kwindla (@kwindla) header
kwindla (@kwindla) profile picture

Infrastructure and developer tools for real-time voice, video, and AI. @trydaily // ᓚᘏᗢ // @pipecat_ai

San Francisco, CA
Joined September 2008
Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)Photo by kwindla (@kwindla)
When the model guy and the harness guy pull up in their casuals for a talk, you know it’s gonna slap.
1
1
24
4,388
Replying to @aiDotEngineer
@aiDotEngineer Concierge is now live! Talk to the concierge and build your custom schedule for the next three days along with a list of people ( speakers ) who you should definitely meet! This was built on the go with @pipecat_ai and @GradiumAI Try it out: aie-concierge-web.onrender.c…
One shotted a fun voice agent with @pipecat_ai and @GradiumAI to figure out with sessions to attend based on my profile at @aiDotEngineer . My agent recommended @danielhanchen ( ofcourse! ) and few more. Should I ship this so that everyone coming to conf can find their way around? Wdyt @swyx ?
7
18
4,482
Replying to @lmazare
@lmazare and I are in SF this week for @aiDotEngineer . Two talks, including one with @kwindla, where we'll try to design the system and neural architecture that will power the next generation of voice agents. And one workshop where you'll build your digital clone to make phone calls on your behalf!
3
1
15
3,221
kwindla retweeted
The AI automatically renamed my ship from "Shippy McShipface" to "Shippy McCourrierface" when I bought a courrier, AGI confirmed 🤯
Sub-agents in (latent) space! We’ve been working on a side project. As far as I know, this is the first massively multiplayer, completely LLM-driven game. Come play Gradient Bang with us. See if you can catch me on the leaderboard. This whole thing started because I wanted to explore a bunch of things I’m currently obsessed with, in an application of non-trivial size, that felt both new and old at the same time. So … a retro-style space trading game built entirely around interacting with and managing multiple LLMs. Factorio, but instead of clicking, you cajole your ship AI into tasking other AIs to do things for you. Some of the things we’ve been thinking about as we hack on Gradient Bang: - Sub-agent orchestration - Partial context sharing between multiple LLM inference loops - Managing very long contexts, and episodic memory across user sessions - World events and large volumes of structured data input as part of human/agent conversations - Dynamic user interfaces, driven/created on the fly by LLMs - And, of course, voice as primary input If you’ve been building coding harnesses, or writing Open Claw agents, or doing pretty much anything that pushes the boundaries of AI-native development these days, you’re probably thinking about these things too! This is all built with @pipecat_ai, the back end is @supabase, the React front end is deployed to @vercel, and all the code is open source.
2
1
8
3,025
We @minimax_ai are excited to co-host the private dinner tonight with @kwindla~ If you missed our talk yesterday @aiDotEngineer and are interested to learn about M2 and our Voice model Speech 2.6, apply here. Link is in the comments.
I'm in NYC for the @aiDotEngineer Code Summit and am hosting a small dinner on Friday night for people who are working on voice AI, multi-agent systems, or both. Holding a couple of seats for people I haven't met before. DM me if you want to join us; tell me what you're building!
1
4
12
3,547
This macOS local voice agent repository from @kwindla with @pipecat_ai via @trydaily is the only local voice stack I’ve tried that 'just works' out-of-the-box. It even supports interruptions (see video)! I got this pointing to gpt-oss hosted by @ollama, and it works flawlessly.
Here's voice agent code that runs entirely locally on macOS: github.com/kwindla/macos-loc… Models you can run locally are getting better and better. You still need a pretty high-end machine to run an LLM that's good at conversation and tool calling. In my opinion, the ~30B parameter models are now good enough to be really interesting. But depending on what you're doing you might find a smaller LLM works well, too.
1
8
1
24
4,779
I'm so excited to be speaking at the Daily Voice AI Meetup on Wednesday evening here in SF. I'll be going deep on our new Mist v2 TTS model, with some accompanying hot takes on "agents", I'm sure! Link ⬇️ Come hear me speak. I'll be joined by speakers from @OpenAI and @trydaily. Big thank you to @kwindla and the team for having me! And to @Vapi_AI for hosting!
1
1
2
519
Pipecat 0.0.54 is out! 🔥 Performance improvements, a new task management system to track asyncio tasks within Pipecat, an initial Unit Test framework for testing frame processors and more improvements and bug fixes! 🔥 @pipecat_ai github.com/pipecat-ai/pipeca…
2
1
8
1,613
🫱🏻‍🫲🏼 Last night I have the pleasure of attending the Voice and AI Meetup - Jan 2025 at Chroma HQ in San Francisco 🗣️ 🔥 Hot takes from the all-star panel: 🤖 @krandiash — @cartesia "Advances in Robots in 2025" 🧠 ​Niamh Gavin — Stanford University "The next frontier isn't just knowledge recall - it's real-world understanding and reasoning" 🎯 @shresbm — Gemini API "Multi-modal AI is unlocking unprecedented capabilities" 🚀 @swyx — @latentspacepod "Perfect memory becomes a choice in 2025" 💡 @kwindla - @trydaily "The future is hybrid: seamless on-device + cloud AI" 🎯 Industry Impact Alert: - Voice AI already outperforming median human CSRs - Healthcare disruption incoming: AI showing stronger patient interactions than average practitioners - Founders actively deploying admin-facing voice AI across legal, insurance, accounting & therapy
1
1
3
466
A lot of talented people working in the voice ai space! Great meetup hosted by @trychroma @trydaily and @kwindla, discussing the latest in multimodal ai. Was happy to travel out for the event and get the opportunity to share @withsen with a few of the people in the room.
1
1
6
401
kwindla retweeted
my favorite genre of deep youtube: @achowdhery dropping some llama 4 alpha? 231 views on a <2k channel LS Discord goated for deep finds
7
18
3
370
42,704
kwindla retweeted
Developer meetup Tuesday! Join @swyx, @shresbm @GoogleDeepMind, @krandiash @cartesia_, Niamh Stanford, @kwindla @pipecat_ai @trydaily 🍕 🎙️ 🛠️ 🤖
3
8
2
15
8,970
kwindla retweeted
I'm really, really proud of the team I work with at @trydaily . It's exciting to see @pipecat_ai on bigger and bigger stages.
Announced at #CES2025: New #AI Blueprints for #agenticAI empower developers to build AI agents with a single click.⚙️ Explore blueprints ➡️ nvda.ws/4iYbzEJ. Co-created with @crewAIInc @trydaily @llama_index @langchain and @wandb.
2
10
712
Pipecat 0.0.51 is now available. This is a quick bug-fix release that fixes an infinite reconnection issue with websocket-based TTS services. More good stuff coming soon! @pipecat_ai github.com/pipecat-ai/pipeca…
1
7
449
The Gemini 2.0 era begins with 2.0 Flash Experimental release ⚡️ 📈2.0 Flash beats 1.5 Pro across factuality, reasoning, coding, math. 📳 More modalities - image and audio out (in EAP) 🔧 Native tool use for Google Search, code execution and 3P functions 🆕 a new multimodal, realtime API experience 🎬 3 new cool starter apps - spatial understanding, video analyzer and map explorer
We just released Gemini 2.0 Flash Experimental ⚡ Available in the Gemini API and Google AI Studio for testing, it allows developers to build interactive experiences with better performance and multimodal capabilities. goo.gle/3BriaXd
6
14
181
13,832
What's coming to @pipecat_ai this week: Input audio filters (for noise reduction), output audio mixers (for background sounds), @tavus twins, garbled audio fixes, avoiding premature LLM interruptions and more!
2
1
6
527
happening tomorrow!!!!!! lu.ma/r1ntmmh2
Probably going to be the best diffusion party of the year. If you are in SF at 18th and want to join a crazy team of diffusion people including the creators of Stable Diffusion series, FLUX and AuraFlow, join us!!!
2
1
9
3,551
Replying to @hassaanraza
@hassaanraza97 from @tavus live with @kwindla from Daily, Matty from @HelloCivitai, and Brandon from @cartesia at the Daily + @ProductHunt meetup!
2
5
630
When @karpathy announced he is launching @EurekaLabsAI and he mentioned that subject matter experts would not be able to personally tutor all 8 billion people on demand we thought we would prove him wrong... Demo: educationbot.cerebrium.ai/ #aiagent #GenerativeAI #education
6
17
3
60
10,708