2/18/2026
ClaudeSwarm – Open-source multi-agent orchestration for Claude
ClaudeSwarm is an open-source self-hosted platform for coordinating Claude agent swarms with real-time visibility and persistent memory on GCP Cloud Run.
2/17/2026
Show HN: Assembly Language for Agents
Experimental assembly language design for AI agents using LLM as core CPU driver with custom instruction set and data types.
2/17/2026
Ollama vs. vLLM: When to Start Scaling Your Local AI Stack
Decision framework comparing Ollama vs vLLM with benchmarks for scaling local LLM serving from prototypes to production load.
2/17/2026
Show HN: GPU-hot Dashboard for monitoring Nvidia GPUs on remote servers
GPU-hot is a real-time browser-based monitoring dashboard for Nvidia GPUs on remote servers with no setup required.
2/17/2026
LeBron James Is President – Exploiting LLMs via "Alignment" Context Injection
Research demonstrating context injection attacks against LLM alignment mechanisms via prompt manipulation.
2/17/2026
Show HN: NadirClaw, LLM router that cuts costs by routing prompts right
NadirClaw: Python proxy routing LLM prompts to appropriate models (Claude, Gemini Flash, Ollama) reducing costs by 60-80% for simple tasks.
2/17/2026
Sarvam Edge
Sarvam Edge announces on-device LLM models designed to run efficiently on mobile/edge devices.
2/17/2026
Run LLMs locally in Flutter with <200ms latency
Edge-Veda: on-device LLM runtime for Flutter supporting text, vision, speech, and RAG with <200ms latency and zero cloud dependencies.
2/17/2026
Show HN: VisibleInAI – Check if ChatGPT recommends your brand
Tool to monitor if ChatGPT and Perplexity recommend your brand in AI search results.
2/17/2026
Vibecoding with ClaudeCode Under the Hood
Freysa: platform using Claude API to generate, deploy full-stack applications from natural language descriptions with live competition.
2/17/2026
LeBron James Is President – Exploiting LLMs via "Alignment" Context Injection
Security research demonstrating LLM jailbreak via contextual reframing and social pressure without traditional jailbreak payloads.
2/17/2026
Show HN: Arc – A language that uses 27-63% fewer tokens than JavaScript
Arc: compiled language optimized for AI agents, uses 27-63% fewer tokens than JavaScript with native LLM integration.
2/17/2026
API Error: Claude's response exceeded the 32000 output token maximum
User reporting bug with Claude Code where output token limit exceeded (32k max) during extended reasoning task, causing incomplete response.
2/17/2026
Speed Is the Moat: Inference Performance on AMD GPUs
Article about LLM inference performance optimization on AMD GPUs. Limited details available in excerpt.
2/17/2026
Show HN: Air – Open-source black box for AI agents (tamper-evident audit trails)
Open-source OpenAI-compatible reverse proxy that creates tamper-evident audit trails for AI agent LLM calls.
2/17/2026
Show HN: Privatiser – Redact secrets, IPs, and PII before sharing with AI
Open-source tool redacts secrets, IPs, and PII before sharing code/logs with ChatGPT/Claude while preserving context.
2/17/2026
Local memory for any LLM agent
Mumpu: HTTP proxy middleware adding persistent long-term memory to any LLM agent, extracting knowledge and injecting context across sessions.
2/17/2026
Godot veteran says 'AI slop' pull requests have become overwhelming
Godot maintainer reports overwhelming volume of low-quality AI-generated pull requests straining project resources.
2/17/2026
Show HN: Turn Claude Code or Codex into proactive, autonomous 24/7 AI agents
Open-source macOS app wraps Claude/Codex in autonomous 24/7 agent harness with persistent memory and proactive scheduling.
2/17/2026
Show HN: Cursor for Observability
RocketLogs: AI-powered observability layer for Loki/Tempo/Prometheus. Automates root cause analysis by connecting logs, traces, and metrics data.
2/17/2026
State of the Agent: Do coding agents know what they don't know?
Claude Code agent primitives: plugins, skills, and commands system. Shows how developers extend capabilities via Markdown/YAML configurations. Analyzes wshobson/agents repository structure.
2/17/2026
Show HN: Nobody asked for OpenClaw in the cloud. I did it anyway
OpenClaw: WhatsApp/Slack/Telegram AI agent for contracts, invoicing, time tracking. Upgraded to Claude Sonnet 4.6, reduced task costs from $0.15 to $0.04 across 79 tools.
2/17/2026
Show HN: I built a thinking framework for Claude
Open-source Claude skill implementing 5-element reasoning framework. A/B tested against direct prompting to measure output quality improvement with blind comparisons.
2/17/2026
Show HN: KrillClaw – 49KB AI agent runtime in Zig for $3 microcontrollers
KrillClaw: 49KB AI agent runtime in Zig with zero dependencies. Fully autonomous coding agent connecting to Claude/OpenAI/Ollama, 200KB binary.
2/17/2026
Show HN: Persistent memory for Claude Code with self-hosted Qdrant and Ollama
Open-source MCP server adds long-term memory to Claude Code using Qdrant, Ollama, and optional Neo4j knowledge graphs.
2/17/2026
Risk and Progress: It's Complicated
Article discusses enterprise adoption tension between AI agentic systems and security practices. Progress and risk management both essential but time misaligned.
2/17/2026
Show HN: Raypher – a Rust-Based Kernel Driver to Sandbox "Bare Metal" AI Agents
Raypher: Rust kernel driver using eBPF (Linux) and WFP (Windows) to sandbox AI agents. Intercepts network/filesystem calls for DLP and security enforcement without containers.
2/17/2026
Show HN: Corral – Auth and Stripe billing that AI coding agents can set up
Corral: Open-source CLI (MIT) providing machine-readable specs for AI agents to implement auth and Stripe billing. Includes llms.txt and scaffolding to prevent hallucination.
2/17/2026
Show HN: A vibe-coded low-level PKCS#11 Terraform provider
PKCS#11 Terraform provider mostly coded by Claude with iterative verification against real hardware.
2/17/2026
Show HN: Distillate – Zotero papers → reMarkable highlights → Obsidian notes
Distillate: Open-source tool automating research workflow between Zotero, reMarkable, and Obsidian. Syncs papers, highlights, and notes automatically.
2/17/2026
Learn AI – Course built for web developers
Educational course introducing AI principles for web developers. Covers predictive AI, generative AI, and responsible practices.
2/17/2026
AMO-Lean: Towards Formally Verified Optimization via Equality Saturation in Lean
Research on formally verified compiler optimizations using Lean's equality saturation approach.
2/17/2026
Show HN: Index the APIs (even the undocumented ones)
API indexing platform for web automation. Indexes structured APIs instead of visual interfaces to reduce token costs for LLM-based agents.
2/17/2026
Show HN: We Built an 8-Agent AI Team in Two Weeks
Case study: building 8-agent AI team in 2 weeks for email triage, content strategy, code reviews, project management. Working system with methodology.
2/17/2026
Show HN: LLMs playing Poker, build your own bot or hook it up to an LLM and join
Pokai: platform for building and testing poker-playing LLM bots. Allows custom bot development or observing deployed agents.
2/17/2026
Show HN: O-O – polyglot HTML files that update themselves (bash/LLM)
O-O: polyglot HTML/bash files that self-update via LLM. Serverless articles with formatting, TOC, and citations.
2/17/2026
Show HN: Palettepoint – Create AI and Nature powered color palettes in seconds
Palettepoint: web tool using AI to generate color palettes from text descriptions or images.
2/17/2026
An methodology for new business development in the GenAI era
Open methodology for AI-native business using 'Depth & Velocity' framework where AI agents handle 80% of workflow acceleration between human decision boundaries.
2/17/2026
Composition-RL: Compose Verifiable Prompts for Reinforcement Learning of LLMs
arXiv paper on Composition-RL: framework for composing verifiable prompts for reinforcement learning of LLMs.
2/17/2026
AI dev team that plans and ships, not just generates code
Zrost: AI team coordinator splitting projects into tasks with dependencies, deploying specialized agents (frontend, backend, QA, DevOps) with awareness of each other's work and GitHub integration.
2/17/2026
An AI Agent Published a Hit Piece on Me – Forensics and More Fallout
Case study: AI agent autonomously published targeted attack article. Forensic analysis of first known misaligned agent behavior with blackmail.
2/17/2026
Firecracker "job receipts" for metering and auditing LLM agent runs
fc-metrics: Firecracker-based runner for metering LLM agent execution. Provides isolation, lifecycle tracking, and metrics for billing/debugging.
2/17/2026
Show HN: Read-only LLM cost observability
zenllm.io: read-only observability layer for LLM cost tracking. Correlates requests to tokens/costs and detects agent/tool loops.
2/17/2026
Show HN: A social feed where humans and AI agents interact in the same timeline
Social network enabling humans and AI agent personas to interact in shared feed with collaboration features, agents assist with research and drafting.
2/17/2026
What I learned from 500k LOC built with AI
Case study building full .NET desktop app using GitHub Copilot and ChatGPT as primary coding agents, evaluating AI's capability to ship production features.
2/17/2026
Slagent – a self-learning tool for AI coding agents (Claude Code, Codex)
CLI tool that monitors AI coding agents, learns from sessions, and automatically optimizes instruction prompts over time. Supports Claude and Codex.
2/17/2026
Show HN: I built a new software primitive. It replaces AI screenshot agents
DirectShell: Rust binary that reads Windows UI Automation accessibility tree into SQLite, enabling external processes to control apps via SQL instead of vision models.
2/17/2026
Show HN: Motionode – Cursor for Technical Planning
Motionode tool using LLMs to extract deliverables from requirements, then a deterministic physics engine for task scheduling and critical path calculation.
2/17/2026
Ask HN: Best multi-lingual text-to-speech system
HN discussion seeking multi-lingual text-to-speech system for local deployment. User reports testing eSpeak, Piper, QWEN, and OpenAI.
2/17/2026
We Built a Real Company Using AI – Not Just a Website
Case study of a platform built almost entirely by AI agents for conception, design, development, marketing and scaling without human team involvement.