Do Software and Programmers Still Have a Future?

Background

Six months ago, when we released NocoBase 2.0, we published our second recap: No AI, No VC, Just 17K Stars and Real Revenue. At that point, NocoBase had already reached around $1.4 million in annual revenue.

The plan had been simple: one recap a year, sharing NocoBase's growth as it happened. This third piece was supposed to wait until the end of this year, when NocoBase 3.0 comes out. But the world is changing too fast. AI is hitting this industry with a new shock wave almost every day. In a jolt like this, NocoBase was never going to remain untouched. We have been pulled, over and over, into all kinds of uncertainty.

So we had to move faster. We had to stay nimble enough to answer the shock as it came.

That is why this recap is arriving half a year early, together with NocoBase 2.1. Better to write it now than wait another six months and wonder whether this industry will still look the same by then. Or whether anyone will still care about an open-source product like this. Or whether anyone will still care about software, or programmers, at all.

Where We Are

It has now been five years since we committed the first line of code on GitHub.

Compared with six months ago, our team size has not changed. We are still 14 people. We still do not have a dedicated sales team. For the most part, users still find us, not the other way around. But beyond SEO, we have also started paying serious attention to generative engine optimization, and we already have quite a few paying users who discovered NocoBase through ChatGPT and Claude.

Here are a few other numbers:

• GitHub stars: 22.7k
• Contributors: 115
• Git clones: 3K/day

Revenue

Now that the revenue figure is becoming more substantial, we will stop disclosing exact numbers in future recap articles. But we will keep sharing the trend lines, and the shape of what is happening underneath them.

In the first five months of 2026, our revenue was exactly double what it was in the same period of 2025. In our best month so far, monthly revenue alone had already reached our entire revenue for all of 2024.

But if I am being honest about how it felt from the inside, this was not the result of some smooth, graceful curve upward. It came after a major turn.

Were We About to Be Killed?

Before 2025, NocoBase was positioned as a no-code platform. It was a direction plenty of companies had already proven to be valuable and commercially viable, even if it was not the kind of business that usually leads to spectacular fortunes. The familiar products in this category can become solid businesses, even if they rarely produce outsized outcomes. With differentiated positioning, a standardized product, and a global market, we believed we had a chance to build a durable business around a focused product and a lean team.

Starting in 2025, we began introducing AI features into the product. That opened up a new imaginative horizon for the traditional no-code platform. AI could play a supporting role in business workflows and help people get certain kinds of work done more efficiently. At the same time, our revenue was climbing quickly, which seemed to validate the decision.

Then, at the end of 2025, things changed.

When Opus 4.5 was released, it felt as though the weather turned overnight. Social media was suddenly full of awe at the revolutionary changes it seemed to be bringing to programming. Then came the wave of layoff news, followed by all the loud voices declaring that traditional software was about to be killed. And just like that, the whole industry seemed to be living inside the same split reality, one half panic, one half exhilaration.

That mood spread through our team almost immediately.

More than one colleague started to feel that what we were building no longer had any meaning. And once that feeling takes hold, it does real damage. We had spent years building something we were proud of. Now AI appeared able to produce something similar in a few hours. And AI was still evolving at a frightening pace. So what were we still here for? Only a month earlier, I had believed that the stronger AI became, the better it would be for us, unless it planned to take over the entire world. Was that day really about to arrive this quickly?

At the same time, NocoBase's revenue also entered a weaker stretch. I could not tell whether that was just because of the Christmas and New Year holiday season, or because a flood of AI news had genuinely affected how enterprises were making purchasing decisions.

For two full months, I read obsessively. I tried product after product. Every day I discussed, argued, and compared notes with the more radical people on our team, while also trying to keep everyone grounded. I spoke with NocoBase users about how they viewed AI and how it was actually being used inside their companies.

And while I was doing all that, I sank into a serious state of anxiety, the kind I had not experienced in many years. When I was resting, I could hear my own heartbeat. I would fall asleep, wake up quickly, and find it hard to sleep again. The muscles in my abdomen were often so tight that I could feel them beating with my pulse.

NocoBase has not made us a huge amount of money. But everyone on this team has tied some part of their ideals to it, along with some part of their hope for a better life. If it were killed, it would hurt. A lot.

It was not until after February, after many rounds of intense discussion inside the team and many conversations with users, that I felt we had finally thought it through. We then held three all-hands alignment sessions in a row and shared a presentation called The Twilight of the Old Version, the Dawn of the New One. It laid out why NocoBase still deserves to exist, and where it should go next. We made sure everyone understood the reasoning behind the direction and felt confident moving forward.

Looking for What Does Not Change

The day before I started writing this piece, one of the three largest life sciences companies in the world contacted us. They wanted to explore using NocoBase as infrastructure, layering AI capabilities on top of traditional systems like SAP, and rebuilding certain parts of their supply chain on that basis.

For example, they want to identify and structure large volumes of orders arriving from different countries through fax, paper, handwriting, websites, and other channels. They want to plan shipping in batches based on warehouse locations and delivery destinations. They want to process accounts receivable, bills, and invoices more efficiently depending on different payment methods.

This is a giant enterprise with tens of thousands of employees spread across dozens of countries. They have no shortage of technical expertise and access to the most advanced models. So why would they still consider a product like NocoBase, instead of simply instructing an LLM to write the whole thing from scratch?

That was exactly the question we had been discussing and thinking through over the past few months.

Large models and agents are advancing at astonishing speed. They are making code itself incredibly cheap. They are making programmers as a group feel deeply uneasy. But once that same wave reaches traditional enterprises, it loses a lot of force.

The furniture company is still making furniture. The supply chain team is still doing supply chain work. Processes that never even made it to digitalization are still running on A4 paper and fax. These things are not going to undergo some revolutionary transformation just because LLMs can generate code. LLMs are not wish-granting machines. You cannot wave a magic wand and expect an entire enterprise to remake itself.

The bar is still very high. Solving real problems in production, sales, logistics, finance, and the rest of the business means building systems that are solid, secure, able to run for years, able to evolve, and able to absorb AI in ways that genuinely improve efficiency.

Code generation lowers the barrier to generating code. That is all. And in practice, if you use it badly, it may lead you straight into more traps.

In The Twilight of the Old Version, the Dawn of the New One, we listed several things that we believe will remain unchanged for quite a long time. Those are also the reasons we believe NocoBase still needs to exist.

1. If AI is going to truly land inside business, a chat window is not enough. AI needs a real operating environment. NocoBase gives it one, along with a ready-made toolbox. We turn data, workflows, and functional modules into interfaces AI can actually use, while surrounding that access with strict permissions and full logging. That is what allows AI to query data safely, identify risks, carry out complex tasks, and gradually move enterprise software from the old logic of "people looking for data" to a new logic of "tasks finding people."
2. An enterprise-grade system must have standardized data structures, rigorous permissions, strict business workflows, audit logs for every action, and historical records for key data. This layer is complex, and it cannot afford to be wrong. And no matter how much AI improves, it will not make these things disappear, because this is not really a problem of intelligence. It is a problem of people and organizations. Rather than asking AI to rewrite a fresh, half-proven version of this every time, it makes far more sense to build on top of standardized foundational modules that have already been thoroughly tested.
3. Enterprise applications contain a great deal of basic functionality that gets reused over and over again. User systems, authentication, email and SMS delivery, notification centers, data import and export, backend asynchronous jobs, almost every system needs them. As AI develops, these building blocks will keep changing too, and more AI-oriented tools will appear with them. But asking AI to rewrite them from scratch every single time is a massive waste. Reusing mature capabilities that already exist in the platform still creates enormous value.
4. WYSIWYG no-code visual configuration still matters. Its value is no longer just that it reduces the amount of code people need to write. It also gives AI output a visible surface that humans can adjust directly. If AI generates an interface or a workflow, people should be able to understand it at a glance, then refine it with their own hands. That kind of directness preserves transparency in human-AI collaboration. It keeps the system from turning into a black box that only AI understands and no one else can really move.
5. Without truly strong architectural design, the more code AI generates, the harder it becomes to maintain. NocoBase uses architecture to impose a set of physical rules on the system. It keeps interfaces and interaction styles highly consistent. It keeps automation under shared standards. It keeps plugins under shared standards too. That is what makes long-term, stable iteration possible.

Based on these things that do not change, we immediately adjusted our product positioning. We moved from "no-code platform" to "AI + no-code" infrastructure, and within a matter of weeks we completed the corresponding product changes. Of course, AI itself helped enormously, pushing our efficiency to more than twice what it was before. The no-code platform that had originally been built for people formally became a platform built for collaboration between people and AI.

And soon after that, we started receiving strong positive feedback from several large enterprises.

One pharmaceutical company, with more than 20,000 employees and dozens of subsidiaries, had begun promoting AI coding across the whole company months earlier. In the process, they had done a great deal of engineering work to deal with hallucinations, limited context windows, and architectural drift, all the problems that make long-term stable iteration so difficult. Even then, the results were not encouraging. Eventually, they concluded that a more sensible approach was to develop on top of a scaffold suited to AI, one that would preserve efficiency and flexibility while also placing strong constraints on AI behavior. After spending several weeks testing NocoBase, they concluded that NocoBase was exactly that kind of infrastructure.

Another top-tier renewable energy company in the wind power sector used NocoBase inside a single team for several months, then formally rolled it out across the entire company, to tens of thousands of employees, as an AI development platform. It has since gone into production in key scenarios such as review workflows, project management, and AI portals, allowing AI to create dependable value in those areas. They also have more than enough technical and business experts, the most advanced models, and essentially unlimited tokens. But spending time and resources on low-level infrastructure that does not generate returns is simply not worth it. Once they adopted NocoBase, they were able to focus most of their energy on the business itself.

What We Do Next

No matter what we think, AI is still moving very fast. If we want to stay alive for longer, and if we want to keep finding a sense of fulfillment true to our original intentions, then we need to keep reminding ourselves not to forget a few simple facts.

Stand on What Lasts

Under AI's impact, the pace of the world seems to have accelerated dramatically. And with social platforms, industry hype, and recommendation algorithms amplifying everything, we are bombarded by new concepts every day. But from what we have seen in real life, the world of AI vendors and online commentary is very different from the world of actual enterprises.

No matter how intense the conceptual bombardment becomes, real businesses are still making their pharmaceuticals, their cars, their bottled water. Many companies still have not even completed basic digitalization. They still rely heavily on paper and Excel, let alone any AI revolution. If we anchor ourselves in the things we can identify as lasting, and take those real businesses as the people we serve, that will be the foundation of our healthy survival.

Embrace What Changes Fast

Since the company was founded, we had been adding one or two new people to the product and engineering team every year. But starting this year, we decided to stop expanding the engineering team. Instead, we would give the existing team access to the most advanced models and plenty of tokens. In practice, the result is clear: people's output has already increased by at least a factor of two.

Using frontier models every day also keeps the team highly sensitive. It helps us really understand the boundaries of what these models can do, where they fit, and where they do not. Once we understand those things, we can fold them into NocoBase as part of the product itself.

Live in the Present

I have written so much above about "what lasts." But how long is "long-term," really? Six months? One year? Three years? To be honest, we ourselves cannot judge that with confidence.

But there is no need to let that become a source of anxiety. We are a bridge between AI and enterprises. AI is moving fast, so fast that it is hard to say what things will look like this time next year. But between the speed of AI and the speed of most enterprises in the real world, there is still a gap of many months, if not years. As long as we stay grounded in the actual conditions and needs of those businesses, there will always be time for us to adjust, and perhaps even discover more opportunities along the way.

What we most need to avoid in this process is end-state thinking: the belief that if AI can generate code, then it can do anything; the belief that if AI will eventually take over everything, then whatever we do now is meaningless. The real world in front of us does not look like that.

From a Vast Ocean, We Only Need One Drop

If the needs of enterprises around the world are a vast ocean, then we do not need a bucketful. We only need one drop. That is enough for a team our size to move at our own pace, stay focused on what we love, create value for the users inside that one drop, and find a sense of achievement there.

NocoBase has to stay focused on a precise kind of user. We are not trying to make every enterprise our target customer. Many people may not agree with the things we have listed here as long-term constants. That is fine. As long as the users in this one drop agree, that is enough.

So, Do Programmers and Software Still Have a Future?

I think the answer depends mostly on what we want, and on what we are willing to do.

If the ambition is to build another Salesforce, form a monopoly, and change the world, that is beyond our ability to judge. But for a product like NocoBase, built by a team of fourteen people, I believe there is absolutely a future. Not only are we not going to be killed, we may well have more opportunities ahead of us than before.

NocoBase used to live only inside the no-code crowd, and that crowd was very small. Today, NocoBase has also started attracting attention from AI users. More and more enterprises want to bring AI capability into real business operations. As long as we avoid end-state thinking, and do not assume that the leading model companies are capable of doing everything across every industry, every scenario, and every link in the chain, we will see how diverse enterprise demand really is. On the road to adopting AI, companies will need all kinds of infrastructure layers, along with the services that make them usable.

From that point of view, I believe some programmers, and some kinds of software, have a better future ahead of them than they did before.

Final Note

NocoBase 2.1 was released last week. Try connecting your AI agent to NocoBase and experience it for yourself.

GitHub: https://github.com/nocobase/nocobase