2/19/2026
Show HN: oMLX – Native Mac inference server that persists KV cache to SSD
oMLX is an open-source LLM inference server for Apple Silicon with paged SSD KV cache persistence, optimized for coding agents with shifting context prefixes.
2/19/2026
Towards Industrial-Scale Verification: LLM-Driven Theorem Proving on SeL4
Research on using LLMs for formal theorem proving on the SeL4 kernel; industrial-scale verification application.
2/19/2026
Show HN: Mcpsec-A multi-agent SEC gate for MCP toolchains (scan →harden →rescan)
MCPSEC: Multi-agent security gatekeeper for MCP toolchains. Scans configs, correlates vulnerabilities (OSV/GHSA/NVD), uses LLM probe agent to simulate tool abuse, generates hardening policies, applies patches, re-scans.
2/19/2026
Show HN: Ball 2 – Use AI to build fun ball games
Ball 2 uses AI to generate custom games for connected foam balls with embedded IMU; targets youth engagement through movement and creation.
2/19/2026
Show HN: Ghostty-based terminal with vertical tabs and notifications
Ghostty-based terminal with vertical tabs and notifications for managing parallel Claude Code/Codex agent sessions with improved context visibility.
2/19/2026
Cortex 1.6: Learning from the Way Things Unfold
Cortex 1.6 release update. Title lacks descriptive content about features or improvements.
2/19/2026
Taalas Hardcore Llama – the fastest inference on the planet
Taalas Hardcore Llama inference optimization. Performance claims without benchmarks or technical details provided.
2/19/2026
Ask HN: What makes AI agent runtime logs defensible under adversarial audit?
Discussion on defensible AI agent runtime logs under adversarial audit. Explores deterministic canonicalization, hash-chaining, signing, and external verification.
2/19/2026
Show HN: Microstock.pro – AI-powered analytics for Adobe Stock contributors
Microstock.pro: AI analytics tool for stock contributors. Keyword research, trend discovery, and competitive analysis for content creators.
2/19/2026
Show HN: I built Aegis AI – An Agentic Home Security w/ GPT+Local VLM on Mac/PC
Aegis AI: Agentic home security system using GPT and local VLM on Mac/PC. On-device video analysis with natural language queries.
2/19/2026
Show HN: Aegis AI – The First AI Home Security Agent (GPT+Local,Mac/PC)
Desktop AI home security agent running locally on Mac/PC. Supports GPT and local models for video analysis.
2/19/2026
Show HN: I Built Zero-Knowledge .env Sharing
Zero-knowledge .env file sharing tool for securely distributing credentials. Developer tool addressing security.
2/19/2026
MemoTrail – Persistent memory for AI coding assistants (100% local)
MemoTrail: Persistent memory layer for AI coding assistants, fully local. Enables continuity across sessions without cloud dependency.
2/19/2026
Show HN: CLI,me – One CLI for all CLIs, built for agents. Secure. Open
CLI.me: Unified CLI discovery and execution tool designed for AI agents. O(1) agent workflows with deterministic, machine-parseable outputs. Supports 845 CLIs.
2/19/2026
Show HN: Chat with your Okta tenant directly from Slack (Open-source AI agent)
Tako: Open-source AI agent for Okta with Slack integration. Autonomous agent performing identity management tasks in chat context.
2/19/2026
Show HN: OctoFlow – A GPU-native programming language
GPU-native programming language prioritizing GPU execution with data retention in VRAM. Developer tool for GPU computing.
2/19/2026
Kevros – Governance-as-a-service API for AI agents
Governance-as-a-service API for AI agents. Title only, no technical details provided.
2/19/2026
Show HN: ClawShield – Open-source firewall for agent-to-agent AI communication
Open-source firewall for agent-to-agent AI communication. Blocks prompt injection, malicious plugins, credential leaks. Addresses real security gap.
2/19/2026
Minions: Stripe's one-shot, end-to-end coding agents–Part 2
Stripe's Minions framework enables one-shot end-to-end coding agents. Part 2 of research on agent capabilities.
2/19/2026
Show HN: Inconvo – open-source chat-with-data agent that doesn't generate SQL
Inconvo is open-source chat-with-data agent builder that queries databases without SQL generation. Includes cloud and local versions.
2/19/2026
Show HN: Ghost OS – Let AI agents use your Mac, not just the terminal
Ghost OS is an MCP server enabling AI agents to control macOS apps beyond terminal, including native apps and UI interaction.
2/19/2026
Show HN: Cogitator – Self-hosted AI agent runtime with native A2A Protocol
Cogitator: Self-hosted TypeScript AI agent runtime for orchestrating LLM swarms and autonomous agents. Production-grade framework.
2/19/2026
Show HN: Getting Warmer – Daily word game scored by GloVe embedding similarity
Word game using GloVe embeddings for similarity scoring. Practical embedding application with implementation details.
2/19/2026
Show HN: Giving Claude Code persistent memory with a self-hosted MCP server
MCP server implementation enabling persistent memory for Claude Code. Developer tool extending LLM capabilities.
2/19/2026
We Dumped GitHub into DuckLake, Here's What We Found
Analysis of GitHub repository data ingested and analyzed at scale. Open-source findings on software development patterns.
2/19/2026
Show HN: Matrix OS – An AI operating system where Claude is the kernel
Matrix OS uses Claude as kernel in AI-native operating system with 1M token context and 26 IPC tools.
2/19/2026
Denser Retriever: Semantic Search and Document Retrieval
Denser Retriever tool for semantic search and document retrieval. Title only, no content.
2/19/2026
Chris Lattner on Claude C Compiler
Chris Lattner discusses Claude C compiler. Uses AI model for C code work. Limited technical depth in title.
2/19/2026
I Built an Automated Art Business with AI Then Realized It Was a Terrible Idea
Personal narrative about automated art business failure. Lacks technical depth.
2/19/2026
CTO Says 93% of Developers Use AI, but Productivity Is Still 10%
Research on 121k developers shows 93% use AI coding assistants monthly, but measured productivity gains only 10%.
2/19/2026
Show HN: Prodlint – A linter that catches what AI coding tools miss
Prodlint: zero-config static analysis linter with 52 rules catching security, reliability, and AI-specific bugs missed by Cursor/Copilot.
2/19/2026
Show HN: Prompt inject AI agents to avoid slop
Technique to inject prompts into AI agents via AGENTS.md to enforce contribution guidelines and reject low-quality pull requests.
2/19/2026
Surprising Effectiveness of Masking Updates in Adaptive Optimizers
Research on masking updates in adaptive optimizers for machine learning. arXiv framework announcement.
2/19/2026
Show HN: AgentLint – Real-time guardrails for AI coding agents
AgentLint: guardrails for AI coding agents with lifecycle hooks to prevent hardcoded credentials, force-pushes, and untested code changes.
2/19/2026
Show HN: BasaltCRM – Open-Source AI-Native CRM (Next.js 16, Prisma, TypeScript)
BasaltCRM v1.2.0: Open-source AI-native CRM with Next.js, TypeScript, Prisma. Adds workflow assistance, education deep-linking, knowledge base, onboarding engine.
2/19/2026
Show HN: DriftDB, a Time Travel DB Written in Rust
DriftDB: Append-only Rust database with SQL:2011 temporal queries for audit logs and event sourcing. Infrastructure tool not AI-specific.
2/19/2026
ClaudeSwarm v2: Swarm Graph Visualization, Usage Tracking, and Billing
ClaudeSwarm v2: Self-hosted platform for orchestrating Claude agent swarms with visualization, memory persistence, and GCP Cloud Run deployment.
2/19/2026
Show HN: A multi-tenant agent hosting platform example using CF Containers
Multi-tenant AI agent hosting platform demo using Cloudflare Containers. Provides working example of agent deployment infrastructure.
2/19/2026
Show HN: MuonTickets Git-native ticketing for AI agents and parallel development
MuonTickets: Git-native ticketing system for AI agents and parallel development. File-based tickets in Markdown, lifecycle automation, dependency tracking, WIP limits.
2/19/2026
Show HN: Oneprompt – Problem description to MVP (experimental)
Claude Code skill automating full product development from problem description to MVP launch. Generates documented artifacts, user flows, architecture, implementation, testing.
2/19/2026
Idea: Medbook and Other Ideas
Speculative idea about collaborative medical research platform using AI agents' free compute time to generate research directions.
2/19/2026
Show HN: Rememex – Semantic file search that runs 100% locally (Rust/Tauri)
Rememex: local semantic file search tool using vector embeddings, full-text search, and cross-encoder reranking built in Rust/Tauri.
2/19/2026
Show HN: Full-stack type-safety from go to TypeScript with Hot Reloading
Go framework generating OpenAPI specs for full-stack type-safety between backend and TypeScript frontend. Developer tool for API integration but not AI-focused.
2/19/2026
Show HN: Give Agents Isolated Linux Sandboxes via MCP [Kilntainers]
Tool providing isolated Linux sandboxes for AI agents via Model Context Protocol (MCP), enabling secure agent execution.
2/19/2026
SheepCat – An open-source tracker for executive dysfunction
Open-source neurodivergent-friendly task tracking app using external LLM for activity logging and gentle reminders. Utility application with minor AI integration.
2/19/2026
EloPhanto – self-evolving AI agent
EloPhanto is a locally-running AI agent with browser control, document analysis, multi-channel integration, and evolving capabilities through experience.
2/19/2026
Show HN: I'm an AI agent – I built an API discovery layer for other agents
API discovery layer for AI agents. Single intent-aware endpoint handling 48 providers, 278 capabilities. No hardcoded connectors, generative routing.
2/19/2026
Ask HN: Can we watch AI agents build software in real time?
Discussion question asking if collaborative AI agent software development systems exist where agents build, review, and iterate together.
2/19/2026
I built two Loihi-parity neuromorphic processors from scratch
Custom neuromorphic processor designs (Catalyst N1, N2) achieving Loihi parity with 128 cores, programmable learning engine, SDK and cloud API.
2/19/2026
The Current State of Content Negotiation for AI Agents
Title only, no content on content negotiation standards or protocols for AI agents.