Compared to traditional software development, while AI agents still have numerous security and uncertainty flaws, their greatest potential lies in:
First, they can improve their own source code.
Second, they can spontaneously seek new methods to achieve a specific goal.
Third, they can spontaneously engage in complex value exchanges with other AI agents in an open space, accomplishing tasks that were previously impossible or unimaginable for traditional software.
Especially the third point: if a consensus is reached on payment methods, the spontaneous and continuous collaboration and evolution of numerous AI agents will form the rudiments of a living organism. Current tools like OpenClaw are still toys, but they have the potential to build an abstraction layer on top of current hardware operating systems, becoming the primary way for users to interact with hardware. This is incredible; Microsoft and Apple will absolutely not allow this to happen and will certainly build their own AI agent ecosystem, using their powerful distribution capabilities to marginalize competitors.
Nevertheless, the vast majority of developers will ultimately just be cannon fodder, because they have no monopolistic advantage whatsoever, relying purely on the dopodium effect brought by AI agents and hot money to sustain themselves.
In the future, there will likely be new business models that are unimaginable now, capable of successfully capturing value. Just like the microcomputer revolution began in 1975, you could have simply stood by and watched for eleven years until Microsoft went public in 1986, then waited another 40 years and reaped a four-thousand-fold return.
