A Robot Vacuum Giant Just Called on Programmers Worldwide: "Come Use Cursor"

Narwal’s booth at this year’s CES was quite impressive. Who would have thought they’d pull a move like this just a few months later?

Early March in Shanghai is terribly raw and cold—barely above 5 degrees Celsius with a perpetually gray sky. I was curled up in my chair wrapped in a blanket, scrolling through my phone, when I stumbled upon a piece of news that nearly made me drop my donut.

Why is a Robot Vacuum Company Suddenly Calling on Global Programmers?

Here’s the situation: Narwal Intelligence—the company known for the robot vacuums that “wash their own mops”—recently encountered a sudden situation. The specific nature of this “accident” varies depending on who you ask, and the information I can find is quite scattered. But the result is clear: they have released messages across communities and recruitment channels inviting programmers worldwide to use Cursor to participate in their project development.

When this news was reposted in the Linux.do community, the title read: “Narwal really has vision, asking everyone to do AI coding.” The comments section was bustling.

Just picture this: A global Top 5 cleaning robot company, shipping millions of units annually, selling in over 50 countries to more than 4 million users. When a company like this hits a snag, their first reaction isn’t the traditional route of mass-hiring embedded engineers, but directly shouting out: Come, use Cursor, and let’s write code together.

From the J1 in 2019 to now, Narwal’s product line has expanded significantly, and one can imagine the software complexity behind it.

To be honest, my first reaction was that it sounded a bit ridiculous. But after thinking about it carefully, I realized this is actually worth discussing.

Why Cursor?

First, a bit of background. If you haven’t heard of Cursor yet, how should I put this… it is currently the hottest AI programming tool in the global developer circle, bar none. It’s an independent editor heavily modified based on VS Code, but its core is an AI-first design philosophy, where everything is built around Large Language Models (LLMs).

In February of this year, Cursor released a major update. According to CNBC, the updated Agent can test code on its own, screen-record its workflow, and run in parallel on independent virtual machines. Their annualized revenue has broken $1 billion, with a valuation reaching $29.3 billion. Thousands of engineers at Stripe are using it, and 90% of the 20,000+ developers at Salesforce are on it.

I’ve chatted with a few tech friends before, and their consensus is: Cursor isn’t a question of “is it good?” anymore, but rather “if you aren’t using it, you’re wasting time.” A backend developer friend told me that since using Cursor, his speed in handling daily features and bug fixes has roughly doubled. This figure aligns with external benchmarks—reports suggest that using AI tools can boost development efficiency by 126%.

But here is the key point. Cursor’s typical user profile has largely been internet companies, SaaS teams, and independent developers—people with pure software backgrounds. For a hardware company making robot vacuums to jump out and say “we want to use this too,” the signal is actually much more complex than it appears on the surface.

Everyone talks about AI Coding, but for a Hardware Company, it’s a different beast

(Rubbing chin) The interesting part of this story isn’t Cursor itself, but who is using it.

Look at Narwal’s tech stack: it spans SLAM (Simultaneous Localization and Mapping), 3D perception, AI object recognition, robotics structural technology, and big data applications. They have applied for over 1,100 technical patents. This isn’t as simple as building an App or writing a webpage. Robot vacuum code involves embedded systems, real-time control, sensor fusion, path planning… these fields have completely different requirements for code.

For an internet company to use Cursor to write an API interface or build a frontend page, it’s like a fish in water. But to use AI programming tools to handle embedded C++, RTOS task scheduling, or low-level hardware drivers—frankly, the difficulty level is in a different league.

However, having said that, modern robot vacuums aren’t purely embedded affairs anymore. Narwal has an App, cloud services, data platforms, and user interaction systems. In 2025, they secured a $100 million Series E+ round led by Tencent Investment. This means their software-side volume is actually huge. App feature iteration, cloud service maintenance, building big data analysis platforms… these tasks can absolutely be accelerated using Cursor.

So I suspect—and maybe I’m overthinking it—that their “sudden accident” likely occurred on the software side rather than the hardware side. App crash? Server-side issues? A critical system needing an urgent rewrite? In these scenarios, calling for people to help using Cursor makes logical sense.

The data in this chart is actually more exaggerated than I imagined—from 70% to 84%, AI programming tools have become standard configuration for developers.

Speaking of the “Competition”

Cursor isn’t the only player. This field is currently a battlefield.

Anthropic’s Claude Code has surpassed $2.5 billion in annualized revenue. OpenAI’s Codex just released a standalone App, with over 1 million monthly developers using it, allowing free users access while burning subsidies to grab market share. GitHub Copilot has over 26 million users, backed by Microsoft’s full-force promotion.

But there’s an interesting distinction. Cursor is positioned as a complete AI-first IDE—you can understand it as a “total solution.” Copilot is more like a plugin embedded into your existing editor. Codex chose another path—embedding into developers’ existing varied environments rather than building an independent editor.

To put it bluntly, Anthropic has recently been criticized by many developers for increasingly strict usage limits and a closed ecosystem. Cursor’s advantage lies precisely in its support for multiple models—Claude, GPT-4, GPT-5, DeepSeek—which you can switch between as needed. This flexibility is crucial in actual development.

However, there is a question I haven’t reached a definitive conclusion on: As these tools become stronger, will the differences between them be leveled out by the capabilities of the models themselves? After all, much of Cursor’s core competitiveness comes from user experience and workflow integration, while the model capabilities are provided by underlying suppliers. If one day everyone uses the same or comparably capable models, how will Cursor retain users?

A partner at Wing VC wrote a good analysis, stating that Cursor’s moat might be “more like a mirage than a fortress.” I think that’s a bit harsh, but not without merit.

If Hardware Companies Start Playing Like This…

There’s one more thing I want to expand on.

If we look at Narwal’s move against a larger backdrop, it actually represents an accelerating trend: the software-ization of hardware companies. IDC released a report this year titled “Model Driven, Software Defined, Hardware Reconstructed,” discussing how the robotics industry is evolving along the path of a “software-defined architecture.”

I’m thinking, if a hardware company of Narwal’s size starts using Cursor and AI programming tools to restructure their software development process, does that mean the software development models of many traditional industries are being redefined?

You don’t need to maintain a 200-person backend team anymore. You might only need 20 people who understand business logic and know how to use AI tools. Coupled with tools like Cursor, they could cover the work of 100 people from the past. This algorithm isn’t something I made up—Cursor’s website states that after Salesforce’s 20,000 engineers used it, both delivery cycles and PR speeds saw “double-digit improvements.”

Maybe I’m overthinking it, but if this trend continues, the scarcest resource in the future might not be people who can write code, but people who can define what code needs to be written. The ability to pinpoint key problems amidst noisy requirements might be more valuable than the technical implementation itself.

Left: Programmers drowning in code in 2020. Right: Technical developers focusing on strategic thinking in 2025—the change is indeed happening.

The Rain Doesn’t Seem to Have Stopped

Outside the window, it’s still gray. March in Shanghai is just like this—damp and cold, not enough to freeze you to death, but enough to make you drowsy.

As for the follow-up on the Narwal situation, frankly, I don’t have enough information yet to make a final judgment. What level of “accident” did they encounter? How many Cursor users did they actually recruit? What was the final outcome? I’m still watching these details.

But the signal released by this event itself is, I think, worth remembering: When a company making robot vacuums starts shouting to the world “Come use AI to write code,” the way this world operates is indeed quietly changing.

As for me, tonight I plan to use Cursor to see if it can help me refactor the messy backend of my personal blog. If this public account doesn’t update tomorrow, it’s likely because Cursor helped me crash the code. ╮(╯▽╰)╭

References:

—— Lyra Celest @ Turbulence τ

A Robot Vacuum Giant Just Called on Programmers Worldwide: “Come Use Cursor”