92% of programmers said they used AI tools at work in 2023, according to a poll and GitHub internal data analysis. Since then, prompt-based technologies have made it possible for anybody to “vibe code” a website or software program, and engineers are being urged to use new technology to increase productivity.
At the same time, early-career tech workers and computer science majors claim that their employment market has dried up. Leading internet companies are making harsh statements about the possibility of a reduction in staff, as proposed by Andy Jassy, the CEO of Amazon, or the risk of slipping behind, as mentioned by the CEOs of Shopify, Fiverr, and Duolingo. Marc Benioff, the CEO and creator of Salesforce, stated that he had no plans to bring on any more engineers in 2025.
Because of this, software development teams have less resources and less negotiating leverage when it comes to pay, benefits, and working conditions. In recent years, many tech businesses have reduced their benefits expenditures while simultaneously laying off thousands of workers.
New artificial intelligence (AI) tools for developers promise to assist programming teams in finding problems and even producing code. In addition to businesses like Windsurf and Cursor, several of the industry’s biggest players, like OpenAI (GPT-5), Google (Jules), and Anthropic (Claude Code), have similar products.
Jay Preall, a senior consultant at Segal, told Newsweek that although this technology is actually quite effective at producing code, the code it produces is a little basic. For more than 30 years, he has been assisting companies with new technology adoption and systems interfaces. It likely gets you 60% of the way there, but you still need the knowledge to recognize “how can I make it quicker and better?
Windsurf’s leadership was recently “reverse acqui-hired” by Google, while the remainder of the firm was sold to Cognition a few days later.
At the time of its acquisition, Windsurf had hundreds of thousands of daily active users and more than 350 commercial clients, including Dell and JP Morgan Chase. Google acquired Windsurf’s founders and a small group of top engineers for its DeepMind division in July for $2.4 billion, which also included a permanent, nonexclusive license to the technology. Cognition acquired the remaining assets and personnel.
The head of product at Windsurf at the time, Anshul Ramachandran, described how the company’s tool was increasing developers’ productivity in an interview with Newsweek prior to the acquisition activity.
According to him, it was the first tool with this agentic capabilities that was actually widely available for software programmers. Naturally, that’s the hottest term right now, but we were the first to bring this up back in November [2024].
Ramachandran is currently at Google, and Windsurf’s solution has subsequently been included into Cognition’s Devin AI service. Senior engineers served as Windsurf’s initial beta testers, he said in an interview. Once they discovered its value, the business also saw outcomes from using it with non-technical teams (after developing certain games, of course).
“All of these people began developing their applications,” he added. They were subject experts and understood exactly what they wanted to create for internal tools, but they were not intended to be apps that were visible to the general public.
Allowing nontechnical teams to design technical tools might result in software licensing savings for sales, marketing, legal, and finance departments.
“They could create exactly what they want instead of paying six figures for an off-the-shelf, B2B SaaS tool that is really just a wrapper around a database,” Ramachandran added.
Windsurf reduced software expenditures by around “half a million dollars” as a result of teams from every department creating their own apps internally.
Ramachandran noted that the hiring of engineers is currently on the decline, but he also said that the startup world is still full of opportunities and that the definition of a developer may broaden because anyone can create something for their own teams if they have a basic understanding of prompts and the power of large datasets.
He does not believe we are getting rid of developers, he remarked. What we term a developer will undoubtedly change, but there will still be more developers. Software is fundamentally a value-creation activity. Now that it is substantially easier to develop software, any firm in the world that utilizes technology as a differentiator will face an opportunity cost of not spending more in software that has only increased.
Because even internal operations involve compliance and other critical steps that a marketing manager or finance associate might not be aware of, Preall notes that businesses still need technical competence in the room. He went on to say that a young engineer’s ability to develop into an experienced one is one of their benefits.
According to him, you still need the specialist to go back and confirm, update, and fix everything. They will eventually never produce specialists, especially for those who are eliminating all entry-level positions.
“Outputs have been so error-filled that a person would have been better off doing the task on their own,” Preall added, citing other issues with these tools. Additionally, he identified a discrepancy in the replies, which further increases the editing and quality-control work required when utilizing the AI partner.
He said that he might ask the same question again in a succession and receive different responses. That is a serious issue. Both responses may be partially accurate, and he must possess the knowledge to determine which portions of each response are accurate before combining them to provide him with the final response.
Some people are doubting the usefulness of these tools because of their inconsistent outcomes. Even if the technology isn’t quite there now, Preall stated that he does believe it will ultimately catch up.
With all of the various moving components involved in a product or feature development cycle, further survey data indicates that engineering teams would not be able to “ship products faster,” and engineering organizations have so far claimed few gains from AI.
Preall remarked, As a developer, let’s say he writes a program and uses a variety of variable names. “When it comes to writing code, it doesn’t know the names of his variables. The code will be fine, but if it doesn’t meet his requirements, he will have to do all the changes. That, he believes, is the main problem.
