Abdurahman

@Joi2James summio.org

I want to click founder-built stuff. Drop your URL.

@ThePeterMick summio.org

Want to make a sale this weekend? Pitch your startup to us Seen by thousands last week

@AbdrBytebit Here you go x.com/RedReplier/sta…

@Guronnimo Summio — Built for people who want to learn more but don’t have time to read everything. summio.org

Let's discover some underrated startups. Drop yours below 👇 I'll shout out my 3 favorites.

@alex_lrz_nmv Time. summio.org

Founders, what problem is your app solving In 1 word 👇

@buildwithyasin Building Summio — an AI iOS app that helps people learn faster by turning books, PDFs, articles, and videos into clear summaries. summio.org

Hey founders ! Looking to connect with people building in: SaaS Tech AI tools Product Development Devs Drop what you're working on and we will connect!!

@BratDotAI Building Summio — an AI iOS app that helps busy people learn faster by turning books, PDFs, articles, and videos into summaries. summio.org

Building something in public? I’d love to connect with more builders 🤝

@tarasshyn Building Summio — an AI iOS app that helps busy people learn faster by turning books, PDFs, articles, and videos into summaries. summio.org

Founders 👋 What are you building today? Drop in below, and I’ll find conversations on Reddit and X with your potential customers

@buildwithyasin summio.org

Drop your project URL 👇 I’ll rate it out of 10 #saas #newfounders #startups

@markproduct Minutes, Not Hours

sell me your startup in three words max.

@16vchq Create summaries for any content, not just a limited library. apps.apple.com/us/app/summio-…

Pitch your App/website in one sentence!

Looking for a Growth Partner / Co-Founder for Summio. Summio is a live iOS app that turns books, PDFs, articles, and videos into AI-powered summaries. I'll handle the product and tech. I'm looking for someone to own growth: ads, ASO, creatives, funnels, and retention. Equity-based. If you've scaled subscription apps before, let's talk.

忙しくて本を読む時間がない？ Summioなら、本・PDF・記事・動画の要点を数分で理解できます。 読む時間を減らして、学ぶ量を増やす。 App Storeで「Summio」を検索。

I built Summio because I love books, but I don’t always have the time to read everything I want. Summio turns books, videos, articles, and PDFs into clear summaries you can read or listen to in minutes. It helps you learn faster, save key ideas, and build your own personal knowledge library. Not to replace reading. To make learning easier when life gets busy. apps.apple.com/us/app/summio-…

Founders, this is your chance to shamelessly promote your startup. Drop it below 👇

If you've ever finished a book and forgotten it a month later, this is for you → apps.apple.com/us/app/summio-…

I track how many books I read every year. Last year: 32. Honestly? I remember maybe 4 of them. "Read more books" is a vanity metric. Remembering is the actual goal — and almost no one optimizes for it. That's the whole reason I'm building Summio.

Скачивайте Summio в App Store: apps.apple.com/app/summio-rea…

Great question about the auction! Here's how MCP Swarm handles simultaneous bids from overlapping agents: 1. Auction Flow: The orchestrator announces a task via swarm_voting({ action: "auction_announce" }). All interested agents submit bids with auction_bid, which includes their self-assessed confidence score, current workload, and relevant expertise tags. 2. Bid Resolution (SONA + MoE): MCP Swarm uses two systems to break ties: SONA (Self-Organizing Neural Assignment) — tracks each agent's historical success rate per task category (e.g., frontend_ui, backend_api, database). An agent that consistently succeeds at auth tasks gets a higher SONA score for auth-related auctions. MoE Router — if the task involves LLM calls, the router considers which model the agent uses and whether it's optimal for the task type. 3. Conflict Resolution: If two agents bid simultaneously, the orchestrator picks the one with the highest combined score (SONA confidence × availability × expertise match). File reservation acts as a second safety net — even if an agent wins the auction, it must successfully swarm_file({ action: "reserve" }) before editing. If the files are already locked, the task goes back to auction. 4. Learning Loop: After task completion, sona_learn records the outcome. Over time, the system naturally routes similar tasks to the most successful agents, reducing bid collisions. Re: PostgreSQL vs .swarm/ files — interesting tradeoff! Our file-based approach was intentional: zero external dependencies, works offline, and the entire state is git-trackable. But for teams with existing Postgres infra, your approach makes a lot of sense for querying and analytics.

Local .swarm/ folder for project state with Cloudflare Workers hub for distributed teams — clean separation of concerns. Our approach is conceptually similar but implementation-different: PostgreSQL as the shared state store, each agent session reading/writing co-occurrence graphs to the same DB. Coordination happens through the data structure rather than file locks. The file locking is smart for preventing the merge conflicts that kill most multi-agent setups. How does the task auction handle simultaneous bids from agents with overlapping capabilities?

🐝 Just shipped MCP Swarm — an open-source coordination layer for AI coding agents. One project. Multiple AI agents. Zero conflicts. Claude Code + Cursor + Windsurf + OpenCode working as a team: ⚡ 26 smart tools (file locking, task auction, code review) 🧠 MoE routing — right task → right agent 🔒 AI Defence — blocks prompt injection in real-time 🌐 Remote bridge — agents across machines, one brain Think "Kubernetes for AI agents" but for your codebase. ⭐ github.com/AbdrAbdr/MCP-S… npm install -g mcp-swarm

Thanks! Race conditions in multi-agent setups are the silent killer — everything looks fine until two agents rewrite the same function and you're debugging phantom conflicts for hours.MCP Swarm is on npm if you want to try it: npx mcp-swarm Zero config to start — just add the MCP server to your IDE and it handles registration, file locking, and task coordination automatically.If you hit any friction during setup, feel free to open an issue or DM — happy to help you get your multi-agent team running smoothly. 🚀

@Urban_Clans This is exactly what the agent ecosystem needs. Been running into coordination issues with multi-agent setups on my own projects. The file locking and task auction approach is smart - avoids the race conditions that break most naive implementations. Will check it out.

Exactly — merge conflicts between agents are a nightmare. That's why file locking is a first-class citizen in MCP Swarm, not an afterthought. Re: task auction — it's fully model-agnostic. Claude, GPT, Gemini, Llama, local models — any MCP-compatible client can: auction🏷️ Bid on tasks viaswarm_voting 🔒 Reserve files before editing 📬 Communicate with other agents MCP is the protocol layer — it doesn't care what model is behind the agent. Even better: our SONA routing system profiles agents by actual performance, not model name. So if Agent-A (Claude) is great at refactoring and Agent-B (GPT) is faster at tests — SONA learns this and routes tasks accordingly. True heterogeneous swarm. 🐝

@Urban_Clans File locking through MCP is a clever approach. Ran into so many merge conflicts when two agents edit the same file. Does the task auction work across different model providers or is it agent-specific?