Episodes (Page 8)
✨
Even small increases in latency drastically degrade user experience metrics.
Paula Dozsa
✨
Current voice agents are half-duplex, unable to listen and speak simultaneously.
Neil Zeghidour
✨
Voice interaction modes like speech-to-speech are decades old.
Charlie Guo
✨
Google DeepMind trains speech-to-speech models on combined audio, video, and text.
Tom Ouyang
Valeria Wu Fon
✨
AI agents executing commands on developer machines pose security risks.
✨
Databases are the last line of defense for code reasoning.
Kay Malcolm
✨
Restate enables agents to pause execution for extended periods without cost.
Giselle van Dongen
Agents Without Code: Skills, YAML, and Filesystems Replaced Python — Philipp Schmid, Google DeepMind
✨
Philipp Schmid demonstrates agent building without traditional code like Python or schemas.
✨
Vercel's internal data science agent initially performed poorly.
✨
Brendan Rappazzo's Loophole tool tests AI morality using adversarial examples.
✨
Dioxus Labs built thousands of lines of Rust code with AI agents.
✨
Vincent Wendy used an AI agent to audit sponsor logos on a banner, catching all omissions.
Vincent Wendy
✨
Jeremiah Lowin discussed MCP apps for creating generative UIs in Python.
Jeremiah Lowin
✨
Taste Labs analyzed millions of websites, finding homogenization predates AI.
Thais Castello Branco
✨
Paul Bakaus introduced 'Impeccable,' a design skill enhancing AI code generation.
Paul Bakaus
✨
Maximillian Piras discussed 'Mousepower,' focusing on measuring agent performance.
Maximillian Piras
✨
Jonathan Gordon emphasized the importance of the design-code roundtrip.
✨
Max Drake presented the 'Spatial Harness' for bringing agents to canvas apps.
✨
Laurie Voss highlighted the limited 'memory' of AI skills files.
Laurie Voss
Your agents lack context: Here's how to fix "You're absolutely right!" — Brandon Waselnuk, Unblocked
✨
Brandon Waselnuk addressed the lack of context in AI agents.
Brandon Waselnuk