Raw Content

Four Predictions for How AI Will Change Software in 2026

Source: Every podcast episode featuring Dan Shipper and Brandon Gell

Key Predictions

1. Agent-Native Architecture Will Reshape Software Design

Software will be redesigned around AI agents as primary users, operating through three capability levels:

Level 1: Agents access all user-facing features, mimicking human interaction with applications
Level 2: Agents leverage hidden backend capabilities, accessing functions not exposed in standard interfaces
Level 3: Agents implement changes directly—fixing bugs, adding features, or personalizing experiences

Dan notes that Anthropic and Notion are already exploring this paradigm where “both humans and agents are first-class citizens” rather than treating AI as an afterthought.

2. Designers Will Become Power Users of AI Tools

Non-technical creators are transitioning into builders. Design leaders like Every’s Lucas Crespo now develop small applications using AI assistance. However, Brandon raises a practical concern: mainstream adoption requires abstraction of code complexity. Tools like Cursor must hide technical barriers to reach designers at scale.

3. A New Software Engineering Role Emerges

Three distinct developer types are emerging:

Engineers using AI as acceleration within traditional workflows
“Vibe coders” building without deep comprehension
“Agentic engineers” who primarily orchestrate AI agents, focusing on problem definition and agent direction rather than writing code

This represents a fundamental redefinition of software development work.

4. AI Training Will Prioritize Independence

Current alignment training emphasizes predictability and obedience. True autonomy requires different approaches—allowing agents to explore, make mistakes, and learn continuously. Dan predicts 2026 will bring “new training approaches and architectures” enabling longer-running, more independent agent behavior.