Hermes ᯅ

@fullstackizzy Let’s go!

What a day at the NYC Voice AI Meetup! The sun was out and the energy was high. Big thanks to Agora for hosting and to my friend @hermes_f always a blast catching up! I had some fantastic conversations about Skills and voice agents with some truly brilliant people: Gonzalo Adrover and Andres Niño. Great venue, great food, and even better company. Can’t wait for the next one! 🚀

📅 Tuesday 4/21 🎙️ NYC Voice AI Meetup by @hermes_f luma.com/a6la9xte

@akshay81844 It’s got all the vibes

95% of Github repos in 2026 be like:

if you’re in NYC this week and don't know where to connect with people in AI 🗽 check out these events (4/7 - 4/13)👇

📅 tuesday 4/7 🎙️ NYC Voice AI Meetup by @hermes_f lu.ma/680blwvk

@HackingDave 🙋‍♂️ I’ve been an avid Claude Code user and totally agree, over the last few weeks it feels like it got slower and code quality has dropped. A coworker recommended I try Codex and Im impressed with its ability to catch issues Claude missed, and it codes 5x faster

Dude Claude is total trash - seen massive degrading of code quality, bugs, and more over the past several weeks. This week, I can’t even use it or rely on it to complete basic bug fixes or implementations. Codex has been performing substantially better. Anyone else ?

Cursor is coming for v0, lovable, and bolt.

Two great events back-to-back this week: the @AgoraIO + @DeepgramAI NYC Voice AI Meetup and the Idea Makers pitch competition part of Penn State's Startup Week. Really appreciated everyone who came out Tuesday. Spent the night talking with teams building voice agents and AI companions. You can feel how fast things are moving. People aren’t just experimenting anymore, they’re shipping. On Wednesday, I had the privilege of serving as a judge for the Bardusch Family IdeaMakers Challenge pitch competition. Thanks to @ISTatPENNSTATE team for including me! Always inspiring to see what students are building, from early ideas to thoughtful, well-executed products. Different stage, earlier ideas but the same builder mindset: put something real in front of others, and learn fast. That’s the common thread for me this week: Whether you're deep in production or just getting started, momentum comes from building and putting it out into the world. Grateful to be part of both communities. Looking forward to more conversations ahead.

@hasantoxr This needs to add Voice orchestration. They should use @AgoraIO for it

🚨 BREAKING: CHINA just released a Python framework for building AI agents. 100% OPEN SOURCE. It has visual agent design, MCP tools, memory, RAG, and reasoning. All built in. All working together. It's called AgentScope. You describe your agent system. It builds the architecture, wires the tools, and runs the whole thing. You come back and there's a working multi-agent pipeline. Not a prototype. Not a demo. The actual system. Not a wrapper. Not a chatbot builder. A full Agent-Oriented Programming framework that thinks in agents from the ground up. Here's what it does out of the box: → Visual agent builder so you design your entire system before writing a single line of code → Native MCP tool support, plug any external tool directly into any agent in your pipeline → Built-in memory so every agent remembers context, decisions, and history across sessions → RAG pipeline ready to connect your own documents, databases, and knowledge bases → Reasoning modules that let agents plan, reflect, and self-correct without human input → Multi-agent coordination so your agents collaborate as a system, not a pile of isolated API calls Here's how it thinks: You define your goal. AgentScope maps the agent roles. Each agent gets its tools, its memory, its reasoning layer. They coordinate. Results flow back up. You get a finished output. A single complex task might route through a planner agent, a researcher agent, a coder agent, and a critic agent, each doing its job, then converge into one clean deliverable. Here's the wildest part: AgentScope is built by Alibaba DAMO Academy. The same lab behind Qwen. They didn't assemble this from existing pieces. They designed the entire framework from first principles around how agents actually need to think, remember, and work together. Most frameworks give you building blocks. AgentScope gives you an architecture. The community has already started plugging it into data pipelines, research workflows, and full automation systems the team never planned for. 100% Open Source. Apache 2.0 License.

@zodchiii You missed @AgoraIO’s Skills for building real-time voice agents and video streaming apps.

@PatrickHaede Agent

We just mass automated social marketing. Introducing Superscale Agent - the first advanced AI agent for social marketing. What used to take 1000s of hours now takes minutes: → Brainstorm & execute full marketing strategies instantly → Deep-dive competitor & trend reports (connected to the entire web, TikTok trends, Meta Ad Library) → Analyze your own Meta & TikTok ad accounts directly → Generate 100s of ads for TikTok, FB, IG, or Google from a single prompt → Iterate on creatives at insane speed → Build e-commerce store & ad assets on autopilot You give instructions. The agent does the work. Software engineering went agentic. Today, social marketing follows. This is the most complex product we have ever built, and our most advanced update to @superscale_ai - ever. Early customers have been using it for months. The results have been transformative. To celebrate: comment "Agent" and get our 100 most powerful prompts + 3,000 free credits (= 3 videos or 50 static ads). It only gets crazier from here 🚀

@KissMyAds @beffjezos thats you fam.

@beffjezos @hermes_f 👌🏼

How bro starts moving after hitting $1k MRR

@techhalla Awesome work! I saw the spritesheet generator and thought it was interesting, but an entire game engine 🤯

Me: I'm just gonna vibe code a quick game. Also me: accidentally builds an entire AI-powered game engine. This is gonna be wild 👇

NYC builders showed up last night. Huge thanks to everyone who came out to the @AgoraIO’s Convo AI Meetup in the East Village. Great conversations with teams building voice agents and AI companions. We talked about everything from latency in voice pipelines to real-world challenges with turn detection, interruptions, and streaming audio. It’s always energizing hearing what teams in the NYC dev community are building. Thanks to everyone who came out and shared what they’re working on.

@fullstackizzy @AgoraIO Thanks so much for showing up Izzy! Always a great time catching up. It’s been too long!

I had an amazing time at the NYC Voice AI Meetup! The gloomy, rainy weather didn’t stop the shine and energy. It was incredible to see so many people still show up. A massive thank you to the hosts, @AgoraIO and my good friend @hermes_f . It’s always a blast whenever I get a chance to speak with him! I met some truly great people and absolutely loved the discussions about Skills, voice agents, and so much more.

This is super cool to see, but I’d be hesitant to trust it in production. WebRTC on its own is great, but it really doesn’t perform at scale. As an app scales and users/usage increase, you’ll need TURN/STUN/SFU or there are going to be issues. @AgoraIO solves these issues and doesn’t require complex backends, use your existing TanStack setup & sub Agora in for the realtime.

Building realtime voice AI is usually a pain: WebRTC, audio streams, tool execution, state 😵 TanStack AI now handles that w/ useRealtimeChat()🎙️ 🔊 Voice in / Voice out 🧠 Tools/State 🖼 Multimodal ⚡ Realtime transcripts Supports OpenAI Realtime & ElevenLabs 🔗⬇🧵

@thisdudelikesAI Super interesting. The number of browsers built for agents is more than the options a human has. I wonder how this compares to Cloudflare and Vercel’s new headless agent browsers?

🚨BREAKING: Someone just open-sourced a headless browser that runs 11x faster than Chrome and uses 9x less memory. It's called Lightpanda and it's built from scratch specifically for AI agents, scraping, and automation. Not a Chromium fork. Not a hack. A completely new browser written in Zig. Here's why this changes everything for AI builders: ↓

Given Cloudflare’s dominance in the market (especially with Bot detection) I’m curious to see how others can perform if CF gives their browser open access (no verification required) Also curious how CF will handle this traffic attribution for their customers to understand bot crawler vs human users

@hermes_f cloudflare just launched their version too. competition in this space is good, means better tools for everyone

🚨 Someone built a tool that turns any website into clean data your AI can actually use. Give it a URL. It crawls every page. Hands you back perfect markdown. It's called Firecrawl. The web data API that every AI app has been missing. Here's the problem it solves: You paste a URL into ChatGPT. It hallucinates half the content. You try scraping with BeautifulSoup. You get HTML soup with ads, navbars, and cookie banners mixed into your data. Firecrawl fixes this. One URL in. Clean, structured, LLM-ready data out. No sitemap needed. No scraping scripts. No parsing headaches. Here's what it does: → Scrape a single page into clean markdown → Crawl an entire website. Every subpage. Automatically → Extract structured data with a schema you define → Handle JavaScript-rendered pages (SPAs, dynamic content) → Bypass anti-bot protections → Output as markdown, HTML, or structured JSON Here's why everyone building with AI needs this: → Building RAG? Firecrawl turns any documentation site into your knowledge base → Building an AI agent? Give it the ability to read any website properly → Doing competitor research? Crawl their entire site in minutes → Training a model? Convert hundreds of pages into clean training data → Building a search engine? Firecrawl is literally what Perplexica uses under the hood SDKs for Python, Node, Go, and Rust. Integrates with LangChain, LlamaIndex, CrewAI, Dify, and more. Self-hostable. Or use the hosted API. 100% Open Source. AGPL-3.0 License.