This post was originally published on this site.
Yahav Nir (27) from Jerusalem graduated with a degree in computer science from the Open University about two months ago and was sure that the market was waiting for him. During his studies, he worked as a data analyst and systems administrator at Check Point and at a startup, but in recent months he has been looking for a job as a software developer. Some of his classmates have been looking for work for even longer. “The feeling is that there are very few jobs open to bachelor’s degree graduates,” he says.
Nir also says that the jobs offered require prior industry experience. “The requirement is at least two or three years of experience, and preferably with a background in AI. But getting experience like that isn’t so easy. You need huge amounts of data to process, quantities that only companies have. You send in a resume and the recruiters don’t get back to you. They open a job on LinkedIn and close it within an hour because they’ve received enough resumes. It’s understandable; when people can’t find a job that suits them, they send resumes for jobs that don’t suit them.”
Aharon Sinai (26), a third-year computer science student at the Hebrew University of Jerusalem, has for six months been looking for a job in software, testing, or any other entry-level position in the tech industry. “I submitted resumes for 150 jobs, most of which don’t even bother to get back to you,” he says. “It’s different from the way it was in the past – by the third year, people would have already started working. And not just me. All my friends have been searching for months and months; when a job opens, everyone sends in resumes, and then the lucky one gets the job, probably one out of several thousand applicants.”
There are many reasons for this. In addition to the large number of programmers entering the market after years of growth in the number of tech training courses and schools, Ronen Nir, Israel managing director of US-based PSG Equity, explains that tech companies have experienced declines in fundraising, more emphasis on profitability and, of course, the war and callus of reservists. At the same time, he says, more and more companies are introducing development tools, such as Microsoft’s GitHub and Copilot, or Anthropic’s Claude, to replace young software developers. “We’re just at the beginning, and it’s still impossible to prove the connection between AI and unemployment, but I have no doubt that over the next two years we’ll see its impact on the rate of programmer recruitment.”
RELATED ARTICLES
Like many of his peers, Sinai is also concerned that AI tools have already become well-established in tech companies, and some perform tasks that juniors like him have performed in the past. “I haven’t heard about this explicitly from any recruiting company, but we can assume that if someone used to take a few hours to write a particular code, today it can be done in one-tenth the time. However, I still don’t think we can completely give up on human programmers who will work with AI tools, as you need a thorough understanding of what can be done with the code and where its weaknesses lie.”
Meta CEO Mark Zuckerberg openly admitted a few weeks ago on Joe Rogan’s podcast that in 2025, AI systems at Meta and other companies will be capable of writing code like mid-level engineers with, AI engineers eventually replacing human engineers. Around the time, it was reported that marketing software giant Salesforce was laying off about 10,000 employees, and that and that Salesforce founder and CEO Marc Benioff revealed “We’re not adding any more software engineers next year because we have increased the productivity this year with… AI technology that we’re using for engineering teams by more than 30% – to the point where our engineering velocity is incredible. I can’t believe what we’re achieving in engineering.”
Lemonade CEO Daniel Schreiber admitted that AI had helped the company reduce its workforce by 11% in the past two years while at the same time doubling its revenue. “Our development manager thinks that in two years he’ll stop recruiting university graduates, because AI can do the juniors’ work,” he said.
Faster, cheaper, and doesn’t get tired
Some say this is not a future vision but is already happening. “AI has overturned everything,” says the CTO of a growing Israeli startup. “It has set higher standards for joining our company while reducing our willingness to train juniors from the ground up. Our approach now is that new employees must contribute value right from the start.”
The company he co-founded already develops code using generative AI with LLMs, including GitHub, Copilot (which has already become a standard development tool), Claude, (also used for software development purposes), and Perplexity AI, the high-tech “stepdaughter” that is challenging Google with advanced search technology and advanced AI.
Another tool in the kit is early-stage “mouse tracking” startup Cursor, which has collected data on millions of cursor positions made by software developers, to predict the code they will write based on them. Cursor is considered the world’s fastest-growing AI engine startup, and in recent weeks has raised $100 million at $2.6 billion value. In the US, the funding was deemed the fastest in venture capital history, Cursor having raised its seed capital only in August 2023.
Two Israelis also joined the cohort of AI-driven development tools companies this year: Tessl, owned by Snyk founder Guy Podjarny, which raised more than $100 million at a valuation $500 million and Tel Aviv-based Qodo (formerly Codium).
“Team leaders want code written in their own language to fulfill a specific function – the engine writes it for them. They define a problem – it debugs it, and does it better than a junior who doesn’t know our code base,” says the same VP. “But the benefit doesn’t end just with coding or testing. It’s multidisciplinary and brings us insights from physics, mathematics and statistics, and generally from the exact sciences. If you ask it to explain a formula, it provides a much clearer, more detailed explanation than most employees when you ask them. Failures usually happen if it’s not given sufficient context, so you want employees who already have experience with these systems.”
Dror Weiss is CEO of Tabnine, an Israeli company that competes with GitHub and Claude, with a code development engine that works with about 70 giant companies such as Ericsson, AstraZeneca, Cohere, and Credit Agricole. While the new engines are rapidly adopted by growing technology companies, Tabnine works with major corporations looking for turnkey software solutions tailored specifically for them. “All these tools started as a kind of ‘ autocomplete,’ which, as with Google’s engine, completes the line you’re writing based on what it thinks is appropriate. Today, these tools are integrated into the entire software development cycle: development planning, coding, testing, peer review, and project management.”
Weiss says that programmers, especially young ones, need to ask themselves what they can offer that AI cannot. “If once it was enough to bring to the table the ability to translate clear requirements into code, today AI already does it faster, cheaper, and without getting tired. The thing is, that’s how many people started their careers, but today they’re no longer able to compete with technology. When I started out in the 1990s, I excelled at remembering commands and parameters. But if I were starting out today, AI would perform tasks better than me. So, programmers today need to adopt the mindset that characterizes team leaders, one that looks at a problem and breaks it down into its components.”
The jobs most at risk
Grove Ventures general partner Lotan Levkowitz surveyed nearly 100 tech companies and found that 92% already generate at least some of their code using generative AI. “In the previous survey last year, only 50% admitted to this,” says Levkowitz.
However, he said there are areas where organizations still struggle to replace human developers. Testing, for example, where only 21% of managers admitted to automating, project management (10%), infrastructure management and development tools (DevOps) with only 29%.
“When asked what prevents a manager from introducing AI engines, the main concerns are related to the quality and accuracy of the code (58%), legal considerations (44%), and data security issues (42%),” notes Levkowitz. “This means that development departments still don’t fully trust generative AI, which also means it can’t be entrusted to a junior. We see companies where juniors are not allowed to work with AI because they’re afraid of substandard code might be integrated. Because the entire foundation of development departments is changing, they’re optimizing for the near term. If companies like Check Point once built management teams for the coming five years, organizations today can’t afford to do that in an environment where everything is changing.”
According to the survey, software development managers are at the greatest risk of being replaced by AI, followed by software developers and software architects. Product managers are relatively protected from the AI revolution, followed by technical tool developers and data scientists.
How is it that mid-level development managers in particular are endangered? Levkowitz explains that should AI engines take on the bulk of code development work, developers and engineers will be “upgraded” to a kind of product manager or software architect. “The engineer needs to specialize in problem solving, systems thinking and architectural planning,” says Levkowitz. “They need to have the ability to make decisions about design and user interface issues, involve customers in product development, and prioritize tasks.”
According to Levkowitz’s analysis, currently, junior software programmers are at a disadvantage because companies hesitate to hire them, fearing they lack the skills to work with AI. But all is not lost: “My wife is an architect. She learned how to design buildings but has never laid a brick in her life. On the other hand, software architects have had contact with code. Very soon, we’ll reach the point where a programmer van become a software architect through AI-based training, brick by brick. Instead of giving juniors who’ve never flown a plane an F-15, we’ll start them out with a Piper. “
Recruiting juniors was uneconomical
For over 20 years, Miriam Shtilman managed the operational and fundraising aspects of medical algorithm company Algotec, interviewing hundreds of mathematicians, engineers, and AI experts. Today, she is a partner at Tal Ventures, a venture capital fund that invests in dozens of companies, with an emphasis on deep tech. Shtilman says the underlying mechanism for hiring and training juniors is undergoing a major upheaval. “In the past, there was a shortage of engineers, so they would hire outstanding juniors, and after the pool was exhausted, they would hire regular juniors from universities, and then from colleges. They would join an enterprise, and their salaries would increase by 15%-30% each year, reflecting the cost of the organization’s investment in them. It got to the point where the cost of training could reach one-fifth of an experienced engineer’s salary, because a fifth of the team’s time – team leaders, software architects – would be dedicated to training that outstanding junior. This turned out to be uneconomical, as juniors would leave within two to three years.”
“10X engineers” is industry parlance for experienced and sought-after engineers, developers who are ten times more effective than average. “These are people who used to help developers and juniors with less ability and experience with simple, often boring tasks. Those 10X engineers would guide them and share their experience, but today that’s no longer needed. The 10X engineers are becoming 100X, without wasting time on training and mentoring.”
Shtilman sees the accelerated rate of change mainly in the new generation of companies: “We won’t see thousands of engineers being laid off overnight,” she says. “But new startups are already far leaner than was customary on the software development side,” she says. “Someone starting a new company no longer needs to hire five high-quality, experienced engineers at once – that’s an expensive resource. One developer equipped with all the tools they need can produce a huge amount of code.”
Many compare AI to the industrial revolution; the technology will hurt some jobs but overall add lots of jobs to the sector.
“I wouldn’t compare what’s happening here to the industrial revolution. It’s more like the second printing revolution, the one that brought in digital printing and eliminated typesetters.”
Shtilman believes AI’s main impact is in code-writing in all types of languages, less in tangential development areas such as tools development, code maintenance, machine learning modules development or cloud applications. She recommends specializing in data engineering or applying AI to a specific content area, but mostly admits that expressiveness, especially oral and written expression, has become very important: the ability to formulate a comprehensive prompt with the right context and in fluent English. “Writing code is actually writing text, which is where AI comes in, but this means that now, even software architects, seniors and very experienced developers, can go back and write code in any language they want, even Chinese,” she says.
Maybe learn something else?
Even Jensen Huang, CEO AI chip giant Nvidia, was eventually forced to address the matter, as one of its creators. A few months ago, he surprised the market when he claimed that “over the course of the last 10 years, 15 years, almost everybody… would tell you: it is vital that your children learn computer science, everybody should learn how to program. And in fact, it’s almost exactly the opposite. It is our job to create computing technology such that nobody has to program, and that the programming language is human.”
When software coding will be handled by AI, humans will be free to specialize in other professions such as biology, agriculture, education, or industry, he claims. If even Huang is recommending not studying computer science, what should one study instead? “It’s clear there’s a need to deepen your knowledge in areas where LLMs cannot replace you,” says Shtilman, who mentions physics, electrical engineering and statistics as professions with not easily replaceable skills.
Those who did study computer science and are having difficulty finding work often describe the huge gap between their studies and reality. “The university is here to train future researchers, not necessarily to train high-tech workers,” says Yahav Nir, an Open University graduate. “They are trying to take steps toward a more applied program, but you see that people graduate from academia don’t understand their options. I signed up for a software engineering workshop offered by the university in conjunction with Microsoft, which was supposed to introduce us to the development field, but the program was canceled after one meeting.”
Aharon Sinai is in his final year at the Hebrew University. After not finding a suitable job, he began volunteering part-time at a tech company which develops a system for hospitals. He also admits there is “No relationship between curriculum and workplace format and lifestyle,” but is confident that a degree enables graduates to provide better AI commands, and to better understand their output. “That’s the difference between writing code and writing smart code,” he says. “A person who doesn’t master data security and the complexities of code will lose their way very quickly.”
Warnings from industry executives and the difficulties faced by graduates raise questions of whether studying computer science is justified. Today, it is still one of the most sought-after study programs in academia, and until recently, at least, it was considered the main gateway to a promising tech career. Prof. Shimon Schocken, founding dean of the Efi Arazi School of Computer Science at Reichman University, agrees with the claim that you don’t have to study computer science to work in high-tech. He says, “You can study mathematics, physics, statistics, or life sciences; all of these are excellent preparation for the job market. If you want, you can also study theater, if you also take courses in programming and algorithm. The specific subject is less important – more the quality of the university, the lecturers, and the broadening of your horizons.”
The skills beyond AI capabilities
Reichman University, however, is not sitting idly by. It is also developing new applied subjects, in conjunction with industry. The university recently established The Google and Reichman Tech School, offering a combination of academic courses and applied training, such as AI-based systems development, software development, data mining, and more, in a quick and intensive six-month training. “One reason why we established this school is that we have no idea what academia will look like in a decade, and whether, in parallel with academic degrees, modular ‘micro-degrees’ that can be assembled and disassembled as needed, can also be offered,” says Schocken. “This school has become a laboratory where we test new ideas for job training.” He says that by 2023, nearly 90% of Reichman graduates had found places in the industry, but does not deny the phenomenon of junior unemployment. “This phenomenon began two years ago and has worsened over time. There’s no denying it. Juniors are disadvantaged twice: first, by the tendency not to hire juniors in the first place, and second, by the fact that the general demand for workers has declined and, naturally, the market favors those with experience.”
Reichman has also made changes to the regular academic degree in computer science: new courses have been added for software development and product management with AI tools, programming in web, cloud, and mobile environments. At the same time, it has opened new programs that integrate computer science with business, entrepreneurship, cognition, and medicine, and a new master’s degree program in machine learning and data mining. All of this is in parallel with the classic undergraduate, graduate, and doctoral degree programs in computer science. Reichman is also starting to use chatbots designed for specific courses that guide and assist students in independent problem-solving, without revealing the answers in advance.
But will the degree undergo a revolution? Schocken argues that the fundamentals of computer science and common sense are irreplaceable, and these will make the difference for future AI developers. “Almost always, the answer you get from the language model is not satisfactory. You have to treat it with skepticism, a draft that needs to be refined and improved according to many variables, debugged and optimized, adapted to other systems, and ensure that it is efficient in terms of runtime and memory usage.” In general, Schocken argues, “The theory of computer science is critically important, and it is one of the reasons why academia and industry in Israel are also global leaders. From the outset, the founding generation of computer science in Israel emphasized the study of mathematics, logic, algorithmics, and statistics; these are not going anywhere and will continue to stand at the forefront of leading academic programs.”
Prof. Sara Cohen, Dean of the Selim and Rachel Benin Department of Computer Science and Engineering at Hebrew University of Jerusalem, says, “We weren’t surprised by the AI revolution. Hundreds of our students have been studying AI since 2013 as part of a mandatory course, alongside advanced learning courses in a variety of fields such as image processing, natural language processing, and voice processing.”
“In addition, our advanced courses stress skills beyond the capabilities of generative AI systems, including creative thinking, analytical skills, and higher-order reasoning – skills critical to understanding the products of AI- based systems and dealing with the cost of error. After all, there is no absolute solution in AI; several options will always be acceptable, and the human element must therefore know how to make informed decisions based on the professional knowledge accumulated in studies and career.”
Cohen adds “We must view the ‘junior crisis’ in a measured fashion. Ultimately, the human factor must master the professional fundamentals to understand AI’s language and output, monitor, refine and adapt the AI to the desired product, and to improve its efficiency in everyday use. When AI tools are properly leveraged, the juniors entering the job market today are stronger and better prepared than ever before.”
Prof. Roded Sharan, head of the School of Computer Science at Tel Aviv University, also admits that AI is revolutionizing the curriculum. “It is dramatically changing research and teaching; today, more than half of the faculty and students are engaged in AI-focused research,” he says. In fact, last month, the long-standing school changed its name to the “Blavatnik School of Computer Science and AI.”
The school offers about 20 AI-oriented courses. It is in the process of approving a mandatory “Introduction to Artificial Intelligence” course, and planning to launch a new course dealing with programming that will also include engineering fundamentals, in collaboration with the faculties of exact sciences and engineering. It is also conducting courses taught jointly or led by technology experts from Google and Microsoft.
But according to Sharan, there is no substitute for the basics of computer science: abstraction, breaking down complex problems into smaller problems, proof, criticism, and independent thinking. As an example, Prof. Sharan recalls a short story by Isaac Asimov, “Insert Knob A in Hole B.” “Two astronauts travel to a space station with equipment that must be assembled. The problem is, the instructions are complicated, so they ask Earth to send them a robot to read the instructions and assemble the equipment for them. But when the spacecraft lands with the robot, they discover that it, too, comes with complicated assembly instructions.”
“We,” he says, “equip developers so they will understand these machines by themselves.”
Published by Globes, Israel business news – en.globes.co.il – on February 23, 2025.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2025.