This artificial intelligence already writes 20 % of the Salesforce icon. This is why developers are not concerned

when man CEO Dario Amoudi He announced that artificial intelligence will Write 90 % of the code Within six months, he prepared the world of coding for collective extinction. But inside SalesforceA different fact has already been formed.

About 20 % of everyone Summit code He wrote in the last 30 days AgentforceJayesh Govindarajan, Slesforce Ai, told me during an interview with him recently. His team not only follows the code created, but the code has already been published in production. The numbers reveal an acceleration that is impossible to ignore: 35,000 active monthly users, 10 million lines of acceptable software instructions, and internal tools to provide 30 thousand hours developed.

However, Salesforce developers do not disappear. They are developing.

“The vast majority of development – at least what I call the first draft of the code – will be written by artificial intelligence,” Juvenandaragan admitted. “But what the developers do with this first draft may have changed mainly.”

From the code of code to strategic control: How developers have become technical pilots

Software engineering has always been mixed with millley. Now AI deals with the latter, pushing the developers towards the previous.

“You move from a purely technical role to a more strategic role,” Jovindarajan explained. “Not only” I have something for his son, so I will build it, “but” What should we build? What does the customer really want? “

This transformation reflects other technological disorders. When calculator machines replaced manual account, mathematicians did not fade – they dealt with more complex problems. When digital cameras killed the dark rooms, photography expanded instead of contracting.

Salesforce believes that the code works in the same way. As AI drops the cost of creating software, developers earn what they always lack: time.

“If a preliminary working model takes weeks, it will now take hours,” said Juvenandaragan. “Instead of showing clients a document describing what you may create, you can simply deliver the working programs. Then repeat based on their reaction.”

“Mobile Coding” here: Why do software engineers now organize artificial intelligence instead of writing every matter

Programmers began to adopt the so -called “Coding Vepi-The term is formulated by the founder of the Openai co-Andrej Karpathy. This practice includes giving high-level trends instead of accurate instructions, then improving what it produces.

There is a new type of coding that I call “atmosphere encryption”, which completely succumbs to feelings, adopts the si, and forgets that the code exists. This is possible because LLMS (such as the Monnet Monnet) becomes very good. Also, I just talk to the composer with Superwhisper …

Andrej Karpathy (Karpathy) February 2, 2025

“You just give him a kind of high -level direction and let artificial intelligence use his creativity to create a first draft,” said Jovindarajan. “It will not work exactly as you want, but it gives you something you play with it. You improve parts of it by saying:” This looks good, do more, or “these jacy buttons, you don’t need it.”

The process compares the music cooperation: “Amnesty International puts the rhythm while the developer forms the melody.”

While artificial intelligence excels in generating direct business applications, Govindarajan admits that it contains limits. “Will you build the next generation database with a vibrant coding?

The necessity of new quality: Why should test strategies develop because artificial intelligence generates more production code

Artificial intelligence does not write a different symbol – it requires different quality control. Salsforce has developed it AgentFORCE testing center After discovering that the software instructions created by machine guns require new verification methods.

“These are random systems,” Govindarajan explained. “Even with a very high accuracy, there are scenarios where you may fail. You may fail in step 3, step 4, or step 17 out of 17 steps you take. Without the appropriate test tools, you will not know.”

The non -specific nature of artificial intelligence outputs means that developers should become experts in testing borders and preparing handrails. They need to know not only how to write code, but how to evaluate it.

Beyond the generation of the symbol: How artificial intelligence presses a life cycle of developing the entire software

The shift extends beyond the initial coding to include the full program life cycle.

“In the construction phase, the tools are understood and intelligently extended, which accelerates everything.” “Then the test comes – the slope tests are generated, and the creation of tests for the new symbol – which Amnesty International can deal with.”

This comprehensive automation creates what Govindarajan calls a “more strict loop” between the idea and implementation. The fastest developers can test and refine, and more ambitious they can become.

The algorithm thinking is still important: Why do the basics of computer science remain necessary in the era of artificial intelligence

Govendrajan often fields are questions of anxiety about the future of software engineering.

“I was constantly asked if people should study computer science,” he said. “The answer is yes at all, because the algorithm thinking remains necessary. Destroying big problems into controlled pieces, understanding what can solve problems, and modeling user needs – these skills become more valuable, not less.”

What changes is how these skills appear. Instead of writing each letter solution according to the letter, developers direct artificial intelligence tools towards optimal results. Man provides judgment. The device provides speed.

“I still need a good intuition to give the correct instructions and evaluate the output,” Jovindarajan stressed. “This requires a real taste to look at what artificial intelligence produces and get to know what it succeeds and what does not work.”

Strategic height: How developers became business partners instead of technical perpetrators

Since the coding itself becomes goods, the roles of the developer are directly related to the work strategy.

“The developers are taking supervisory roles, directing clients performing their behavior on their behalf,” Jovindarajan explained. “But they remain responsible for what is being published. Back still stops with them.”

This height places developers closer to decision makers and farther than implementation details-upgrade instead of cancellation.

Salesforce supports this transition using the tools designed for each stage: Agentforce for developers, the code generation processing, and enables Builder Agent Thankation, and the Agentforce test center guarantees reliability. Together, they make a platform for developers to grow to become these expanded roles.

The company’s vision is flagrantly contradictory to the narration “the governed developer”. Instead of coding themselves in statute of limitations, software engineers who adapt themselves may find more important than ever.

In a field where innovation is routine, artificial intelligence represents the most powerful software conversion program yet – it is not only converted to how the code is written, but who writes it and why. The developers are ready to upgrade their own mental models, the future seems less similar to finishing and more like a transgression.

Daily visions about business use cases with VB daily

If you want to persuade your boss at work, you have covered VB Daily. We give you the internal journalistic precedence over what companies do with obstetric artificial intelligence, from organizational transformations to practical publishing operations, so that you can share visions of the maximum return on investment.

Read with us privacy policy

Thanks for subscribing. Check more VB news bulletins here.

An error occurred.