Google's AI sprint: Fast models, agentic tools, but delivery gaps frustrate

May 20, 2026 · Lenny's Podcast
🎧 PodShort 14 min squeezed to 2 AI SprinklerAS AI / ML New
Full episode from Lenny's Podcast
Quotable Moments

If you look at 3.5 Flash, you're getting, according to Google's benchmarks, a super smart model, sort of a Codex 4.7 model if you like 4.7, at the speed of something much more like their 3.1 Flash model. So it's super, super fast and super smart.

Claire VoProduct Leader and AI Obsessive
01:20

Google is the source of truth for so much personal information and for companies that use Google Workspace for company information, and I've seen a lot of this in Claude Code and Claude Co-work. People are using the MCP connectors to build artifacts and apps on that data source, and Google is going straight for owning that experience themselves.

Claire VoProduct Leader and AI Obsessive
07:15

The promise is really good for some of these things, but the reality is, if you're not able to use them or they're broken on the day, then people are going to lose patience for some of this.

Claire VoProduct Leader and AI Obsessive
12:55
Key Insights
  • Google announced the Gemini 3.5 family of models, including Gemini 3.5 Flash, which is their fastest and smartest coding model, four times faster than previous models.
  • Google is heavily investing in agentic capabilities, with new features in their IDE (Antigravity 2.0) and CLI that allow agents to perform long-running tasks and ask clarifying questions.
  • The 'Grill Me' slash command in Antigravity is designed to aggressively ask clarification questions to truly understand user requirements, aiming for a more direct approach than existing tools.
  • Google AI Studio now allows users to build apps connected to Google Workspace apps, leveraging personal and company information for productivity and personal assistant use cases.
  • Gemini has a redesigned consumer-facing AI product with improved image generation using Nano Banana and a new video generation model called Omni, capable of creating longer, more consistent, and photorealistic videos.
  • Omni, the new video generation model, allows for conversational editing of videos, changing elements like structures into bubbles, and maintaining character consistency across scenes, making it a powerful video editing tool.
  • Google Flow, a more prescriptive video editing tool, embeds Omni and focuses on cinematic quality production, enabling users to define and reuse characters, and even create avatars of themselves.
  • Despite promising features, many of Google's new announcements, particularly for Google AI Studio and Flow, are not yet fully functional or widely accessible, leading to user frustration.
Metrics Mentioned
  • four times as fast (Gemini 3.5 Flash is four times as fast as other coding models like 5.5, Opus 4.7, and 4.6.)

RevBots.ai View:

  • AI Sprinkler teams bolt on Gemini 3.5 but hit integration walls with Workspace.
  • Agentic features signal shift toward ARM-style autonomous workflows.
  • Video gen tools appeal to SaaS Hoarders adding yet another siloed capability.
  • Google's delivery delays show even tech giants struggle with AI execution.
🎧Full Episode:Lenny's Podcast →