By Om Malik

The shift to AI has gotten me back into tinkering mode. A friend pushed me into using Cursor early, and I am using it all the time—not to write apps or create software, but to figure out how things work.

Cursor is an AI-enhanced development environment, part of a new category of AI-augmented development tools that differ from traditional terminal emulators or standard integrated development environments (IDEs). The company has grown rapidly, reaching nearly $100 million in annual recurring revenue (ARR) since its inception in 2023. While Cursor isn’t alone in this space – V0 from Vercel is another example – it’s the one I personally use.

It reminds me of using a browser in the 1990s, finding a website I liked, and right-clicking on the page to see the source code. That allowed me to figure out how everything worked. I can talk to Cursor in English. Now I can use it to create specialized one-off pieces of software. I don’t want to write software for a living, but I also don’t want to be ignorant about the world around me.

It turns out I’m not alone in seeing the potential of AI in revolutionizing software development workflows and tools. A new AI-enabled software development trend is slowly emerging. It will help developers transform what was previously a highly manual process into a streamlined, faster, and intelligence-augmented workflow. In addition to Anthropic’s Claude, Google’s Gemini, and OpenAI’s models, there are tools like Microsoft’s CoPilot for real-time code suggestions, Cursor for AI-powered code editing and refactoring, and Warp.dev for an AI-enhanced terminal experience that predicts commands and explains complex operations.

Hiding in the shadows are (even) newer tools. Developers (without AI) have had to rely on documentation, Stack Overflow, and years of accumulated knowledge to write code, debug issues, and maintain systems. These tasks now take minutes with AI assistance, helping developers amplify their abilities and allowing them to focus on higher-level problem-solving while AI handles more routine aspects of development.

“I think the population of those doing development is changing, and I think it’s changing in a good way,” said Zach Lloyd, one of those innovators and formerly a principal engineer at Google, who at one time led engineering at Google Docs.

When he left Google and was looking for something to work on, he kept going back to the terminal and how it could be reimagined for new workflows that involved modern user interfaces. “If you want to edit a command, you can’t click and put the mouse someplace and start typing,” Lloyd quipped. “You can’t double-click and hit delete. This is crazy. A product that is used every day by most developers doesn’t work like a modern app.”

So, he started Warp in 2020, a terminal with a much nicer user experience. While Warp is designed for professional developers, it’s built to be more approachable than traditional terminals. “It’s going to be much more familiar to someone who’s coming from Google Docs or VSCode,” he explained. “You don’t have to have taken your first programming course in C to understand how Warp works.”

A year and a half into Warp, the large language model transformation began in earnest. That’s when he started to reimagine Warp as more than just a terminal. “The terminal is a language-based, command-based interface into your computer where you can basically do anything,” Lloyd explained. “What the LLMs do is make the power and flexibility of that command line interface way more accessible than it otherwise would be.”

Lloyd pointed out that the terminal interface is akin to the most popular AI interface — chat. With the terminal, and with Warp in particular, not only can you converse with it and ask it questions, but you can also ask it to just do things. “You can just use the terminal in English, and that lets you do very high-level, complicated, multi-step tasks,” he explained. “It takes you up an abstraction level. You don’t need to know how all the commands work.”

“It’s one of the few programs programmers often live in all day every day,” said Mitchell Hashimoto, who just started Ghostty. In my tech-first view of the world, engineers like Hashimoto have a pole position.

Despite his relatively young age, he is what you would call a Silicon Valley throwback. A developer who started a software project while still in college, turned it into a company, took it public, and then quit the company and walked into the sunset—all before the age of 35.

He created Vagrant in 2010, a tool designed to help developers create and manage portable development environments when he was 21. Two years later, he and Armon Dadgar officially founded HashiCorp, a company that developed software products for managing multiple clouds. The company was ahead of the curve, grew fast, and went public. In 2023, he said goodbye to the company he started.

Over the 2024 holiday break, when he started talking about a new project, Ghostty, on social media, my ears perked up.Had it not been for Hashimoto’s tweets, I wouldn’t have gotten sucked into the terminal vortex. It led me down the path of thinking about a new AI-powered workbench for software developers.

“I started Ghostty in 2022 merely as a way to play with Zig, do some graphics programming, and deepen my understanding of terminals,” Hashimoto wrote on his blog. “I never intended to release it. I didn’t think there was innovation to be had. I thought I would learn a lot over a few months and move on.” Most terminals make you choose between the fastest, the most feature-rich, or the most native. “I felt all terminals made you choose at most two of these categories,” he told me. “I wanted to create a terminal that was competitive in all three categories.”

Hashimoto is keeping things very simple. Ghostty is a fast, feature-rich, and cross-platform terminal emulator that works on Linux and macOS and uses platform-native UI and GPU acceleration. Being native means that its tabs look and feel the same as tabs found in applications like Safari, Preview, and Finder. All the shortcuts on a Mac work with Ghostty. It has very low latency, uses very little memory, and is extremely fast.

Terminals, at one time, were physical devices. Anyone remember Teletype machines (TTY)? By the 1980s, terminals became software programs that allowed personal computers to connect to mainframe computers. Even as computers evolved and adopted advanced graphical user interfaces (GUIs), terminal emulator software remained critical. Despite the decades that have passed, terminals have remained essential, becoming more useful with integration into modern development environments.

With most software now being cloud-native, it is no surprise that more and more people are looking at terminal emulators, along with the entire developer stack, as ripe for reinvention, especially with the emergence of AI. It is not hard to compare integrating AI in terminals to putting an electric motor on your bicycle. Now, you only pedal when necessary — and you can go very fast.

Looking ahead, Lloyd believes development will be transformed by AI within the next three years, though the exact shape of that transformation remains unclear. “If you’re not embracing AI, you’re gonna find yourself at a disadvantage in terms of productivity,” he warns. He sees AI proficiency becoming a crucial skill for developers, particularly in how they prompt and interact with these tools. “There’s a skill in how you use the AI… how you get it to give you the right answer and use it to save you time. The best developers are going to be the ones who can figure out how to use it to make themselves more productive.”

However, Lloyd pushed back against the notion that AI will completely replace human developers. While tools like Replit or V0 may enable non-professionals to create simple applications, he believes professional developers will remain essential for tackling complex problems.

“There are enough hard problems with real existing apps that professionals will still need to build those apps for now,” he argued. The key difference will be in how these professionals work – their effectiveness will increasingly depend on their ability to leverage AI tools effectively. “Just like the best professionals have traditionally been those who learn how to use their tools the best, whether it’s the terminal or the IDE, I think going forward, you have to learn how to use the AI to your advantage.”

Hashimoto also believes that the emergence of AI means that a lot more efficiency can come into the coding realm. But that doesn’t distract him from the mission of building Ghostty and a vibrant community around the open-source project. His hope is to turn Ghostty into something that is extensible, with its community helping shape its future. There is a lot of work that remains to be done.

If anything, my conversation with Hashimoto, Lloyd, and a few others has reminded me that in order to understand the future and get your line-of-sight right, you have to talk to the engineers. The chattering classes on social media would have you believe that “developers” are toast tomorrow. Not even close. There is a reason smart people are building things.They see the problems. They smell the opportunity. They know others like them need those things.

Disclosure: Due to my ongoing affiliation with True Ventures, I received and continue to own HashiCorp shares.