Fragmented - AI Developer Podcast · Kaushik Gopal, Iury Souza

301 - The AI coding ladder

January 19, 2026·25 min·1 clip

An agent is a system that uses an LLM to reason through a problem, create a plan to solve that problem, and execute the plan.

Fragmented podcast hosts Kaushik and Yuri map three generations of AI coding paradigms: Generation 1 is probabilistic autocomplete (GitHub Copilot) replacing rule-based IntelliSense - trained on vast codebases enabling function-level generation from comments. Generation 2 is chat-oriented programming via Cursor IDE, which solved Gen1's copy-paste problem through UX (side panel, code highlighting, file tagging) enabling mainstream adoption but trapped users in synchronous interaction. Generation 3 introduces AI agents using the ReAct framework: LLM reasons, creates multi-step plans, and executes with tools - enabling asynchronous progress without synchronous back-and-forth. The episode positions agent-based coding as the frontier, enabling context maintenance and task decomposition without hallucination drift, establishing a technical foundation for understanding AI coding's evolution.

As heard by us

A clear map of AI coding's progression, from autocomplete to agents and thinking models.

The episode treats AI coding as a sequence of shifts rather than a single breakthrough. Kaushik and Yuri move from generation one, the super autocomplete era tied to GitHub Copilot, into agent loops and thinking models, which gives the conversation a clean sense of movement.

Read the full review in PlayNext →

Why you'd press play

Figure out which AI-coding rung youre actually on, from autocomplete to agents.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Fragmented - AI Developer Podcast