Nikunj Kothari's Insightful Take on AI's Evolution

In a recent tweet that's sparking conversations across tech circles, Nikunj Kothari, a partner at FPV Ventures, shares his experiences with Claude Code and draws a fascinating parallel to the early days of personal computing. The original thread on X (formerly Twitter) and his blog post paint a picture of how AI is ushering in what he calls the "DOS Era of AI."

Kothari describes spending hours in Claude Code's black-screen interface, reminiscent of old-school DOS terminals. But unlike the intimidating command lines of the 1980s, this modern version leverages AI to handle the heavy lifting. He recounts fixing a Wi-Fi issue not by typing commands himself, but by simply explaining the problem to the AI, which then diagnosed and suggested solutions faster than he could have manually.

From Problem-Solving to Building Without Code

The real magic, according to Kothari, lies in how Claude Code shifts the focus from coding to problem-solving. He built event posters, investment memos, and complex spreadsheets—all without typing a single command. Just describe what you want, and the AI executes it, pulling from local files, GitHub, databases, and more.

This approach contrasts with tools like Cursor (for coding) or Replit (for shipping products). Claude Code assumes you have a problem to solve, making it accessible even to non-technical users. As Kothari puts it, "When you only need to explain problems, not understand solutions, everything shifts."

He draws a historical analogy to DOS, where breakthrough apps like Lotus 1-2-3 and WordPerfect started in clunky terminals before graphical interfaces like Windows made them user-friendly. Today, we're in a similar "terminal phase" with AI, but it's evolving rapidly. Tools like Conductor, Terragon, and Magnet are already building more intuitive layers on top.

The Model Context Protocol and Beyond

Kothari also highlights the Model Context Protocol (MCP), which enables seamless integration between systems—like Google Drive flowing into terminal commands or local files connecting to cloud databases. This reveals how current apps limit our view of what's possible, offering only "keyhole" perspectives.

In this phase, AI creation is raw and unfiltered: no app interfaces or product manager assumptions—just direct conversation with intelligence. The bottleneck? Not tech knowledge, but imagination. Kothari shares spending time describing an animation simply because he struggled to articulate his vision.

Implications for Meme Token Creators in Blockchain

While Kothari's post focuses on general AI applications, its implications for the meme token ecosystem are profound. Meme tokens thrive on community, creativity, and quick execution, often launched by non-technical enthusiasts inspired by viral ideas.

Imagine a world where you don't need to code a smart contract or set up a token launch. Just explain to an AI like Claude Code: "Create a meme token based on this viral cat image, with fair launch mechanics, integrated liquidity pool on Solana, and automated social media promotion." The AI could handle the technical details—deploying on blockchain, integrating with DEXs like Raydium, or even analyzing market sentiment via on-chain data.

This democratizes blockchain innovation, allowing more people to experiment with meme tokens without barriers. Tools evolving from this "DOS era" could lead to AI-assisted token design, where imagination drives creation. For blockchain practitioners, it's a call to explore these AI terminals now, before drag-and-drop versions make them mainstream.

As Kothari notes, the DOS phase was short but foundational, birthing giants like Microsoft and Oracle. In AI's terminal era, early adopters in crypto could build the next big meme token empires.

Looking Ahead

Kothari sees the black screen not as "coding," but as the dawn of an "AI computer." For meme token enthusiasts, this could mean faster iteration, better problem-solving, and more inclusive participation in blockchain.

If you're in the meme token space, tools like Claude Code might be your next edge. Dive in, explain your problems, and watch intent turn into execution. The future of crypto creation is looking a lot like 1985—but smarter.