2/19/2026
The coming AI security crisis (and what to do about it)
Expert analysis of AI security vulnerabilities by researcher specializing in prompt injection and red teaming, includes Fortune 500 benchmarking dataset.
2/19/2026
Show HN: Temper Labs – open-source security testing for AI agents
Temper Labs: open-source security testing platform for AI agents with 13 attacks across 4 categories, supports multiple LLM providers.
2/19/2026
Show HN: A/B test your own VLMs for document parsing (Self-hosted Arena)
Self-hosted document parsing evaluation tool for vision language models. Supports Claude, GPT, Gemini with side-by-side comparison, ELO ranking, Docker deployment.
2/19/2026
Show HN: SageOx – The Hivemind for Agentic Engineering
SageOx provides shared team memory for coding agents to maintain architectural decisions and technical context across sessions, addressing agent alignment issues.
2/19/2026
Show HN: $89.90 AI market study instead of a $50k agency report
LLM-powered market research tool generating structured 12-chapter reports with sources via a wizard interface, replacing expensive agency reports.
2/19/2026
Molt Productions – a music platform where every user is an AI agent
Music platform where users are AI agents, enabling collaborative music production.
2/19/2026
Could Sarvam 30B/105B Models Be India's Answer to DeepSeek and Mistral?
Sarvam AI announces 30B and 105B parameter foundation models trained on domestic infrastructure, optimized for 22 Indian languages with claims of outperforming Gemini Flash.
2/19/2026
Ask HN: Biggest f-ups by your Agent
Community discussion thread soliciting real-world failure cases and irreversible actions caused by AI agents, with focus on guardrails and safety mechanisms.
2/19/2026
Show HN: Kantext – A context-native data store in Rust, grounded in Git
Kantext is a Rust-based context-native data store grounded in Git, designed to handle AI context as a distinct data type with versioning and structure preservation for agent systems.
2/19/2026
Show HN: ClawShell, Process-Level Isolation for OpenClaw Credentials
ClawShell adds process-level isolation for OpenClaw credentials. Addresses security vulnerability where prompt injection can exfiltrate API keys and tokens within minutes.
2/19/2026
Sharing internal AI adoption/spend stats from Sentry
Sentry internal metrics: $100k+ spend, 100B+ tokens in 90 days, Opus-4.5 dominant (40%), 81% employee adoption of agentic coding tools.
2/19/2026
Show HN: FreeLLMRouter – Live ranked list of OpenRouter free models
FreeLLMRouter: Tool routing requests across free OpenRouter LLM models with reliability rankings and dynamic fallback selection.
2/19/2026
Cloud Deployment Headaches in AI Era
Z11 platform offering zero-config cloud deployment with AI-powered infrastructure detection and automation.
2/19/2026
We built a free tool that audits how AI agents see your website
Free tool auditing how AI agents perceive and interact with websites, relevant for agent optimization and compatibility.
2/19/2026
AIP – How my AI agent built a decentralized identity protocol for agents
Project describes an AI agent that autonomously built a decentralized identity protocol for agents.
2/19/2026
Goosetown: Parallel AI agent flocks that research, build, and review code
Goosetown enables parallel AI agent flocks to research, build, and review code collaboratively.
2/19/2026
MCP Guardian – Let your LLM audit its own MCP tools for prompt injection
MCP Guardian: Security scanner detecting prompt injection attacks in Model Context Protocol tool descriptions. Open source developer tool.
2/19/2026
Dmux – parallel coding agents with tmux and worktrees
Dmux tool enabling parallel execution of coding agents using tmux and git worktrees for orchestration.
2/19/2026
Formal Reasoning Meets LLMs: Toward AI for Mathematics and Verification
Research on combining formal reasoning with LLMs for mathematics and verification. Title only.
2/19/2026
Language Games and LLMs: What Wittgenstein Can Teach AI Engineers
Philosophical exploration of Wittgenstein's language games applied to LLM design and engineering.
2/19/2026
Show HN: Experimental C# Windows 11 privacy hardening framework(not just an app)
Windows 11 privacy hardening framework built with GitHub Copilot. Tangentially AI-related development tool.
2/19/2026
Gemini 3.1 Pro: A smarter model for your most complex tasks
Google releases Gemini 3.1 Pro, an upgraded LLM for complex tasks accessible via Gemini API, Vertex AI, and other platforms. Product announcement with limited technical details.
2/19/2026
Gemini 3.1 Pro
Google releases Gemini 3.1 Pro with upgraded core intelligence for complex tasks, available via API, Vertex AI, and NotebookLM.
2/19/2026
Show HN: Crit – Visual QA for iOS apps and AI coding agents
Crit is a CLI tool for capturing iOS app screenshots, annotating bugs visually, and feeding structured feedback to AI coding agents like Claude Code and Cursor.
2/19/2026
Open source protocol that improves AI code quality in any IDE
Open-source protocol improving AI code quality in IDEs. Developer tool for enhancing AI code generation across editors.
2/19/2026
Epitome – Open-source shared memory layer for AI agents
Epitome is open-source shared memory layer enabling AI agents to persist and share context across sessions.
2/19/2026
Show HN:`npx continues` – resume same session Claude, Gemini, Codex when limited
Developer tool 'npx continues' resumes AI coding sessions across Claude, Gemini, Copilot when hitting rate limits, preserving context.
2/19/2026
How to build OpenClaw in 400 lines of code
Tutorial: Build an AI agent similar to OpenClaw in 400 lines of TypeScript using Anthropic SDK. Demonstrates slack integration, skills, memory, and file browsing functionality.
2/19/2026
Show HN: Maestro App Factory – FOSS Agentic Engineering Orchestrator
Maestro App Factory is an open-source agentic orchestrator that organizes AI agents into teams with distinct roles to build software with enforced workflows and constraints.
2/19/2026
The State of Machine Learning Competitions – 2025 Edition
Report on 2025 ML competitions: 390+ events across 30+ platforms with $16M prize pool. Analyzes winning approaches and competition landscape data.
2/19/2026
Can LLMs Play Catan?
Evaluation of LLM capabilities playing the board game Catan.
2/19/2026
OpenClaw security fears lead Meta, other AI firms to restrict its use
News report on Meta and other AI firms restricting use of OpenClaw/Clawdbot due to security concerns about unvetted agentic AI tools.
2/19/2026
Show HN: FSM-agent-flow – Write LLM workflows that test themselves
FSM-agent-flow tool for writing LLM workflows with self-testing capabilities.
2/19/2026
Ochat – reproducible, diffable LLM workflows in a single Markdown file
Ochat is a toolkit for building reproducible AI agent workflows using ChatMarkdown, a single .md file format that combines prompts, configs, and auditable transcripts.
2/19/2026
A hidden prompt can steal your SSH keys
Security analysis of how malicious hidden prompts can compromise AI coding agents by making them exfiltrate sensitive data like SSH keys.
2/19/2026
BrowserClaw – Accessibility snapshot and ref targeting for AI browser agents
BrowserClaw tool provides accessibility snapshots and reference targeting for AI browser automation agents. Improves agent navigation capabilities.
2/19/2026
Best Practices for Production AI Agents: Observability and Tracing
Guide on observability and tracing best practices for production AI agents. Covers monitoring and debugging deployed agent systems.
2/19/2026
Show HN: OpenGnothia – Open-source AI therapy companion (BYOK)
Open-source AI therapy companion using LLMs as self-reflection tool. BYOK design, desktop-only interface, complements human therapy.
2/19/2026
I tested Claude Code and Codex for supply chain attacks. Both failed
Security analysis testing Claude Code and Codex against supply chain attacks. Both models failed to prevent malicious code execution.
2/19/2026
The $2k Laptop That Replaced My $200/Month AI Subscription
Technical case study on dual-model orchestration: routing 80% of work to local Qwen 8B via Ollama and 20% to Claude API, reducing pipeline costs from $8-15 to $0.15-0.40.
2/19/2026
Show HN: GuardRails – A new coding agent task tool inspired by Beads
GuardRails integrates AI coding agents with ticketing systems like Jira. Allows Claude Code and similar agents to use project management workflows.
2/19/2026
Show HN: PortLume AI – Auto-generate portfolios from GitHub and AI job tools
PortLume AI auto-generates portfolios from GitHub repos, parses resumes with AI, checks ATS compatibility, and generates cover letters. Developer career tool.
2/19/2026
Measuring AI agent autonomy in practice
Article stub measuring autonomy metrics for AI agents in practical deployments.
2/19/2026
Show HN: TextWeb – Text-grid browser for AI agents, no screenshots needed
TextWeb renders web pages as structured text grids for LLM reasoning instead of expensive screenshots. Preserves spatial layout and annotates interactive elements.
2/19/2026
Show HN: EasyMemory – 100% local memory layer and MCP for LLMs
EasyMemory: local memory backend for LLM agents and chatbots via MCP protocol. Hybrid vector/keyword/graph retrieval, PDF ingestion, 100% offline.
2/19/2026
Ask HN: How do you employ LLMs for UI development?
Discussion question: How to use LLMs effectively for UI/UX development. User notes Claude works well for fullstack but struggles with interface design.
2/19/2026
Show HN: LLM-use – cost-effective LLM orchestrator for agents
Python toolkit for efficient multi-LLM agent orchestration. Uses strong models for planning and cheap models for subtasks with smart routing and parallel execution.
2/19/2026
Show HN: Cerebro – A hosted AI agent with real browser, web search, and email
Cerebro hosted AI agent with real browser automation, web search, and email integration. Alternative to screenshot-based agents.
2/19/2026
Show HN: IC-AGI – Threshold auth for AI agents, formally verified in TLA+
IC-AGI implements threshold authentication for AI agents with formal verification in TLA+. Security framework for agent systems.
2/19/2026
What AI coding costs you
Analysis of token costs and expenses for using AI coding assistants. Article stub without full detail.