Autogenetic AI Agents and the Future of Ruby Development - RUBY 682

·1 hr 1 min·5 clips

Valentino built a Ruby gem where AI agents can self-assemble to accomplish tasks, reusing created agents for future work.

The episode begins with host Charles Maxwood and guest Valentino Stoll introducing themselves and their current work, with Valentino at Gusto working on AI features and Charles at Prize Picks. Valentino explains his exploration of autogenetic AI, meaning self-generating, through his Ruby gem 'agentic'. He built this to enable plan-and-execution workflows where an LLM creates a plan and executes it, with the system assembling agents as needed. Valentino describes how agents can be reused for future tasks, and the system saves artifacts for later use. He shares an experiment where he tasked the system with building a self-assembling agent system in Ruby, which created agents like a coding agent and project manager before hitting limits. The discussion shifts to the evolution of AI, with Valentino predicting trends like self-healing systems and agent-to-agent protocols, referencing Google's early work. He talks about the blurring line between local and production environments as AI tools run more on personal machines. Valentino introduces the concept of progressive disclosure for coding agents to manage context limits, comparing it to CLI tools. When asked about production use, he says it's not yet cost-effective due to model limitations but sees potential as models improve. The conversation covers how AI changes development work, with Valentino noting developers now write more prompts than code but still need conceptual understanding of systems. He advises that newcomers can use AI to learn but should rely on traditional courses for foundational concepts. Valentino mentions his other projects, like an AI software architect and seed box for LLM memory exploration. Charles shares his picks, including the game Infiltrators and AI tools like Cursor's plan feature, and promotes his upcoming 'Code for 75' challenge and training sites railsgeniuses.com and aidevgeniuses.com.

As heard by us

A Ruby Rogues episode on AI agents, shifting interfaces, and what changes when the system starts doing the work.

Ruby Rogues spends this hour on AI agents and agentic systems, with Valentino Stoll discussing his move from Doximity to Gusto and the AI work he is doing there before the panel turns to what happens when a system starts to analyze, act, and improve.

Read the full review in PlayNext →

Why you'd press play

You want Ruby talk about AI agents that gets into what changes when software can take actions on its own.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Ruby Rogues