From Product Hunt
Native Mac apps and developer tools dominate today. Kuku brings a Tauri-based markdown editor that rivals Obsidian, and NetNewsWire 7 gets the Liquid Glass treatment.
Kuku — Native Markdown Editor for Mac
Built with Tauri, not Electron. Local-first with plain .md files, wikilinks, backlinks, and graph view. AI agent that edits files with Cursor-style reviewable diffs. Positioned as "Obsidian + Cursor, without the bloat."
NetNewsWire 7 — Free OSS RSS Reader
Version 7 brings Liquid Glass design update. Free, open source, native Mac and iOS. Supports iCloud sync and multiple services (Feedbin, Feedly, FreshRSS). Fast, accessible, Apple Silicon optimized.
Edit Mind — Self-Hosted Video Search
Index videos with AI (transcription, frame analysis, multi-modal embedding). Semantic search via vector DB. Privacy-focused, runs entirely locally. Docker support. Uses Whisper, YOLO, supports Gemini/Ollama.
Chrome DevTools MCP — AI Debugging for Claude Code
Official Chrome DevTools MCP server lets your coding agent control and inspect a live Chrome browser. Claude Code sees DOM, console, network, screenshots. Open source, 100% local.
PopStack — Dock-Style Folder Popovers
macOS utility that replaces native Stacks with dock-style folder popovers. Click desktop folder to open fast popover (icon grid or list view) without opening Finder. Optional auto-sort.
Komo Playbook — AI Agents from SOPs
Teach once via SOP, screen recording, or plain English. Komo learns goals and nuances, runs continuously via triggers (schedules, webhooks). 100+ tool integrations. Spawns parallel agents for complex workflows.
From Reddit
Google's Agentic Vision in Gemini 3 Flash, DeepMind's AlphaGenome in Nature, and Figure's Helix 02 shows full-body autonomy. Local LLM scene sees BitMamba bringing ternary weights to Mamba-2.
Google introduces Agentic Vision combining visual reasoning with code execution to ground answers in visual evidence. Enables more accurate computer use and visual analysis tasks.
AlphaGenome Published in Nature
DeepMind's AlphaGenome processes up to 1 million DNA base pairs in a single context window. Predicts thousands of genomic signals including gene expression, splicing, chromatin structure. Outperforms prior models on 25/26 benchmark tasks. Critical for rare disease, cancer, gene therapies.
Figure AI Helix 02 — Full-Body Autonomy
Most advanced humanoid robot from Figure. Unified neural network enables full-body autonomy. Completes 4-minute dishwasher task autonomously. Walks while carrying delicate objects. Three neural systems: learned movement from 1000+ hours human motion data, sensory integration, semantic reasoning.
LM Studio 0.4.0 — Major Release
Introduces llmster for headless server deployment. Parallel processing via continuous batching. New /v1/chat stateful endpoint with MCP support. Split View for concurrent chat sessions. Built-in documentation. New lms chat CLI command.
Quick hits from the local LLM scene.
BitMamba-2-1B — 1.58-bit Mamba-2 Model
Ternary weights {-1, 0, 1}. Runs on CPU at 50+ tok/s on Intel i3. 621 MB RAM for 1B model. Apache/MIT licensed.
ByteDance Stable-DiffCoder-8B
First diffusion-based coding model. 8B parameters, 8192 context. New architecture direction for code generation.
From Reddit
ComfyUI gets a professional 3D posing environment, NVIDIA releases FastGen for diffusion models, and world models continue to advance with open-source LingBot-World.
VNCCS Pose Studio for ComfyUI
Professional 3D posing and lighting environment running entirely within ComfyUI. Interactive viewport with bone manipulation gizmos, undo/redo. Dynamic body generator with age/gender/weight/muscle sliders. Advanced lighting control. Free, open source.
LingBot-World — Open-Source World Model
World simulator from video generation. Maintains high fidelity across diverse environments (realism, cartoon, scientific). Long-term memory with minute-level horizon. Real-time interactivity under 1s latency at 16fps. Applications in gaming, content creation, robot learning.
NVIDIA FastGen Library
Plug-and-play library for turning diffusion models into few-step generators. Supports SD 1.5, SDXL, Flux, WAN, CogVideoX.
SkyReels V3 — Multi-Subject Video
New video models with audio-guided generation (A2V), multi-subject from reference images, and V2V. 19B params.
From Reddit
Mac utilities, developer tools, and self-hosted solutions. Canvas-based task management, Spotify metadata for self-hosting, and a markdown-based Postman alternative.
Forma — Canvas-First Tasks & Notes
Spatial organization instead of lists. Drop ideas anywhere, move them around, group visually. 100% local, native macOS, 3MB. One-time purchase $9.99.
Self-Hosted Spotify API Clone
Self-hosted Spotify API emulator using Anna's Archive dataset. 30GB parquet files for 256M tracks with albums/artists. FastAPI app for local music tagging and discovery. Like having Spotify's metadata on your own machine.
Voiden — Markdown-Based Postman Alternative
Offline-first, git-native API tool. Requests stored as .void Markdown files. No cloud dependencies, no accounts. Works completely offline.
More Mac utilities and developer tools.
LiquidFetch — Visual System Fetch
Native macOS system info app. Graphical neofetch alternative. Free.
Boom Video — OBS Alternative for Mac
Real-time screen recording with OBS-level control. Native, no Electron. 50% off this week.
Parall v2.0 — Dock Shortcut Bundles
Launch multiple app instances with separate data. Each shortcut gets its own icon and environment.
Sonexis — System-Wide Audio Enhancer
Free audio enhancer for macOS. Real-time effects via visual chains. Preset sharing in v1.0.1.
Hexa — Menu Bar Hex Calculator
Shows hex/decimal/binary at once. Evaluates expressions live. Open source, free.
Pam — CLI/TUI for SQL Databases
Minimal SQL client for terminal. Multi-database, Vim navigation, BubbleTea UI. Export as CSV, JSON, SQL.