I Built 50+ Apps Using AI Without Coding (2026 beginner's guide)

Key Concepts

• Vibe Coding: A term used to describe coding with AI agents, emphasizing exploration and learning over traditional programming skills. Ben Tossel argues it’s more than just a “vibe” and involves genuine skill development.
• Agents.md: An open format file used to provide context and instructions to AI coding agents, acting as a “readme” for agents.
• CLI (Command Line Interface): A text-based interface for interacting with a computer, favored by Tossel for its capabilities and visibility into the agent’s work.
• VPS (Virtual Private Server): A virtual machine rented as a service, used for running applications and keeping data persistently available.
• Programmable Layer of Abstraction: The shift from mastering drag-and-drop no-code tools to learning how to effectively work with AI agents.
• Fail Forward: A methodology of learning through building, encountering issues, and using those issues as opportunities for understanding and improvement.

Learning to Code Without Coding: Ben Tossel’s Approach with AI Agents

This discussion centers around Ben Tossel’s documented experience learning to build software using AI agents, despite lacking traditional programming expertise. The core argument is that anyone can leverage these tools to create functional applications, learn about software development, and even contribute to professional projects. The insights are based on an article by Tossel detailing his workflow, tools, and learnings over four months of spending three billion tokens with AI agents.

I. Tossel’s Achievements & The Power of Observation

Tossel, previously involved in the no-code space (founder of a no-code ed company acquired by Zapier), has demonstrably shipped a significant number of projects using AI agents. These include:

• Personal Website: Revamped to resemble a terminal CLI tool.
• Social Tracker (Factory): Monitors mentions of “Factory” across Twitter, GitHub issues, and Reddit. (Open Source)
• Factory Wrap: A prototype product that was integrated into Factory’s core offering.
• Custom CLIs: Python CLI used for customer support queries.
• Crypto Tracker: Predicts market signals (positive, negative, neutral).
• DroidMiss: 12 experimental games exploring AI agent concepts.
• AI-Directed Video Demo System: Automates video creation, screen recording, and editing.
• Telegram Bot: Syncs local repositories to a VPS.
• Numerous other (unmentioned) projects.

Crucially, Tossel emphasizes that he doesn’t read code in the traditional sense, but meticulously reads the agent output. This process has provided him with a practical understanding of how code works, project structures, and common failure points. He states, “I don’t read the code but I read the agent output religiously and in doing so I picked up a ton of knowledge around how code works how projects works where things fail and where they succeed.”

II. Workflow & Tooling: A Non-Technical Approach

Tossel’s workflow prioritizes direct interaction with AI agents and a focus on iterative development. Key elements include:

1. CLI-First Approach: He exclusively uses the Command Line Interface (CLI) over web-based interfaces, citing its greater capability and transparency. This is presented as a crucial step for non-technical users.
2. Project Initialization (Droid): He initiates projects using Factory CLI (Droid), a tool described as turning scripts into “agent armies.”
3. Contextual Input & “Spec Mode”: He begins by providing the model with context about the desired outcome, then enters “spec mode” – a phase of questioning assumptions and exploring alternative approaches. He approaches this like a philosophical inquiry, asking “why” repeatedly.
4. Agent Execution (Opus 4.5): He utilizes Opus 4.5 with high autonomy, allowing the agent to generate code. He actively watches the process, intervening when errors occur or to redirect the agent.
5. Testing & Iteration: He tests the generated code, provides feedback, and iterates on the design. He emphasizes building first and identifying learning opportunities from the resulting gaps.
6. Agents.md Implementation: He leverages agents.md files – a standardized format for providing instructions to AI agents – to define project parameters, coding standards, and testing procedures. He actively refines his agents.md setup for consistency across projects.

III. Key Learnings & Skill Development

Through this process, Tossel has acquired a range of skills and insights:

• Bash Commands: He rediscovered the utility of Bash commands for automating tasks and interacting with the operating system. He created a slash command flow using Droid to automate the changelog process.
• CLI Preference: He now consistently favors CLI versions of tools (Superbase, Vercel, GitHub) over their GUI counterparts, citing simplicity and reduced context usage.
• VPS Utilization: He learned to use a Virtual Private Server (VPS) to host applications requiring persistent uptime, such as his crypto tracker and Telegram bot. He uses sync thing to keep local repositories synchronized with the VPS.
• Skills Integration: He’s begun incorporating “skills” (predefined functions or knowledge bases) into his workflow.
• Testing Importance: He now prioritizes end-to-end testing to catch errors and improve code quality.

IV. The New “Technical Class” & The Programmable Layer

Tossel argues that the landscape of software development is shifting. Instead of needing to master traditional coding languages, the new challenge is learning how to effectively work with AI agents. He defines this as a “new programmable layer of abstraction.” He draws a parallel to the no-code movement, where the abstraction layer was drag-and-drop tools like Webflow and Zapier. Now, the abstraction layer is prompting, context management, and understanding how to guide AI agents.

He rejects the label of “non-technical,” instead identifying as part of a “new technical class” – a group that leverages AI agents to build software without necessarily being traditional programmers. He feels the term "vibe coding" is dismissive of the skills involved.

V. Learning from Others & Embracing Exploration

Tossel emphasizes the importance of learning from other developers, particularly by:

• Reading Open-Source Code: Cloning and exploring existing projects to understand how they work. He specifically mentions Peter Steinberger’s work as an example of simplicity.
• Following Twitter: Observing the workflows and insights shared by other developers.
• Asking “Silly Questions”: He encourages asking fundamental questions without fear of judgment, as this is a crucial part of the learning process.

VI. Future Outlook & Call to Action

Tossel predicts an explosion of software development, with a significant portion being of low quality (“AI slop”). However, he believes that the increased accessibility of AI agents will empower more people to build and contribute to the software ecosystem. He encourages viewers to:

• Embrace Play: Treat AI-assisted coding as an experiment and a learning opportunity.
• Pick a Tool & Stick With It: Focus on mastering one agent platform (e.g., Droid, Cursor, Cloud Code) rather than spreading efforts across multiple tools.
• Build, Fail Forward, and Ship: Iterate quickly, learn from mistakes, and consistently release projects.
• Daily Exposure: Regularly engage with AI tools to develop proficiency and unlock new possibilities.

Concluding Synthesis:

Ben Tossel’s experience demonstrates that building software without traditional coding skills is not only possible but also a powerful learning opportunity. By embracing AI agents, prioritizing iterative development, and focusing on understanding the underlying systems, anyone can become a “new technical class” contributor and unlock a world of creative and professional possibilities. The key takeaway is to shift the focus from writing code to directing code, and to treat the process as a continuous learning experience.