3/31/2026
Show HN: LLminate – LLM Detector
Proof-of-concept LLM detector leveraging training data properties to identify LLM-generated content.
3/31/2026
Show HN: Gha-outrunner – Ephemeral GitHub Actions runners without Kubernetes
Go tool providing ephemeral GitHub Actions runners using Docker/KVM/Tart without Kubernetes, open source MIT licensed.
3/31/2026
Show HN: I built a visual ML trainer that exports to standalone PyTorch code
MLForge visual ML trainer exports to standalone PyTorch code. Node-based interface with GPU/CPU/MPS support and early stopping.
3/31/2026
Multimodal file search and sharing for AI agents
Title-only entry on multimodal file search and sharing for AI agents. Insufficient content.
3/31/2026
Show HN: SkillForge – Turn code and docs into instructions AI agents can follow
SkillForge converts code and documentation into instructions for AI agents to execute.
3/31/2026
Using LLMs to amplify human labeling and improve Dash search relevance
Dropbox Dash uses LLMs with RAG pattern for enterprise search. LLMs amplify human labeling to improve retrieval and response relevance.
3/31/2026
Claude Code Internals: An AI-Assisted Analysis of the Leaked Source
Title-only entry analyzing Claude Code internals from leaked source. Insufficient content.
3/31/2026
Show HN: Agent Wellbeing Kit – boundary protection for humans running AI agents
Title-only entry on safeguards for humans operating AI agents. Insufficient content.
3/31/2026
Autodebug: Telemetry-Driven Inference Optimization Loop
Autonomous agent continuously deploying inference services, benchmarking, analyzing telemetry, and redeploying with optimized configurations.
3/31/2026
Ask HN: Academic study on AI's impact on software development – want to join?
NYU and City University of London recruiting software developers for interview study on AI tool adoption and impact.
3/31/2026
Show HN: Team agents that manage your CMS from Slack, WhatsApp, and Telegram
Cosmic launches Team Agents—persistent AI agents in Slack/WhatsApp/Telegram with real capabilities for content management and publishing.
3/31/2026
Claude Code Auto Mode: A Comprehensive Technical Summary
Claude Code Auto Mode uses ML classifier to approve/deny tool calls. Second Claude instance evaluates execution safety in real-time.
3/31/2026
Show HN: Lazy-tool: reducing prompt bloat in MCP-based agent workflows
Go tool reducing prompt token bloat in MCP-based AI agent workflows. Single binary with SQLite catalog, no containers or external deps.
3/31/2026
From "Show Me" to "Do It": How OdooClaw Takes Action in Your Odoo
OdooClaw enables action-capable AI agents for Odoo ERP systems with concrete examples of data modification vs read-only AI.
3/31/2026
Show HN: PhAIL – Real-robot benchmark for AI models. The gap to humans is 20x
PhAIL benchmark for evaluating AI models on real robotic tasks, showing 20x performance gap versus humans.
3/31/2026
AI agent that writes and hot-reloads its own Python modules at runtime
AI agent system that autonomously writes and hot-reloads Python modules at runtime, demonstrating self-modifying code capabilities.
3/31/2026
Ask HN: What are you building with AI coding agents / tooling?
HN thread soliciting projects built with AI coding agents and tooling, encouraging developers to share repos and implementations.
3/31/2026
Create Issues from Slack with Copilot
Integration tool using Copilot to create GitHub issues from Slack messages.
3/31/2026
ClawDesk – Agent orchestration layer on top of OpenClaw
Orchestration layer built on top of OpenClaw for coordinating AI agent workflows.
3/31/2026
Show HN: Stop your agent running Terraform destroy or Git push origin main
Safety guardrails tool preventing AI agents from executing dangerous operations like Terraform destroy or force git pushes.
3/31/2026
Show HN: Solitaire – identity layer for AI agents, not just another memory tool
Solitaire: Identity layer for AI agents managing persistent context and behavior. Framework for agent personalization beyond memory systems.
3/31/2026
Show HN: Gravimera, AI(LLM) driven 3D world editor and explorer
Gravimera: LLM-driven 3D world editor and explorer. AI application for interactive 3D content creation.
3/31/2026
Show HN: Prawduct, a product development framework for Claude Code
Framework of prompts and templates optimizing Claude Code for product development with structured requirements and test generation.
3/31/2026
We accidentally built a machine that prints viral hooks. It scared us
Team built system generating viral social media hooks using LLM. Commentary on AI-generated content with limited technical details.
3/31/2026
Dot – A Siri Replacement learns skills through Apple Shortcuts
Dot: Siri replacement that learns skills via Apple Shortcuts. AI agent application with custom skill extension.
3/31/2026
Show HN: Reprompt – Analyze what you type into AI tools, not what they output
Reprompt: Tool analyzing input prompts to AI systems rather than outputs. Developer tool for prompt engineering analysis.
3/31/2026
OpenClaw for Dummies (getting to a minimum viable OpenClaw agent)
Tutorial building a minimum viable OpenClaw agent with step-by-step implementation guide and examples.
3/31/2026
Clawdia – Claude Code/Codex and with full OS control – electron app with browser
Electron desktop app integrating Claude Code and OpenAI Codex as AI execution engines with MCP tool bridging, terminal emulation, and concurrent agent execution.
3/31/2026
Aki.io – Open-source AI models via API on EU infrastructure (OpenAI-compatible)
Aki.io: OpenAI-compatible API serving open-source AI models on EU infrastructure. Developer tool for LLM access.
3/31/2026
Show HN: Migas – Meeting copilot with live speaker labels (no bot, no cloud STT)
Migas: Local-inference meeting copilot with on-device voice fingerprinting for speaker identification and context-aware queries. No cloud STT.
3/31/2026
Show HN: I built a way for AI to remember how you like things done
AEP (Agent Experience Protocol): JSON-based framework to save and reuse AI agent workflows, constraints, and preferences across sessions.
3/31/2026
Git history is not a reliable record of work. I tried to make it one
Developer built context extraction layer using LLMs and heuristics to reconstruct work from git history amid AI-assisted coding volume.
3/31/2026
Show HN: PromptQL – AI-Native Slack
PromptQL: Slack integration for AI-native workflows. Developer tool for LLM applications.
3/31/2026
The day my AI editor went silent: Debugging Git worktrees, and lost weights
Debugging story involving AI editor issues with Git worktrees and model weights. Technical troubleshooting narrative.
3/31/2026
Show HN: Vibe Check – UX Benchmark for vibe designs
Hivemind is a self-hosted multi-agent AI platform emphasizing user data sovereignty and local execution.
3/31/2026
What Memoir Scandals Tell Us about Two LLM Writing Scandals
Open source eBPF firewall for GitHub Actions that prevents untrusted connections from LLM agents and CI runners.
3/31/2026
Totem – proxy that detects if your LLM has been tampered with
Quadrotor simulation project using Three.js and Python, mentions potential LLM agent applications.
3/31/2026
(OSS) >90% of LLM calls don't need an LLM and could be replaced by classic ML
Totem is a proxy tool that detects tampering or compromise of LLM model behavior.
3/31/2026
Show HN: AppSwitcher – Bringing rcmd-style app switching to Windows
Open source analysis showing 90% of LLM API calls could be replaced with traditional machine learning.
3/31/2026
Show HN: "Have your LLM talk to my LLM"
Project enabling LLM-to-LLM communication and interaction patterns.
3/31/2026
CLI-agent-lint – audit any CLI for AI agent-readiness
Open source tool to audit CLI tools for AI agent compatibility and readiness.
3/31/2026
AI benchmarks are broken. Here's what we need instead
Analysis of limitations in current AI benchmarking methodologies and proposals for improved evaluation frameworks.
3/31/2026
I Handed an AI Agent 27 Domains and a Deadline. 72 Days Later
Personal experiment documenting use of an AI agent to manage 27 domains over 72 days, sharing observations about future workflows.
3/31/2026
Stop Prompting, Start Governing – Why healthcare AI needs contracts, not prompts
Healthcare AI expert argues governance frameworks and contracts are more critical than prompts for responsible AI deployment in medicine.
3/31/2026
I stopped "chatting" with AI and 10x'd my code output using a 4-phase protocol
Former manager describes a 4-phase protocol for using AI tools to improve code output, sharing techniques for effective AI-assisted software development.
3/31/2026
AgentVeil – EigenTrust reputation and sybil detection for AI agents
AgentVeil implements EigenTrust-based reputation system and sybil detection mechanism for AI agents.
3/31/2026
AI Won't Lower Your Engineering Bar. Poor Adoption Will
Analysis of engineering team SLA misses attributed to over-reliance on AI-generated code without adequate review; advocates for proper adoption practices.
3/31/2026
A curated list of high-signal analyses from the Claude Code source leak
Analysis of Claude Code source leak via exposed npm sourcemap, curating high-signal technical insights from 1,900+ TypeScript files.
3/31/2026
OMLX – LLM inference, optimized for your Mac
oMLX is a macOS-native MLX server for LLM inference with persistent KV cache optimization; reduces latency from 90s to 5s for coding agents.
3/31/2026
Show HN: Opal – run GitLab pipelines locally using Apple container
Opal CLI runs GitLab pipelines locally with TUI interface; supports macOS container runtime and includes Ollama AI agent integration.