Yesterday I was chatting with a friend. I mentioned a project I've been working on—using AI to reduce hardware TCO costs, with automated tuning and remote operations.
He suddenly said: "Aren't you making yourself public enemy number one?" (This project later became our open-source AIMA.)
I was taken aback.
He explained: a large company with hundreds of employees used to have an entire team dedicated to performance testing and tuning. Now you write a piece of software, stuff a few Agents in it, and the job is done. Aren't you basically putting those hundreds of people out of a job? And the operations team—what used to require opening a ticket and going through a process is now handled automatically by an Agent. Their KPIs disappear entirely.
After hearing him out, I had a different thought.
Big Companies Have Become a Burden
The processes, systems, and organizational structures accumulated over the past are not moats in this AI transformation—they are baggage.
A small team of three to five people can say "let's try this" and decide on the spot. A week later, they're already using the best models available.
What about big companies? They need pilots, approvals, and have to consider fairness—who gets access and who doesn't? Will others complain? Back and forth for two or three weeks, and maybe only one or two people are piloting it.
Having worked at major tech companies, I've seen this scenario plenty of times. Just take "AI tool procurement"—it has to go through security review, compliance review, legal review, and IT review. Every department has its own processes and concerns, and each round of review takes one to two weeks. By the time the tool actually reaches engineers, the small team has already iterated three or four versions.
What's worse is internal politics. You give one team Claude Pro accounts, and other teams immediately ask: why them? Either everyone gets it, or no one does. Everyone? Where does the budget come from? Another round of approvals. The "fairness" culture of large companies becomes the biggest obstacle when rapid trial and error is needed.
Then there's data security. Large companies are inherently conservative in their data strategies; many AI tools aren't allowed to touch internal data. The result is that engineers get the tools but can only use them for trivial tasks. The scenarios that could actually improve efficiency are completely locked down.
How Much Difference Can One Week Make Now?
With access to the best models, one week of productivity can mean a world of difference. Not to mention two or three weeks.
Here's a real example. Our team of three built an MVP for an AI operations system from scratch—automated tuning, intelligent alerting, remote diagnostics. From the first line of code to a working demo, it took five days. In the old days, just the requirements review and technical design would have taken two weeks, and development scheduling at least a month.
This isn't a 10% or 20% efficiency improvement. It's an order-of-magnitude difference.
It suddenly hit me: everyone has been leveled.
One person, or three to five people, uses the same tools as ByteDance or Google. The gap isn't as large as imagined. More importantly—no one knows how to collaborate yet. One person runs incredibly fast, but how do others keep up? How do they coordinate? No one has figured this out.
This is exactly the window of opportunity.
How Entrepreneurs Should Fight
Since big companies are slow, entrepreneurs must be fast. But fast doesn't mean charging ahead blindly; it means being fast with strategy.
First, choose the right battlefield. Don't compete with big companies on infrastructure—large model training, compute platforms, that's their home turf. Look for gaps that big companies "can see but can't reach." For example, AI implementation in vertical industries: no matter how complex an industry's processes are, three to five domain experts plus Agents can be ten times more efficient than a twenty-person team from a big company.
Second, build moats with speed. During this window, every week you run ahead is another week of accumulated knowledge and data. While big companies are still discussing whether to do it, you've already completed three rounds of iteration with customers. This first-mover advantage can't be caught up with money.
Third, embrace the identity of "public enemy." My friend was right—you are indeed competing with hundreds of people for their jobs. But look at it another way: if you can do with three people what takes them hundreds, your value density is one hundred times theirs. Customers don't care how large your team is; they only care whether you can solve the problem. And honestly, those hundreds of jobs will be affected by AI sooner or later. The only difference is whether it's your product or something else. Rather than letting big companies slowly optimize internally, it's better to let entrepreneurs drive this process with better solutions.
FOMO Is Not Unfounded
In a month or two, someone will always figure out an optimal organizational paradigm and a value-creation process. By then, how much time will you need to spend experimenting and finding your rhythm from scratch?
Looking back at the mobile internet wave, companies that entered around 2012 and those that entered around 2015 had drastically different outcomes. Meituan and Didi—China's dominant food delivery and ride-hailing platforms—both rushed in during the earliest stage of the window. Doing the same thing two years later meant the market was already saturated, and funding was no longer available. This AI window is shorter than the mobile internet one, because AI itself is accelerating everything—including how fast competitors catch up to you.
I used to think that "can't wait" anxiety was a bit overblown. Now I understand.
Trillions of dollars in value—the puzzle is already 80% complete. What's left is for a small team to spend some time and execute it well.
At times like this, every second you wait is a waste.
I never used to understand why so many people would give up stable positions and high salaries to start companies. Now I get it—that feeling of being on the verge of realizing some grand vision makes you feel like waiting even one more second is a waste of time.
It's not about personal gain or loss. It's because some societal value truly can't wait.
Your First Step
So I want to say: go try AI coding. The barrier to acquiring these capabilities isn't high right now. Coding Agents Are a Meta-Ability—they allow ordinary people to control their computers with an extremely low barrier to entry. They give everyone a relatively equal opportunity to ride this wave.
You don't need to quit your job, raise funding, or even have a complete team. Start with a weekend project—use AI to automate the most annoying thing in your work. When you personally experience "solving in three hours what used to take three days," you'll understand why I say this window won't wait.
Miss it, and it might really be gone for good.