GenAI Diary (page) ... Savvy computer users moving from chatbots to agents to writing their own agents using natural language

Last update: Sunday 5/11/25

The editor of this blog recently read a fascinating article in VentureBeat, then watched VentureBeat’s YouTube podcast. Both claimed that genAI is entering a new phase wherein non-coders can build their own agents using natural language. We aren't there yet, but this new phase is coming fast ... (GenAI Diary home page)

So the editor asked ChatGPT on GPT-4o to provide an overview of the latest developments with regards to the development of agents using natural language prompts. 

The editor also posed this question to Gemini on Flash. Gemini's initial response asserted that current apps were mostly low-code, not no-code, although no-code was on the way. When I pressed for specific examples, it provided a few, but did not include anything from Google. When I reminded Gemini that VentureBeat had declared that Google itself was an active pioneer in no-code, it "apologized" and cited "Google Cloud Vertex AI" as an app that savvy computer users might be able to use to create simple agents with no-code.

Now here is ChatGPT's's overview. Note that it is labelled "ChatGPT's Response" because of the editor's concern that its extensive attention to OpenAI's activities may reflect possible built-in bias in the online materials searchable by ChatGPT. Indeed, the editor had to explicitly prompt ChatGPT to include Anthropic and Cursor ... 😎

<ChatGPT’s Response>

The Rise of No-Code Agent Creation: AI Tools for the Non-Coder
For years, the ability to automate complex workflows or build intelligent software agents belonged almost exclusively to people who could code. But in 2025, we’ve crossed a critical threshold: you no longer need to write code to create AI agents that think, plan, and act.

This post reviews what’s now possible, who’s making it possible, and how major players like OpenAI, Anthropic, and Cursor are moving aggressively into this space.

From Prompting to “Vibe Coding”

“Vibe coding” is shorthand for a powerful shift: telling an AI what you want to happen, and letting it figure out the technical details. Rather than learning Python or wiring up APIs, users now express goals like: “Summarize my morning emails, extract anything urgent from my lawyer, and text me if needed.”  With the right platform, that’s no longer a wishlist—it’s a deployable agent.

OpenAI’s Move: Windsurf Acquisition

OpenAI’s reported $3 billion acquisition of Windsurf (formerly Codeium) points to a broader strategy: make agent creation a first-class citizen inside ChatGPT. Windsurf is an AI-powered development tool known for high-performance code generation. By integrating it with OpenAI’s tool use and memory features, ChatGPT could evolve from an assistant into a true agent-building platform. This would empower users to:

• Create agents with persistent behaviors
• Chain tools together across tasks
• Use natural language to define workflows, rules, and goals

This isn’t a far-off vision—some of it is already here.

Platforms Enabling No-Code Agent Design

Here are the leading platforms that support agent creation today—with no programming required:

1. OpenAI Custom GPTs (ChatGPT Pro)

• Memory, tool access, and custom instructions
• Great for task-based agents (e.g., content creation, email triage)
• Still sandboxed—can’t yet act freely on your computer or the open web

2. Cognosys

• Create browser-based agents that perform tasks live on websites
• Ideal for scraping data, booking events, automating web searches

3. MultiOn

• Personal assistant that operates your browser
• Accepts voice or chat commands like: “Search YouTube, summarize top 5 videos”

4. Dust.tt

• Document-centric AI agents with persistent memory and tool plugins
• Especially useful for internal research or enterprise knowledge tasks

5. Relevance AI

• Drag-and-drop visual builder for AI task flows
• Geared toward business teams, customer support, and data enrichment

Anthropic’s Claude: Powerful Agent Building Without Code

Anthropic has also entered the agent space in a big way. With Claude 3.5 and 3.7, users can now:

• Create structured agents that use tools and APIs
• Leverage Claude Code, a command-line interface that responds to natural language
• Integrate with external data through their open-source Model Context Protocol

Claude is especially good for:

• Planning multi-step tasks
• Customizing assistants (e.g., travel planners, report writers)
• Operating across documents, tools, and APIs—without code
  

Cursor: AI Agent Mode for Developers and Power Users

While geared toward coders, Cursor introduces an “Agent Mode” that’s relevant to non-coders who collaborate on technical work.

• You can describe a desired change or feature.
• Cursor’s agent searches the codebase, identifies dependencies, and makes modifications across files.
• Works well for mixed teams—where one person writes prompts, and another edits the output.

Cursor effectively blurs the line between developer and strategist.

The Big Picture: A New Kind of Power for Non-Coders
We’ve entered a new era where:

• Agents are becoming composable, persistent, and personalized
• Natural language is now a software development interface
• Strategy and clarity—not syntax—are the new bottlenecks

If you know what you want done and can explain it, you now have the tools to build agents that carry it out—no engineering team required.

This is the beginning of a broad democratization of automation and software design. Whether through OpenAI, Claude, or Cursor, the future belongs to those who can think clearly and direct intelligently.

</ChatGPT’s Response >