AI will set software engineering on the path to a huge disruption as it continues to evolve. This prediction by Abacus AI chief Bindu Reddy comes in the backdrop of advanced AI tools that are already increasing productivity and are expected to bring about even greater improvements in AI for engineers in the upcoming months.
“Currently, coding assistants increase productivity by around 15%. This number is going to skyrocket to almost 50% in the next 6 months,” Reddy estimated.
Notably, the FAANG companies – Meta (previously Facebook), Apple, Amazon, Netflix, and Google – known for their notoriously low acceptance rates, receive millions of applications annually. Apple accepts approximately 3% of applicants, Meta about 5% after the onsite interview stage, Amazon less than 2%, Netflix under 2% (as of 2019), and Google just about 0.67%.
Now, with AI tools like Cursor, GitHub Copilot, and Claude Sonnet preparing to automate coding tasks, many ‘ghost engineers’ at FAANG companies face the risk of being rendered jobless.
Not to forget, when Tesla chief Elon Musk bought X (formerly Twitter), he laid off over 6,000 employees—roughly 80% of the company’s staff—and the app continues to function smoothly today.
Who are Ghost Engineers?
Recently, in a post on X, Deedy Das, VC at Menlo Ventures, highlighted how several software engineers, not limited to FAANG companies, manage to get away with underperformance while continuing to draw high salaries.
“Everyone thinks this is an exaggeration but there are so many software engineers I personally know, not just at FAANG, who literally make approximately two code changes a month, few emails, few meetings, remote work [for] under five hours per week, for approximately $200,000-$300,000,” Das wrote.
Yegor Denisov-Blanch, a Stanford researcher, defined a ‘ghost engineer’ as someone who performs at 0.1x the efficiency of a median software engineer. “It’s an engineer who severely underperforms, does hardly any work, and often pretends to be ‘very busy’ or working on something ‘very important’,” he told AIM.
“This [practice] unfairly burdens teams, wastes company resources, blocks jobs for others, and limits humanity’s progress. It has to stop,” Blanch added.
Interestingly, Price’s Law, named after sociologist Derek J. de Solla Price, suggests that in any group, the square root of the total members accounts for half the output. For example, in a team of 100, about 10 people contribute 50% of the work.
Blanch predicted that companies, with the help of AI’s increased ability to evaluate work performance and detect inefficiency, will become much better at identifying these ‘ghosts’.
“In our research, we found that there are many ‘ghost engineers’ working from home, but there are also incredibly skilled engineers who thrive in the same environment, free from distractions and endless meetings,” he added.
Even OpenAI’s co-founder Andrej Karpathy agreed with Das’s opinion. “My moment of realisation was when a small group of them I met once openly laughed about it, saying, ‘Yeah, we didn’t do anything for months, our manager is remote and doesn’t care,’ and they all laughed. I realised it’s not even an individual here and there – it’s a shared secret,” Karpathy posted on X.
Sharing a similar observation, Yuchen Jin, co-founder and CTO of Hyperbolic Labs, said, “My realisation came during my PhD when I interned at Microsoft. Honestly, many engineers there were so chill, working just three to four hours a day and rarely writing meaningful code, so I decided not to pursue a FAANG job.”
Meanwhile, Jonathan Grey, who works at Anthropic, said, “I know someone at FAANG whose team or project got cancelled, and they were given four to six months to find a new team. During that time, there were literally no expectations. They just went climbing all day. Near the end, they spent a week or two chatting with people and found a team.”
Blanch said the current situation of software engineers mirrors the well-known 1970s paper ‘The Market for Lemons’, which discusses the economic impact of poorly performing products – in this case, engineers.
“Companies have historically tolerated it, assuming that 10-20% of their workforce will be ‘ghost engineers’. However, this brings down the overall productivity and affects the market value of engineers,” he said.
However, Meghna Chaudhary, a software engineer at Google, disagrees with the report. “I am surrounded by a lot of hard-working people. I haven’t seen this around me, personally,” she said.
Will AI Replace Underperforming Engineers?
In a recent interview, Anthropic CEO Dario Amodei said coding will soon become obsolete, and this shift will happen exponentially. “We’ll find that when AIs can do 80% of a coder’s job, including most of it that’s literally writing code with a given spec.”
He believes that for humans, it will be more about high-level system design, assessing whether the app is well-architected, and focusing on design and UX aspects. “Eventually, AI will be able to handle those as well,” Amodei added.
AIM reached out to Das to get his opinion on the matter. He thinks that jobs at FAANG companies are unlikely to be eliminated due to efficiency gains from AI, as this has historically never happened. “The only reason for headcount reduction is typically when revenue growth cannot support an increase in headcount.”
One can do five hours of work a week and still be very valuable because someone else cannot just come in and do that work, Das said, stressing that people acquire expertise in extremely specific niches.
Similarly, Blanch said that there will always be a demand for skilled engineers and top-tier talent. “The key difference will be in how we identify and retain them.”
“To get a job, you need to be smarter than today’s AI and more agentic than tomorrow’s AI,” said Sahil Lavingia, founder of e-commerce platform Gumroad. He believes software engineers must upskill faster than anyone else.
Lavingia believes AI will “kill remote work” since it can deliver products quickly. He explained that in-office teams can use AI more effectively by working together in real time and getting instant feedback.
François Chollet, the creator of Keras, predicted a few months ago that there would be around 10 million more coding jobs in the next five years, particularly for people with expertise in writing Python code. He added that most of the coding, however, would be done by AI.
“If you could fully automate software engineering (my job), I think that would be great since I could then move on to higher-leverage things. Making software is a means to an end, not the end,” Chollet had earlier said. He added that software engineering is not just about copy-pasting code, but about developing mental models of problems and their solutions.
Research and advisory firm Gartner said in its latest report that GenAI will spawn new roles in software engineering and operations through 2027, requiring 80% of the engineering workforce to upskill.
The post AI is Coming for FAANG’s ‘Ghost Engineers’ appeared first on Analytics India Magazine.
