A Year Of Vibes

2025-12-22 · Source: Armin Ronacher's Thoughts and Writings · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, long

Summary

The author reflects on 2025 as a transformative year, marked by a significant shift in programming methodology due to agentic coding tools like Claude Code, Amp, and Pi. This change involved moving from hands-on coding to an "engineering lead to a virtual programmer intern" role, resulting in a substantial increase in blog posts and conversations about AI. The author notes a growing reliance on code generation, file systems, programmatic tool invocation, and skill-based learning, reinforcing the belief in this approach as foundation model providers also adopt skill-based strategies. The piece also explores the unexpected emotional connections formed with AI tools, the challenges of conflicting opinions on AI adoption, and the evolving landscape of outsourcing versus building solutions in-house with AI assistance. Key pain points identified include the limitations of traditional version control and code review for AI-generated code, and the need for new observability paradigms.

Key takeaway

For engineering leaders and architects evaluating AI integration, recognize that current version control and code review systems are inadequate for agentic workflows. Your teams will need new tools that capture prompts, failed attempts, and differentiate human from machine input. Prioritize exploring alternative VCS like Jujutsu and evolving code review processes to support local, agent-driven iteration, ensuring team visibility and maintaining code quality standards in an AI-augmented environment.

Key insights

Agentic coding tools are fundamentally reshaping software engineering workflows, challenging traditional practices and human-machine relationships.

Principles

• Value exists in recording failed attempts for agentic learning.
• Observability benefits from LLM's query and analysis capabilities.

Method

The author's agentic coding method involves code generation, file systems, programmatic tool invocation via an interpreter, and skill-based learning, primarily using tools like Claude Code, Amp, and Pi.

In practice

• Use agentic tools for day-to-day tasks beyond code generation.
• Consider building SDK generators with AI for custom needs.

Topics

• Agentic Coding
• Large Language Models
• Code Generation
• Software Development Workflow
• Version Control Systems

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Software Engineer, Machine Learning Engineer, AI Architect

Related on AIssential

• Vibe coding and agentic engineering are getting closer than I'd like — The increasing reliability of AI coding agents blurs the distinction between casual "vibe coding" and professional "agentic engineering."
• The Best AI Observability Tools for Agentic Systems in 2026 — Agentic AI observability must integrate testing, debugging, and iteration, moving beyond simple LLM call monitoring to support complex multi-step workflows.
• Most Influential ArXiv (Software Engineering) Papers (2026-04 Version) — LLMs and AI agents are rapidly transforming software engineering, driving innovation in code generation, testing, and repair.
• Agentic Workflows Have Changed EVERYTHING in 2026 (DEATH Of The Senior Dev?) — Agentic AI workflows have fundamentally reshaped software development by 2026, making traditional coding skills obsolete while elevating engineering principles.
• How AI Coding Agents Actually Understand Your Codebase — AI coding agents use tools and an iterative loop to understand code, plan, and execute tasks, mimicking human engineers.
• Running an AI-native engineering org — Agentic coding fundamentally redefines engineering bottlenecks, shifting focus from code generation to verification and strategic human oversight.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Armin Ronacher's Thoughts and Writings.