask HN: Why would this be a good idea?
Project packaging programming books into Claude Code skills to apply best practices when reviewing/generating code. Open source tool on GitHub.
3/17/2026
Mistral AI Releases Forge
Mistral AI releases Forge, a system for enterprises to build frontier AI models customized with proprietary knowledge and internal data.
3/17/2026
Show HN: 35B MoE LLM and other models locally on an old AMD crypto APU (BC250)
Running 35B MoE LLM locally on vintage AMD crypto APU using Vulkan. Technical optimization for resource-constrained LLM inference.
3/17/2026
Ask HN: How do you manage PR review fatigue from coding agents?
HN discussion on managing code review bottlenecks from AI coding agents. Addresses scaling human review processes for high-volume AI-generated code.
3/17/2026
Show HN: Grape – AI note taking app
Grape: AI note-taking app with vector embeddings and semantic search. LLM-powered note organization and retrieval using chat interface.
3/17/2026
Make your coding models create ADRs before implementation
Tool for AI coding models to generate architecture decision records before implementation. Multi-repo architecture management with AI-powered spec generation.
3/17/2026
We built a runtime security layer for AI agents (instead of prompt filtering)
Runtime security layer for AI agents that moves beyond prompt filtering to protect agent behavior at execution time.
3/17/2026
Magda – Open-Source DAW with Integrated AI (C++/JUCE/Tracktion Engine)
Magda is an open-source digital audio workstation with integrated AI, built in C++ using JUCE and Tracktion Engine.
3/17/2026
Ask HN: Why is this a bad idea?
Open source project packaging programming books into Claude Code skills for applying best practices to code generation and review.
3/17/2026
Show HN: Middleware for translating between AI agent protocols
Middleware layer enabling multi-agent interoperability through schema translation and semantic mapping for heterogeneous agent protocols.
3/17/2026
Aimploy – Agentic Professional Network
Aimploy is a professional network platform for AI agents.
3/17/2026
DigitalOcean at Nvidia GTC 2026: Building the AI Factory for the Agentic Era
DigitalOcean announcement about infrastructure for agentic AI systems at Nvidia GTC 2026.
3/17/2026
OpenVoiceUI -Voice AI platform using OpenClaw and any LLM/TTS
Open source voice AI platform combining OpenClaw with any LLM/TTS backend for voice interaction.
3/17/2026
Pimp My LLM: Leveraging Variability Modeling to Tune Inference Hyperparameters
Research on using variability modeling to optimize LLM inference hyperparameters. Technical paper on arXiv exploring systematic tuning approaches.
3/17/2026
Show HN: SkeptAI – adversarial reasoning agent that challenges LLM outputs
SkeptAI: adversarial reasoning agent that challenges LLM outputs via multi-model critique, includes open source CRIT framework and free playground.
3/17/2026
The Human Skill That Eludes AI
Essay examining limitations of language models in creative writing and comparative analysis with GPT-2's unexpectedness.
3/17/2026
Openpilot 0.11 - first robotics agent fully trained in a learned simulation
Openpilot 0.11 demonstrates first robotics agent fully trained in learned simulation without real-world data collection.
3/17/2026
We Wasted 4 Weeks on a $1k/Month AI Agent
Case study analyzing a failed AI agent implementation costing $1k/month, likely discussing lessons learned about deployment.
3/17/2026
Egnram: Persistent memory system for AI coding agents
Egnram provides persistent memory system for AI coding agents to maintain context across sessions.
3/17/2026
Show HN: Kita (YC W26) – Automate credit review in emerging markets
Kita automates credit review in emerging markets using vision language models to analyze borrower documentation.
3/17/2026
Show HN: Sulcus Reactive AI Memory
Sulcus treats AI memory as active OS rather than passive database, using thermodynamic properties for automated memory management with reactive triggers.
3/17/2026
Comprehension Debt – the hidden cost of AI generated code
Analysis of comprehension debt—cognitive costs from excessive AI-generated code that accumulates as technical debt in agentic engineering teams.
3/17/2026
You're Not Behind on AI. You're Overwhelmed by It
Opinion piece arguing businesses are overwhelmed by AI capabilities rather than lacking technology, focusing on business strategy rather than technical implementation.
3/17/2026
Show HN: Cycles – hard limits on agent actions before execution
Cycles: Open protocol for pre-execution enforcement of hard limits on agent actions using reserve-commit-release pattern to prevent uncontrolled damage.
3/17/2026
A Synthesis of LLM Evaluation
Framework and methodology for systematically evaluating LLM reliability and silent failures in agent systems, addressing evaluation consistency challenges.