Framework for 'Agentic Engineering' Emerges
A design framework called the "Universal Tool Design Cheatsheet" is gaining traction for building AI tools that serve both creative and technical users. The framework outlines a process-driven approach that includes understanding the user's problem, setting clear boundaries for the AI, and providing transparent results. This methodology emphasizes user empowerment by making the AI's contributions explicit and auditable rather than "magical."
- Agentic frameworks like LangGraph, CrewAI, and Microsoft's AutoGen provide structured approaches for building multi-step AI workflows, moving beyond single-prompt interactions to more complex, goal-oriented tasks. LangGraph, for instance, uses a graph-based structure giving developers tight control, while CrewAI is designed for production systems with role-based agent delegation. - The concept of "agentic AI" is shifting development from manual coding to automated, intelligent workflows where AI agents can independently plan, code, test, and deploy applications, potentially reducing manual workloads by up to 60%. This paradigm treats the AI as an active partner that can manage repetitive tasks, allowing developers to focus on strategic problem-solving. - In creative fields like photography and design, human-AI collaboration is being framed as a partnership where AI handles tedious tasks like culling images or organizing assets, allowing professionals to remain in a "flow state" longer. Adobe is integrating agentic AI into its Creative Cloud and Express platforms to act as a creative partner, scaling a single user's output and accelerating creative work. - For developers, AI IDEs and terminals are becoming central to new workflows. Windsurf, a fork of VS Code, offers features like "Cascade AI Flows" and project-wide context awareness, aiming to reduce the context-switching common with other AI assistants. It competes with tools like Cursor, another AI-native code editor, by offering different models such as Claude 3.5 and Gemini 2.0 Flash. - The terminal is a key area of innovation, with tools like Warp, a Rust-based terminal, focusing on GPU-powered speed and a block-based interface to make reading logs and command outputs more efficient. While Cursor integrates a terminal, its primary strength is codebase-aware AI within the IDE, making Warp a preferred tool for terminal-centric tasks like DevOps and scripting. - Multi-tool workflows are becoming more formalized, with platforms like n8n and Make allowing developers to chain together different AI services and APIs visually. This approach enables the creation of complex creative pipelines, such as using an AI to generate text, which then feeds into another AI for image generation, followed by an automated optimization step. - A key challenge in agentic systems is the risk of "deskilling," where over-reliance on AI for routine coding and debugging could erode engineers' fundamental problem-solving abilities. The current pragmatic approach is to automate the 80% of repetitive work while ensuring human oversight for the remaining 20% that requires nuanced judgment and validation. - OpenAI's Codex app, launched in February 2026, exemplifies the shift towards multi-agent orchestration, providing a "command center" for managing parallel AI workflows on a single project. This allows multiple agents to work on the same repository concurrently without merge conflicts by using isolated "worktrees," enabling developers to explore different solutions in parallel.
