From Product Hunt
Local-first AI tools dominate today. Media search, memory layers, and browser automation — all running on your machine.
Clipto — Fully local, natural language search over terabytes of media
Like Google Photos, but fully local. Tags people, dialogue, and scenes across video/audio/files. Search by describing what you're looking for. M1+ Mac with 24GB+. Indexed 2TB in 24 hours on M5.
Second Brain for AI — Persistent memory for Claude, ChatGPT & Cursor via MCP
Self-hosted memory layer on Cloudflare Workers (free tier). Store context once, recall by meaning across any MCP client. Includes Obsidian plugin, Chrome extension, CLI. One-click deploy, MIT licensed.
TabTasker — 50+ offline utilities running 100% in your browser
Edit PDFs, process images, transcribe audio, fix JSON — all via WebAssembly and ONNX Runtime. Zero uploads, everything stays on your device. Free, no signup.
Web Clipper for NotebookLM — Clip in, export out
One-click save any page, PDF, AI chat, Reddit thread, or YouTube playlist to NotebookLM. Export flashcards to Anki, mind maps to Obsidian Canvas, reports to Word/PDF. 25K+ users.
Uindow — Programmable browser with local LLM automation
Record automations with no-code recorder, optionally call a local LLM for AI reasoning. Trusted interactions, exported as .js.yaml. Free for personal use, source-available on GitHub.
From Reddit
Anthropic's Dynamic Workflows unlock massive parallelism in Claude Code, Microsoft readies its own coding model, and Anthropic overtakes OpenAI in valuation.
Update: Dynamic Workflows lets Claude Code orchestrate up to 1,000 parallel subagents (16 concurrent). The plan lives in script variables, not Claude's context, so only the final answer returns to your session.
Jarred Sumner (Bun creator) used it to migrate ~750K lines of Rust in 11 days. SWE-bench Pro up to 69.2%, USAMO 2026 math at 96.7%. Requires Claude Code v2.1.154+, available on Max/Team/Enterprise.
Microsoft to Unveil Homegrown Coding Model for GitHub Copilot at Build
Project Polaris — Microsoft's first in-house coding model, debuting at Build 2026 (June 2-3). Led by Mustafa Suleyman's division, features chain-of-thought reasoning for multi-file refactoring. Part of broader move to reduce OpenAI dependence.
Anthropic Surpasses OpenAI as Most Valuable AI Startup at $900B
Raised $30B+ at $900B valuation, surpassing OpenAI's $852B. Q2 2026 revenue expected at $10.9B (doubled QoQ). Annualized run rate targeting $50B by end of June. IPO exploration as early as October.
Gemini 3.5 Pro Expected June 2026 — Flash Already Shipping
Sundar Pichai confirmed "give us until next month" at I/O 2026. Flash is GA since May 19. Pro entering Vertex AI allowlist for enterprise GCP customers. Prediction markets price late-June launch.
OpenClaw Crosses 300K GitHub Stars — Run Your AI Assistant Locally
Personal AI assistant on your own devices across WhatsApp, Telegram, Slack, Discord, iMessage, and 15+ channels. Bring your own API key (Claude, DeepSeek, GPT). Fastest-growing GitHub project ever.
MLX 0.31 + macOS 26.2 Unlock M5 Neural Accelerators — 4x Faster TTFT
Major local AI runtime updates: MLX unlocks M5 Neural Accelerators. Ollama 0.24.0 adds Gemma 4 MTP speculative decoding. vLLM v0.21 stabilizes DeepSeek V4 on Blackwell.
From Reddit
AMD goes all-in on ComfyUI acceleration, Kling 3.0 dominates video generation, and ComfyUI ships audio + geometry support.
AMD x ComfyUI: Professional-Quality GenAI on Ryzen + Radeon
AMD partners with ComfyUI for native ROCm acceleration on consumer Radeon GPUs and Instinct MI355X. Official ROCm support now on Windows for ComfyUI Desktop. MI355X outperforms B200 for ComfyUI workflows.
Kling 3.0 Leads Video Generation Arena — 2127 Elo Score
As of May 31, Kling v3 by Kuaishou tops text-to-video (2127 arena score, 731 blind votes). Seedance 2.0 Fast at 1993, Happy Horse 1.0 at 1962. Veo 3.1 praised for realism + native audio.
ComfyUI v0.22.0 — Stable Audio 3.0, HiDream-O1, MoGe Models
Released May 20. Stable Audio 3.0 support, MoGe geometry processing, HiDream-O1 with area conditioning, LTX2.3 VRAM optimization, downscaled IC-LoRA for LTXV.
From Reddit
Developer tools for the AI coding era — repo guardrails, JSON fixers, and a full open-source alternative database.
Open Source Alternative Finder — Self-Hosting Ratings for 60+ Tools
Daily-updated comparison site: pricing tables, self-hosting difficulty (1-5 scale), step-by-step migration guides, and AI verdicts. Built by a solo dev on $0/month infra (GitHub Actions + Groq AI + Pages).
Harness Starter Kit — Repo Guardrails for AI Coding Agents
Turn fragile AI coding prompts into durable repository rules. AGENTS.md, drift checks, failure memory, adoption reports. Works with Claude Code, Cursor, Codex, Copilot. Prompt-first approach.
JSON Kit — Fix Broken AI JSON + 6 More Browser Tools
Fixes trailing commas, single quotes, Python literals, markdown fences, truncated output from LLMs. Also: formatter, JSON-to-TypeScript, JSON-to-Zod, JSONPath, diff. 100% local, free.