AI Coding Assistants: The "Killer App" for Generative AI?

AI-powered coding assistants have attracted nearly £750 million in funding since the start of last year, a clear signal that software engineering is emerging as the first "killer app" for generative artificial intelligence.

Companies such as Replit, Anysphere, Magic, Augment, Supermaven, and Poolside AI have raised a combined £340 million in funding this year alone, bringing the total since January 2023 to £720 million, according to Dealroom.

This surge in investment highlights the significant impact AI is having on the world of computer programming. Many believe it signifies the beginning of a profound transformation in the software development landscape.

"Software engineering and coding are currently the primary areas affected by AI," commented Hadi Partovi, CEO of the educational non-profit Code.org and a seasoned Silicon Valley investor. "At this juncture, software engineering without AI is akin to writing without a word processor."

The growing conviction in Silicon Valley regarding the benefits of AI coding stands in contrast to the scepticism among some investors about the economic benefits of generative AI and the likely returns on Big Tech's projected multi-billion-pound investment in computing infrastructure to support the technology in the coming years.

Hannah Seal, a partner at Index Ventures, which has invested in Augment, alongside Eric Schmidt and others, highlighted the ease of monetising AI when embedded into existing workflows. "The benefit becomes instantly visible," she explained.

Seal believes the key questions for AI tools are their time to value and the significance of the value-add. "With coding co-pilots, the answer is very clear," she asserted.

The AI enthusiasm has led to a surge in competition within the sector, with start-ups and tech giants like Microsoft, Amazon, Meta, and Google vying for dominance. All are developing AI assistants and agents that can write and edit computer code.

An executive on Code.org's board, which includes David Treadwell, Amazon's head of ecommerce, and Kevin Scott, Microsoft's chief technology officer, recently informed Partovi that their company would cease hiring individuals who code without AI by the end of the year.

"The easier programming becomes, the higher the demand grows, as much more technology can be built," Partovi added.

Microsoft-owned GitHub, the world's largest software development platform, was among the first to leverage a large language model — the technology behind ChatGPT, capable of generating text, images, or code — to create a coding assistant.

"When using GPT-3, OpenAI's first major model, we quickly realised that it was so proficient at writing code that we could build a product around it," said Thomas Dohmke, CEO of GitHub, which was acquired by Microsoft for £6 billion in 2018.

This prototype evolved into GitHub Copilot, an AI coding assistant launched widely in 2022 and now boasting nearly two million paying subscribers. "The model currently writes code that surpasses the average developer," Dohmke declared.

As of April, GitHub's revenue had increased by 45 per cent year-on-year, and according to Microsoft CEO Satya Nadella, its annual revenue run rate reached £1.6 billion at the beginning of this month.

"Copilot contributed over 40 per cent of GitHub's revenue growth this year and is already a larger business than all of GitHub was when we acquired it," Nadella revealed during a July 30 earnings call.

Over 77,000 organisations, from BBVA, FedEx, and H&M to Infosys and Paytm, have adopted the two-year-old tool, reflecting a 180 per cent increase year-on-year, Nadella stated.

Despite the enthusiasm, IT departments within large companies remain cautious about the security implications of using automated programming tools to produce production-grade code.

Dohmke assured that AI-generated code would not be deployed without manual checks and balances.

"In general, we observe productivity gains ranging from 20 to 35 per cent among enterprises that have shared their internal statistics," Dohmke said, citing customers like Latin American ecommerce giant Mercado Libre and professional services group Accenture.

A McKinsey analysis from last year indicated that the direct impact of AI on software engineering productivity could range from 20 to 45 per cent of current annual spending on the function. The benefits include generating initial code drafts, code correction, and refactoring.

"By accelerating the coding process, generative AI could shift the required skill sets and capabilities in software engineering towards code and architecture design," McKinsey concluded.

Software engineers confirm that they have already incorporated AI assistants into their daily workflow, finding them not only faster but also more creative.

"I personally code daily with GitHub Copilot, often alongside ChatGPT," remarked Marc Tuscher, a deep learning scientist and CTO of Sereact, a German robotics start-up.

GitHub's tool is particularly helpful for "repetitive tasks," such as user interfaces and back-end development, while he uses ChatGPT for more abstract problem-solving.

"ChatGPT will suggest some classical ideas, recent research papers, and then you can ask, 'how would this be done in Python?' and it produces code," Tuscher explained. "Both tools are truly remarkable."

While all programmers he knows use these products, and it "fundamentally changes how we work," Tuscher maintains that the tools are powerful assistants rather than replacements for coders.

"No GenAI understands good software architecture or how to integrate systems," he added. "That's something we still need to think through ourselves."