2/26/2026
Unleashing Low-Bit Inference on Ascend NPUs: A Comprehensive Evaluation of HiFloat Formats
Evaluates HiFloat low-bit formats on Ascend NPUs for LLM inference. Compares INT8 and 4-bit floating-point for efficiency-accuracy tradeoffs.
2/26/2026
Evolutionary System Prompt Learning for Reinforcement Learning in LLMs
Evolutionary System Prompt Learning method jointly improves LLM contexts and weights via reinforcement learning. Enables autonomous self-improvement for agentic systems.
2/26/2026
MALLVI: A Multi-Agent Framework for Integrated Generalized Robotics Manipulation
Multi-agent LLM framework for robotic manipulation with closed-loop visual feedback. Integrates language and vision models for task planning in dynamic environments.
2/26/2026
OptiRepair: Closed-Loop Diagnosis and Repair of Supply Chain Optimization Models with LLM Agents
LLM agents diagnose and repair infeasible supply chain optimization models. Demonstrates closed-loop agent task decomposition for operations research problems.
2/26/2026
Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks
Research on prompt injection vulnerabilities in LLM agents via skill files. Identifies security risks in agent supply chains and skill-based attacks.
2/26/2026
Some Simple Economics of AGI
Economic analysis of AGI's impact on labor and growth. Argues human verification becomes the bottleneck as AI decouples cognition from biology.
2/26/2026
I built an open-source AI Gateway that sits between your apps and LLM providers
Self-hosted open-source API gateway supporting multiple LLM providers with per-token limits, statistics, and OpenAI compatibility.
2/26/2026
Show HN: PullMaster – Recommends code reviewers from your repo history
Developer tool using repo history to recommend code reviewers with awareness of availability and expertise.
2/26/2026
How to make LLM native User Interfaces - Post LLM Workflow
Discussion of UI design patterns optimized for LLM-native workflows post-foundational models.
2/26/2026
Show HN: ZSE – Open-source LLM inference engine with 3.9s cold starts
Open-source LLM inference engine optimizing memory efficiency and cold starts for serverless deployments.
2/26/2026
(paper money) Hedge Fund staffed by AI Employees (experiment)
Minimal post about hedge fund experiment staffed by AI employees using paper money. No technical details.
2/26/2026
Show HN: Bloomfilter – A service for AI agents to register and manage domains
Bloomfilter service allows AI agents to register ICANN domains via single API call using on-chain payments. MCP server compatible.
2/26/2026
Examining Bias and AI in Latin America
Study analyzing gender and social stereotypes in Spanish-language LLMs using 4,156 test questions from Latin American researchers.
2/26/2026
Show HN: WebMCP Core – AI agent tool definitions from any site
WebMCP Core tool converts websites into Model Context Protocol definitions for AI agents. Open source CLI with playground and A/B testing.
2/26/2026
Anthropic is dropping its signature safety pledge amid a heated AI race
Anthropic reportedly weakens AI safety principles amid competitive pressure. News article about company policy shift.
2/26/2026
Open-Source Agent Operating System
Production-grade open-source agent operating system written in Rust with 137K LOC, 14 crates, comprehensive testing.
2/26/2026
Claude Code Video Toolkit
Skills and MCP servers for Claude Code to generate videos programmatically using Remotion and FFmpeg.
2/26/2026
Which web frameworks are most token-efficient for AI agents?
Comparative analysis of web framework token efficiency for AI agent code generation.
2/26/2026
Show HN: RubyLLM:Agents – A Rails engine for building and monitoring LLM agents
Rails engine for building and monitoring LLM agents in production with cost tracking, retries, circuit breakers, and observability dashboard.
2/26/2026
Show HN: OpenTrace – Self-hosted observability server with 75 MCP tools
Self-hosted observability server exposing logs, database, and metrics as 75 MCP tools. Works with Claude Code and Cursor.
2/26/2026
Multi-agent workflows often fail
Analysis of multi-agent workflow failures, identifying three engineering patterns for reliable agent systems. Technical guidance on agent design.
2/26/2026
Show HN: Open-source MCP servers for self-hosted homelab AI
Open-source bundle of 8 MCP servers for homelab services (Proxmox, Grafana, Ollama, etc). 40 tools total, Python implementation.
2/26/2026
Nkmc – a virtual filesystem that lets AI agents call any API with ls, cat, grep
Nkmc virtual filesystem allows AI agents to call APIs using standard Unix commands (ls, cat, grep). Minimal details provided.
2/26/2026
Tool use and notation as shaping LLM generalization
Analysis of how tool use and notation reduce task complexity for LLMs rather than increasing model capability. Examines agent design patterns.
2/26/2026
Show HN: I built an LLM comment detector for HN (I got banned)
Developer built LLM comment detector for HackerNews after being flagged for excessive AI-assisted posting. Personal experience account.
2/25/2026
Show HN: Deff – Review AI-generated code changes
Deff tool streamlines review of AI-generated code changes. Surfaces diffs with vim motion support for faster comprehension.
2/25/2026
Show HN: Clerk – Simple invoicing for freelancers built with AI agents in 7 days
Clerk invoicing app built with AI agents in 7 days. Uses natural language chat for invoice generation and PDF parsing.
2/25/2026
Show HN: Provision Stateless GPU Compute with Claude Code's Remote Control
Claude Code MCP integration for stateless GPU provisioning across cloud providers with conversational control and cost optimization.
2/25/2026
Show HN: Edictum – Runtime governance for LLM agent tool calls
Edictum is a runtime governance library for LLM agents that enforces safety contracts at tool-call boundaries. Tested on 6 frontier models across 17,420 interactions, identifying a 'GAP' where models refuse harmful text requests but execute them via tool calls.
2/25/2026
Tldraw moves tests to closed source to avoid AI copies
Tldraw moves test suite to closed source to prevent AI-assisted reimplementation of open source libraries. Discusses implications for open source projects with commercial models.
2/25/2026
Show HN: Unworldly – A flight recorder for AI agents (tamper-proof, HIPAA)
Unworldly is a tamper-proof audit trail system for AI agents with real-time behavior monitoring, file/shell command tracking, and HIPAA/ISO 42001 compliance. Records and replays agent sessions.
2/25/2026
Show HN: Tesseract – 3D architecture editor with MCP for AI-assisted design
Tesseract is a 3D architecture editor desktop app with built-in MCP server for AI-assisted code visualization. Enables Claude integration to display codebase analysis visually rather than in text.
2/25/2026
Say No to Human-AI Workflow Segregation
Opinion on documentation quality for both AI agents and humans. Argues against segregating workflows between human and AI use, advocating unified documentation standards.
2/25/2026
Show HN: I challenged an LLM to find a hidden problem in my telemetry data [video]
LLM autonomously discovered hidden Rails performance bug in telemetry data using MCP server, then built dashboard and alerts. Demonstrates agent capability for observability analysis.
2/25/2026
Ask HN: Is meaningful privacy possible with hosted AI models?
Discussion thread on privacy tradeoffs when using frontier AI models, exploring options for accessing models without identity-linked accounts.
2/25/2026
Show HN: AI-runtime-guard – Policy enforcement layer for MCP AI agents
AI-runtime-guard is an MCP server enforcement layer that intercepts file and shell commands from AI agents before execution. Enforces policies without retraining or prompt engineering.
2/25/2026
Show HN: Running hallucination detection on a $200 GPU (RTX 3050, 4GB)
SIB-ENGINE detects LLM hallucinations by monitoring geometric drift in hidden states, achieving 54% detection with 7% false positives on RTX 3050 GPU with minimal overhead.
2/25/2026
Reinforcement Learning for LLMs
Intuition-first guide to reinforcement learning concepts behind RLHF, PPO, and GRPO. Explains RL principles for LLM alignment without heavy notation and mathematical density.
2/25/2026
Show HN: LLM Colosseum – A daily battle royale between frontier LLMs
Interactive battle royale simulation pitting Claude, GPT, Gemini, and Grok against each other in a game environment. Built with React, Canvas, Bun, and Hono.
2/25/2026
Deterministic Programming with LLMs
Discussion on deterministic programming approaches with LLMs for code generation, examining ethics and best practices for industry adaptation.
2/25/2026
Show HN: AgentGuard – Open-source EU AI Act compliance middleware for LLM apps
Open-source middleware for LLM applications ensuring EU AI Act compliance. Developer tool for regulatory requirements in AI systems.
2/25/2026
Show HN: AgentMD – CI/CD for AI agents, makes AGENTS.md executable
CI/CD platform for AI agents that executes AGENTS.md specifications with sandboxing, governance, and execution tracking. Open-source tool for agent workflow automation.
2/25/2026
Automated pentesting with MCPwner (finds 0-days)
Model Context Protocol server for autonomous security vulnerability discovery and exploitation automation.
2/25/2026
Hoofy– MCP server with persistent memory, adaptive pipelines, and a Clarity Gate
MCP server providing AI agents persistent memory, specifications, and adaptive pipelines with 32 tools in single Go binary.
2/25/2026
Swival – A coding agent for open models
Open-source coding agent for LM Studio and HuggingFace models with zero-setup local inference and custom Python implementation.
2/25/2026
From Postgres Migrations to AI Pipelines
Essay on software engineering philosophy transitioning from database migrations to ML pipeline design and systematic thinking.
2/25/2026
'Probably' doesn't mean the same thing to your AI as it does to you
Research on semantic interpretation differences between humans and AI models regarding probabilistic language and uncertainty expression.
2/25/2026
Connect your AI agent to every chat platform
Lightweight daemon enabling AI agents to communicate across multiple chat platforms: Slack, Discord, Telegram, WhatsApp, IRC, Matrix, Twilio, Zulip via single interface.
2/25/2026
Show HN: Opty – A Zig-based HDC that reduces token use by up to 90%
Zig-based MCP server using Hyperdimensional Computing to reduce token usage by up to 93% in LLM context windows.
2/25/2026
Show HN: Context Mode – 315 KB of MCP output becomes 5.4 KB in Claude Code
MCP server that compresses Claude Code tool outputs by 95% through sandboxed processing and summarization, supporting 10 language runtimes and SQLite FTS5 search.