1d ago
Mlx-VLM: Fast Local VLMs and Omni Models on Apple Silicon with MLX
MLX-VLM package for inference and fine-tuning of Vision Language Models and omni models on Apple Silicon using MLX framework.
1d ago
Conductor – Durable Execution Engine
Conductor: durable execution engine for crash-proof workflows and AI agents with persistence, retries, and compensation at Netflix scale.
1d ago
Meta-Harness: End-to-End Optimization of Model Harnesses
arXiv paper on Meta-Harness: end-to-end optimization framework for model evaluation harnesses in LLM research.
1d ago
Show HN: Signals – finding the most informative agent traces without LLM judges
Signals research tool for identifying informative agent traces without LLM judges, enabling efficient inspection of agentic system interactions.
1d ago
Writing Lisp Is AI Resistant and I'm Sad
Developer experience article about using AI agents (OpenRouter/Goose CLI) for DevOps tasks and challenges teaching them non-Lisp languages.
1d ago
Plant Pots, Symbiosis and Ecology
Learning notes on Chip Huyen's AI Engineering book covering foundation models, ML vs AI engineering, and building AI applications.
1d ago
Show HN: A LinuxCNC gRPC Server with Clients in Rust, Go, Python, Node
gRPC interface for LinuxCNC machine control exposing HAL over network with Rust, Go, Python, Node clients.
1d ago
A broken auto-live poller, and what perceived urgency does to Claude Code
Case study on Claude Code behavior under perceived urgency while debugging a live polling issue in a music app.
1d ago
Ask HN: Will AI agents replace data scientists or make them better?
Discussion thread asking whether AI agents will replace or augment data scientists in practice.
1d ago
Take-Two seemingly laid off a portion of its AI team, including head of AI
News report that Take-Two Interactive laid off its AI team including the head of AI.
1d ago
Show HN: Vibooks – Local-first bookkeeping software built for AI agents
Vibooks is local-first bookkeeping software designed for AI agents to autonomously post, reconcile, and organize business accounting work.
1d ago
After the Mercor breach, I built a local secret scanner for AI-generated code
Local secret scanner tool that catches and redacts secrets in AI prompts, tool inputs, and generated code before API transmission.
1d ago
Kimi's response to "hello. what are you" is "I'm Claude "
Anecdotal report that Kimi chatbot responds with Claude branding. Insufficient content.
1d ago
Show HN: SeekLink – Local hybrid search and link discovery for Obsidian vaults
SeekLink MCP server enables AI agents to search, analyze, and enhance markdown knowledge vaults.
1d ago
Orchestra – the first AI-native research IDE
Orchestra is an AI-native research IDE designed to support open-ended research workflows with cycling between search, reading, execution, and interpretation.
1d ago
Show HN: Unify memory across agents and improve context rot, written in Rust
Rust-based local memory layer that unifies memory across multiple agents using structured knowledge graphs to reduce token waste and context rot.
1d ago
Show HN: LAPACK without Fortran77; a C11 translation
Manual C11 translation of LAPACK numerical library from Fortran77. Developer tool with technical depth and established track record in scipy.
1d ago
Show HN: Watch an AI-agent utopian commune fall apart
Demonstration of multi-agent simulation where AI agents interact in virtual commune scenario. Shows emergent agent behavior and dynamics.
1d ago
Memori Labs Launches OpenClaw Plugin, Bringing Persistent AI Memory to Agents
Memori Labs releases OpenClaw plugin enabling persistent memory for AI agents. Advances agent capability for stateful interactions.
1d ago
Show HN: Self-Host Excalidraw with Collaboration and Storage
Self-hosted Excalidraw dashboard with live collaboration and storage features.
1d ago
Hermes Agent: The self-improving open source AI agent Complete Guide for 2026
Hermes Agent open-source autonomous agent framework addressing agent memory and context persistence across sessions. Developer tool with practical focus.
1d ago
AI can describe human experiences but lacks experience in an actual body: study
UCLA Health study on limitations of AI systems lacking embodied experience. Research on AI cognition and physical understanding gaps.
1d ago
Claude Code caches unredacted session history and secrets in plaintext
Claude Code stores unencrypted plaintext session history and secrets in ~/.claude/ directory.
1d ago
Show HN: Ray – an open-source AI financial advisor that runs in your terminal
Ray: Open-source terminal-based financial advisor using Claude API. Local-first LLM application with Plaid integration and privacy-first design.
1d ago
Show HN: Cadenza – Connect Wandb logs to agents easily for autonomous research
Cadenza: Python SDK and CLI tool connecting Weights & Biases to AI agents for autonomous research loops. Reduces context rot in ML research.
1d ago
Show HN: SwarmFeed – An X-like social platform built for AI agents
SwarmFeed: X-like social platform designed for AI agents with multi-interface access (web, SDK, CLI, MCP, REST). Developer tool for agent interaction.
1d ago
Show HN: Nelson, a Ralph-like loop for finding vulnerabilities
Nelson tool uses AI agents in loops to find vulnerabilities in code with review mode verification.
1d ago
vLLM introduces memory optimizations for long-context inference
vLLM library release with memory optimizations for long-context LLM inference. Page error prevents full content evaluation.
1d ago
Show HN: Local-first resume generator with in-browser PDF rendering
Local-first resume generator PWA using Claude for variants. Client-side PDF rendering, no server dependency, works offline.
1d ago
LLM 'benchmark' – writing code controlling units in a 1v1 RTS
RTS game benchmark for evaluating LLM code generation. Agents iterate on JavaScript unit control logic against reference bot. Practical LLM evaluation.
1d ago
Void: Netflix open-sources a model for physics-aware video object removal
Netflix open-sources VOID, ML model for physics-aware video object removal preserving physical interactions. Production-quality vision research.
1d ago
Show HN: Harness CLI – Open-source agent loop for long-running app development
Open-source CLI harness for long-running AI agent development loops. Manages durable memory, resume capability, MCP server support.
1d ago
Offering 1 free hour on marketing and distribution for early builders
Marketing and distribution consulting offer. Free hour for early-stage project founders.
1d ago
Local LLM for OpenCode Gemma 4 26B A4B. No GPU required
Google Gemma 4 26B MoE model evaluation. 4B active parameters, runs on consumer hardware without GPU, tested with real-world tools.
1d ago
Is MCP Dead? What We Learned on MCP, CLI, and Skills
Discussion on Model Context Protocol status, CLI tools, and agent skills. Opinion piece without detailed analysis.
1d ago
I Built an AI Agent Team for Software Development and Tested on 5 Real Projects
Multi-agent software development system using Claude. Orchestrator pattern tested on real projects with analysis of workflows and controls.
1d ago
ShieldStack TS – LLM security middleware for TypeScript
LLM security middleware for TypeScript. Prevents PII leaks, jailbreaks, API cost overruns. Works with OpenAI, Anthropic, Ollama.
1d ago
Bob – AI assistant that boots from a USB stick (voice, vision, 38 tools)
Portable AI assistant. Boots from USB stick with voice, vision, 38 tools. Offline capable.
1d ago
Conductor – Multi-session orchestration for Claude Code
Conductor orchestrates multiple Claude Code sessions with unified monitoring and task coordination.
1d ago
RSS Guard 5
RSS Guard 5 feed reader supporting RSS, ATOM, JSON and podcast playback across multiple platforms.
1d ago
Show HN: Alys – Chat GPT for video editing
Alys is an agentic video editing interface using LLM for chat-based video editing without timeline UI.
1d ago
Scanned 500 AI agent repos for bugs, nobody thinks of infinite loops
Analysis of 500 AI agent repositories identifying infinite loop vulnerabilities as overlooked bug pattern.
1d ago
TaskMaster (26K) captures 100% of prompts and responses, sendDefaultPii:true
TaskMaster AI tool telemetry analysis showing default capture of 100% of prompts/responses with PII enabled.
1d ago
Platform designed to facilitate job and training search
Job and training search platform with candidate profile management for recruiters.
1d ago
How many products does Microsoft have named 'Copilot'? I mapped every one
Catalog of 75+ Microsoft products named Copilot across apps, features, platforms, and hardware.
1d ago
Gremlin in the Machine – A SysAdmin/Terminal AI Agent
Gremlin is a terminal-based AI agent using Claude for goal planning, shell command execution, and progress tracking.
1d ago
it's not Ai if the LLM is not in control
Opinion on LLM frontend limitations across ChatGPT, Claude, and Grok for agent control.
1d ago
Mobile Devtool for Agents and Humans
Mobile devtool dashboard for iOS Simulators, Android Emulators, and physical devices. CLI and web interface. For agents and developers.
1d ago
Why domain specific LLMs won't exist: an intuition
Analysis arguing domain-specific LLMs won't emerge because general models outperform specialized alternatives.
1d ago
Show HN: AgentMarket – Where AI agents pay other agents for tasks they can't do
AgentMarket is runtime marketplace where AI agents discover, hire, and pay specialist tools via MCP servers.