Meet Brad, Tavus’ real-time AI avatar that listens, sees, and responds in milliseconds, slashing latency to under one second utterance-to-utterance. Tavus’ API lets you design avatars with unique personas, connect to external tools, and interact asynchronously via webhooks. Watch for a peek into the future of AI-human interaction.
In this One-Shot episode, Maxwell Wang, SpaceX software engineer and Dispatcher creator, shows how a drone runs real-time vision on-board while using cloud LLMs for semantics—voice-driven commands translate via function-calling into actions, with hybrid on-board GPU and cloud compute enabling hands-free control, target tracking, and return home.
Join this One-Shot with Kwindla Kramer, Daily CEO, as we unpack Pipecat—an open-source framework that abstracts away infrastructure for real-time voice agents. Real-time LLMs under 800 ms, hybrid local-cloud AI, and multi-model pipelines. Check Daily’s GitHub and build your first voice agent with the AI Tinkerers community.
Varun Pant and team won the grand prize at the AI Tinkerers Humans-in-the-Loop Hackathon in Seattle for SmartNourish, a health-monitoring system fusing CGM data and genetic data with LangChain and LangGraph and Anthropic Claude to predict, advise, and adapt through a real-time, human-in-the-loop agent architecture to tackle prediabetes risk.
AI Tinkerers showcases Yohei's Baby AGI 2—an agent that dynamically generates, tests, and improves its own toolset. It searches API docs, generates code from scratch, stores and catalogs new functions, and uses existing ones as context. Levels 0–3 span tools to anticipatory function creation. Bonus: an agent-focused investment fund.
Neon’s serverless Postgres uses a Rust storage layer that separates compute from storage and copy-on-write, letting AI agents spin up isolated databases instantly. Thousands of databases daily; Neon Authorize adds auth and row-level security for apps. Watch session and spin up free DB in under 5 minutes at neon.tech.
Launched today, Stagehand is Browserbase's open-source tool that tightens AI LLM browser control with a simple three-command interface: act, extract, observe. It offers headless browser infra and seamless LLM integration for fault-tolerant, practical automation. In an interview with Paul Klein, we explore its AI-use cases and the open-source, agent-friendly abstraction.
Top practitioners showcased focused, in-depth AI project sessions featuring live demos and technical deep dives for global attendees.
Join the exclusive AI Tinkerers One-Shot as E2B unveils an $11.5M seed round, with Vasek Mlejnsky detailing a secure cloud runtime and sandboxed AI code execution for agents and apps; includes a live Perplexity AI data analysis and visualization demo and a speed ~150 ms goal.
Discover how MultiOn builds autonomous AI browser agents for price comparisons and shopping. In this One-Shot episode, we cover MultiOn's API and Chrome extension, the tech and challenges of generalized browser agents, and embedding agents in your projects. A glimpse under the hood of MultiOn. Watch: https://www.youtube.com/watch?v=L0lX4iqXBcA
