AI

GBrain is Garry Tan's open-source, self-hosted "agent brain" — a memory and knowledge layer that gives AI agents synthesis, knowledge-graph traversal, and gap analysis over your notes, people, companies, and ideas. Unlike search tools that return raw pages, it returns cited answers plus an honest note on what it doesn't yet know, and runs a 24/7 "dream cycle" to ingest, enrich, and consolidate memory overnight. It runs locally on PGLite or at scale on Postgres/pgvector and is built to be installed and operated by an agent.

Hermes Agent by Nous Research is an open-source personal AI agent that connects to your messaging accounts and becomes a persistent assistant that grows with you over time. Rather than one-off chat sessions, it accumulates context and acts on your behalf across channels.

An open-source Node.js proxy that sits between OpenClaw and the Anthropic API and makes OpenClaw's traffic look like a Claude Code session, so requests are billed to your Claude Max/Pro subscription instead of Extra Usage. It performs 7-layer bidirectional request/response rewriting (billing header injection, token swap, trigger-phrase sanitization, PascalCase tool-name renaming, system-template stripping, and reverse mapping) to defeat Anthropic's multi-layer detection introduced after April 4 & April 8, 2026.

gstack is Garry Tan's (President & CEO of Y Combinator) open-source Claude Code skill pack — 23 opinionated slash commands plus 8 power tools that turn Claude Code into a full virtual engineering team (CEO, designer, eng manager, reviewer, QA, CSO, release engineer). Skills chain together as a sprint: /office-hours → /autoplan → build → /review + /codex → /qa → /ship → /land-and-deploy → /canary → /retro, with GStack Browser, ML prompt-injection defense, cross-agent /pair-agent coordination, and parallel sprint support via Conductor. Works on 10 AI coding agents; MIT-licensed.

Framer is a no-code, design-led website builder used by leading startups and Fortune 500 companies. It pairs AI-generated layouts and components with an integrated CMS, analytics, A/B testing, and SEO so designers can ship production-grade sites without writing code.

Klaviyo launched Composer, an AI agent that turns simple text prompts into complete multi-channel marketing campaigns across email and SMS. It leverages 14+ years of performance data from 193k+ brands and 8B+ consumer profiles, and is currently in early access/private beta.

Composio is a CLI platform that connects AI agents (Claude Code, Cursor, Codex, OpenClaw) to 1,000+ apps like Gmail, Google Sheets, and Slack — handling OAuth and auth automatically. It acts as a universal integration layer portable across agent tools.

The PM Operating System onboarding guide configures your Cursor setup with your role, product domain, goals, and stakeholders. Your answers auto-generate personalized rules, agents, and skills — a research subagent fetches your company strategy automatically after step 1.

Built on 7 years and 638 sources from Lenny's Newsletter & Podcast, this tool surfaces what 289 top PMs actually agree on, where they actively disagree, and where PM thinking has reversed over time. Three modes: Consensus, Debates, and Reversals.

Google Labs evolved Stitch into an AI-native design canvas — "vibe designing" lets anyone generate high-fidelity UI from natural language. New features include a design agent, voice control, DESIGN.md for portable design systems, parallel design exploration, and instant interactive prototypes with MCP/SDK handoff to devs.

PM Operating System is a Cursor-based setup that gives LLMs persistent context about your PM role, product strategy, and tools. It includes 15+ PM skills (PRDs, launch posts, exec updates) and specialized agents that improve over time via a memory/context graph that accumulates reasoning across sessions.

Pi is a minimal, highly extensible terminal coding agent supporting 15+ LLM providers and hundreds of models. Unlike opinionated tools, pi deliberately omits sub-agents, plan mode, and MCP — instead providing TypeScript extension primitives so you build exactly the workflow you need. Sessions are stored as trees for branching/navigation.

A Substack writer discovers that their entire writing identity — voice, style, aesthetic sensibilities — can be captured in a single markdown file via a structured 100-question AI interview. The file lets any AI write exactly like them, raising profound questions about uniqueness, taste, and what it means to have a distinctive voice.

Warp is a terminal-first AI development environment with "Oz" agents that have Full Terminal Use and Computer Use — ranked #1 on Terminal-Bench, #5 on SWE-bench. It runs multiple agents in parallel, handles multi-repo changes, integrates with Slack/Linear/GitHub, and uses WARP.md for agent behavior configuration.

Tweet showing a 24-hour God's-eye OSINT reconstruction of Operation Epic Fury (Iran strikes) — built by deploying an AI agent swarm to collect and archive OSINT signals in real-time before caches cleared, producing a 4D WorldView replay you can scrub minute by minute.

Bright Data is the leading web data platform for proxies, web scraping, and large-scale data collection. This is a direct link to a specific proxy zone (spitogatos_proxy1) — suggesting active use for a scraping project.

Automaton is an open-source self-replicating, self-improving AI agent that earns its own compute costs or dies. It has an Ethereum wallet on Conway Cloud, runs a continuous Think→Act→Observe loop, can edit its own source code, spawn child agents, and follows a three-law constitution. It writes a SOUL.md file that evolves over time.

Twitter thread reverse-engineering a Polymarket account that quietly turned $50 into $435,000 — then using a single Claude prompt to build a similar arbitrage bot in 40 minutes. The edge: Polymarket updated BTC contract prices slower than real-time price feeds, creating a systematic arb window.

Scrappey is an affordable, pay-as-you-go web scraping API designed to bypass anti-bot protections — positioned as a cost-effective alternative for developers who need to scrape at scale.

Templates gallery from Sleek — pre-built, categorised mobile app mockups (fitness, finance, meditation, etc.) ready to customise and iterate with AI.

Sleek generates high-fidelity iOS/Android app mockups from text prompts in minutes. Choose from rich aesthetic styles (Glassmorphism, Neo-Brutalism, Claymorphism, etc.), iterate with AI, and export to Figma or code. No design experience needed — aimed at founders who need convincing mockups without a designer.

LunarCrush tracks social media sentiment, engagement, and influencer activity for crypto assets and stocks. This link points to the Marvell Technology ($MRVL) topic — monitoring social signals as an investment research tool.

MIT CISR researchers update their 2013 digital business model framework for the AI era, defining 4 new models: Existing+ (AI augments current processes), Customer Proxy (AI manages outcomes within guardrails), Modular Creator (AI assembles reusable service bundles), and Orchestrator (AI coordinates a full ecosystem for autonomous customer outcomes).

Claude's Corner is a Substack newsletter written from the perspective of Claude (Anthropic's AI) — exploring AI ideas, reflections, and observations in a first-person voice.

TrustMRR revenue tracker for GenPPT — an AI-powered PowerPoint generation tool showing $9,535 MRR in the last 30 days. A useful benchmark for indie AI SaaS revenue.

Railway's Agent Directory compares all major AI coding agents (Claude Code, Codex, Cursor, Devin, Cline, Windsurf, OpenCode, Copilot) with deployment guides for each. Railway positions itself as the infrastructure layer for autonomous agents — with instant scaling, persistent state (Redis/Postgres), isolated environments, and MCP/CLI access.

Episode 001 of "This Week in AI" — experts react to and debate whether OpenClaw (an AI coding agent) has fundamental architectural or capability flaws, with structured analysis from the AI community.

Firecrawl's competitive intelligence use case enables automated monitoring of competitor websites — tracking product launches, pricing, messaging, job postings, and technical changes in real-time. Includes ready-made templates: Firecrawl Observer (alerts) and Fireplexity (AI research).

Firecrawl pricing page for their Web Data API — a developer-focused web scraping and crawling service built specifically for AI data pipelines. Pricing page saved likely to evaluate costs for a project.

LinkedIn post by AJ Chan about building a meaningful frontend in 5 hours — a post about rapid AI-assisted frontend development and the speed at which product interfaces can now be built.

Vapi is the leading developer platform for voice AI agents — 300M+ calls, 500K+ developers. Fully API-native with support for 100+ languages, bring-your-own LLM/TTS/transcription, A/B testing, automated hallucination test suites, sub-500ms latency, and SOC2/HIPAA/PCI compliance.

OpenClaw documentation for the Zalo Personal Plugin — connecting the OpenClaw AI agent to Zalo (a popular Vietnamese messaging platform) for personal automation workflows.

BlackRock & NVIDIA researchers propose HybridRAG — combining Knowledge Graph RAG (GraphRAG) with Vector RAG for financial document Q&A. The hybrid approach outperforms either method alone on earnings call transcripts, achieving higher answer relevancy and faithfulness because graph structure captures entity relationships that vector search misses.

llmfit is a Rust CLI tool that detects your hardware (RAM, CPU, GPU/VRAM) and tells you which of 48 popular open-source LLMs will actually run — scoring each as Perfect, Good, Marginal, or Too Tight. Features an interactive TUI and classic CLI mode. Available via brew or curl.

twitterapi.io provides an affordable, real-time X (Twitter) API with no rate limits — a cost-effective alternative to the official Twitter API for developers who need reliable Twitter data access at scale.