How AI will change software engineering – with Martin Fowler

In this podcast episode, Martin Fowler, a renowned software engineer and author, discusses the transformative impact of AI on software engineering, comparing it to the historical shift from assembly language to high-level languages. He emphasizes the shift from deterministic to non-deterministic systems, which challenges existing software practices like testing and refactoring. Fowler also highlights the potential of AI to enhance prototyping and understanding legacy systems, while cautioning against over-reliance on AI-generated code without understanding, which could hinder learning. He reflects on the evolution of agile practices since co-authoring the Agile Manifesto, noting improvements but also areas for further advancement. The conversation touches on the importance of mentorship, the enduring value of core software skills, and adapting to AI's influence on the industry.

• The transition from assembly language to high-level languages marked a significant shift in software development, comparable to the current impact of AI.
• This shift involved moving from hardware-specific programming to more abstract, flexible coding paradigms.
• The introduction of high-level languages allowed for the creation of abstractions, which are crucial for maintainable and flexible code.

"The comparable thing would be the shift from assembly language to the very first high-level languages."

• This quote highlights the magnitude of the shift AI is causing in the tech industry, likening it to the historical move from assembly to high-level languages.

"The biggest part of it is the shift from determinism to non-determinism."

• This statement emphasizes the fundamental change AI introduces, moving from predictable to unpredictable environments, altering how developers approach problem-solving.

• Vibe coding is useful for quick explorations and disposable projects but not suitable for long-term applications.
• It lacks the learning loop, an essential component for developing sustainable and scalable software solutions.

"It's good for explorations. It's good for throwaways, disposable stuff, but you don't want to be using it for anything that's going to have any long-term capability."

• This quote underscores the limitations of vibe coding, suggesting its role is more experimental rather than foundational in software development.

• High-level languages enable developers to create their own abstractions, which are essential for solving complex problems.
• Building abstractions involves creating a language within a language to address specific issues effectively.

"The building of abstractions that's I think to me a key element of high-level languages."

• This statement highlights the importance of abstractions in programming, which allow for more efficient and effective problem-solving.

"You're both solving the problem and creating a language to describe the kinds of problems you're trying to solve in."

• This quote illustrates the dual role of programming: developing solutions while simultaneously crafting a language to articulate those solutions.

• AI introduces non-determinism into software development, requiring a new mindset and approach to coding.
• Developers must consider tolerances and potential security issues due to the unpredictable nature of AI-driven tools.

"Suddenly you're working in an environment that's non-deterministic which completely changes."

• This quote points out the paradigm shift AI brings, necessitating a departure from traditional deterministic thinking in software development.

"We need probably some of that kind of thinking ourselves. What are the tolerances of the non-determinism that we have to deal with?"

• This statement suggests that developers must adopt engineering principles like tolerances to manage the uncertainties introduced by AI.

• AI enables rapid prototyping and experimentation, allowing developers to create and test ideas quickly.
• The ability to generate code and abstractions through AI tools opens up new possibilities for software engineering.

"One area is one that has got lots of attention already is the being able to knock up a prototype in a matter of days."

• This quote highlights the potential of AI to accelerate the development process, making it easier to prototype and iterate on ideas.

"Using the LLM to co-build an abstraction and then using the abstraction to talk more effectively to the LLM."

• This statement describes a novel approach to leveraging AI for creating and refining abstractions, enhancing communication between developers and AI tools.

• Rapid exploration allows individuals to explore ideas quickly without extensive commitment.
• There's potential for non-developers to engage in software exploration, though caution is advised.
• The concept of "throwaway explorations" is valuable within certain boundaries.

"People can go, hey, I not really quite sure what to do with this, but I can spend a couple of days exploring the idea much more rapidly than I could have before."

• This quote highlights the ease and speed of exploring new ideas without heavy investment.

• LLMs can aid in understanding and modernizing legacy systems.
• ThoughtWorks has adopted LLMs for legacy code, indicating significant success in this area.
• The process involves semantic analysis and graph databases to trace data flow in legacy systems.

"We put actually understanding of legacy systems into the adopt ring because we said yeah you if you're doing any work with legacy systems you should be using LLMs in some way to help you understand."

• This quote emphasizes the importance of LLMs in comprehending and updating legacy systems.

• Modifying existing systems with LLMs remains uncertain and challenging.
• The effectiveness of LLMs in safely modifying legacy code is still under exploration.

"Well, we know that that LLMs can help us understand legacy code. Can they help us modify legacy code in a safe way? It's still a question."

• This quote underlines the ongoing uncertainty in using LLMs for modifying legacy systems safely.

• AI's role in team-based software development is still evolving.
• Teams are necessary despite AI advancements due to the constant demand for software.
• Effective collaboration with AI in team environments is an area of active investigation.

"We still need a team of 10 people to build what a team of 100 people needed to build and we will always want this stuff."

• This quote stresses the continued need for teamwork in software development, even with AI advancements.

• Vibe coding involves generating code without deep understanding or review.
• It is suitable for exploratory and disposable projects but lacks a learning component.
• The absence of a learning loop in vibe coding hinders skill development and code refinement.

"When you're using vibe coding in this kind of way, you're actually removing a very important part of something which is the learning loop."

• This quote highlights the critical downside of vibe coding: the lack of learning and understanding.

• The use of AI tools leads to increased code output, complicating code review processes.
• Engineers face challenges in maintaining rigorous code reviews with the increased volume of code.

"There's a lot more code going out there, a lot more code to review, and they're asking, 'How can I be vigorous at code reviews when there's just more and more of them than before?'"

• This quote illustrates the difficulty of maintaining quality code reviews amidst increased code production.

• LLMs facilitate learning in unfamiliar environments and technologies.
• They provide starting points and exploratory guidance in new domains and APIs.

"With the LLM, he can learn a bit about it because he can try things out. And if you take it with that exploring sense..."

• This quote showcases the exploratory learning potential of LLMs in unfamiliar areas.

• Testing remains crucial when integrating LLMs into software development.
• LLMs may produce unreliable results, necessitating thorough testing and verification.

"You've got to really focus a lot on making sure that the tests work together."

• This quote emphasizes the necessity of rigorous testing when using LLMs in development.

• Specification-driven development with LLMs involves creating detailed specs for AI to execute.
• The approach requires careful balance to avoid the pitfalls of waterfall development.
• Tight loops and thin slices of development are crucial for effective spec-driven development.

"To me, the key thing is you want to avoid the waterfall problem of trying to build the whole spec first."

• This quote highlights the importance of iterative development in specification-driven processes.

• LLMs can aid in creating domain-specific languages that bridge communication between developers and domain experts.
• The goal is to make business logic understandable to non-programmers, enhancing collaboration.

"What you're trying to do, of course, is create that communication route where a non-programmer can at least read what's going on and understand it enough to be able to find what's wrong about it."

• This quote underscores the potential of LLMs to improve communication and understanding between technical and non-technical stakeholders.

• Large enterprises face unique challenges with legacy systems, regulations, and complex business needs.
• LLMs offer potential solutions but require careful consideration and implementation.

"Suddenly the software developers are a small part of the picture and there's very complex business things going on that we've got to somehow interface in."

• This quote highlights the complexity of software development in large enterprises and the potential role of LLMs in addressing these challenges.

• Startups are more likely to adopt new technologies quickly due to having less to lose and more to gain, whereas larger enterprises are more cautious due to established customer bases and risk aversion.
• Even traditionally slow adopters like government organizations are evaluating and aware of new technologies, indicating widespread awareness.

"Startups... have nothing to lose. They have zero customers. They have everything to gain."

• Startups are driven by the need to innovate and gain a competitive edge, often leading them to embrace new technologies rapidly.

"You start to have a few customers in a business you're starting to be a bit more careful."

• As businesses grow, their approach to technology becomes more cautious due to increased responsibilities and risk factors.

• Large enterprises are not monolithic; different parts of the same company can have varying levels of risk tolerance and innovation.
• Small, adventurous teams within large organizations can drive significant innovation, despite the overall cautious nature of the enterprise.

"Small portions of these companies can be very adventurous and other portions can be extremely not so."

• Within large companies, there are pockets of innovation that drive change, even if the overall company culture is risk-averse.

• Refactoring involves making small, incremental changes to code to improve its structure without altering its external behavior.
• The concept of refactoring was popularized in the late 1990s and has been a key part of software development practices.
• The second edition of the book "Refactoring" was updated to reflect changes in programming languages and practices over 20 years.

"I was always someone who liked going back to something I'd already written and make it more understandable."

• The speaker values clarity and comprehensibility in code, which refactoring helps achieve by improving code structure.

"Each step is so small that it's not worth doing but you string them together and you can really do amazing things."

• Refactoring focuses on making a series of small, manageable changes that collectively lead to significant improvements in code quality.

• Refactoring has become a common term in software development, although it is sometimes misused to describe any change to a program.
• Automated tools have made refactoring more accessible, but understanding the principles remains important for effective code management.

"Refactoring became a word and of course like all of these words got horribly misused."

• The term "refactoring" is often used incorrectly to describe any code change, rather than the specific process of small, behavior-preserving transformations.

• The rise of AI tools generating large amounts of code quickly increases the importance of refactoring to maintain code quality.
• Combining AI with traditional tools could enhance the effectiveness of refactoring practices.

"If you're going to produce a lot of code of questionable quality, but it works, then refactoring is a way to get it into a better state while keeping it working."

• Refactoring can improve the quality of AI-generated code, ensuring it remains functional and maintainable.

• Software architecture patterns were once a major focus in the industry, providing a shared vocabulary for discussing complex systems.
• Patterns help in understanding and communicating system designs, although their popularity has waned in recent years.

"What you're doing with patterns is you're trying to create a vocabulary to talk more effectively about these kinds of situations."

• Patterns serve as a language to facilitate communication and understanding of software architecture among developers.

• The decline in pattern usage may be attributed to the rise of cloud services and well-architected building blocks provided by hyperscalers.
• Startups and newer companies often eschew traditional patterns in favor of more flexible, informal approaches.

"Patterns are only useful in certain contexts. So, you've very much got to understand the context of when to use them."

• Understanding when and how to apply patterns is crucial, as they are not universally applicable.

• The Agile Manifesto was created by a group of 17 people, including the speaker, to formalize and promote agile software development principles.
• The manifesto aimed to provide a cohesive framework for various agile methodologies, such as Extreme Programming and Scrum.

"Bob Martin being the one who was really insistent on I want to make a manifesto."

• The creation of the Agile Manifesto was driven by a desire to formalize agile principles and provide a unified framework for software development practices.

• The Agile Manifesto was initially created as a way to improve mutual understanding among its creators.
• It was surprising to the creators that the manifesto had a significant impact, although it was often misused or misinterpreted.
• The manifesto is a snapshot of the creators' thoughts and practices at a specific point in time (2001).

"And then of course the fact that it made a bit of an impact was kind of a shock. And then of course it it gets misused by by most of the time because there's there's that lovely quote from Alistister Cobin that your brilliant idea will either be ignored or misinterpreted and you don't get to choose which of the two it is."

• The Agile Manifesto's impact was unexpected, and its misinterpretation is a common occurrence, highlighting the challenges of conveying complex ideas.

• Agile practices have evolved significantly since 2000, allowing for more flexible and iterative approaches in software development.
• The traditional, lengthy planning and execution processes were replaced by shorter, more frequent cycles.
• Agile's success lies in enabling organizations to work in ways that align with their preferred methods.

"To me the great thing about agile is that we can actually go into organizations and operate it much closer to the way that we'd like to be able to do."

• Agile has facilitated a shift towards more flexible and iterative work processes, aligning better with desired methods of operation.

• AI is becoming increasingly prevalent, and its integration with Agile practices is still being explored.
• There is a potential for AI to enhance the speed and efficiency of Agile processes.
• Despite AI's potential, the focus remains on maintaining frequent, small iterations for effective learning and adaptation.

"What AI hopefully will allow us to do is to be able to do those slices faster um and maybe do a bit more in each slice, but we need to it's I'd rather get smaller, more frequent slices than more stuff in each slice."

• AI could potentially increase the speed and efficiency of Agile processes, but the emphasis remains on frequent, small iterations.

• Learning about AI involves engaging with experts and staying updated through trusted sources.
• Collaboration with experienced professionals and reading from reputable sources are key strategies for staying informed.
• It's important to critically evaluate information and understand the context and nuances behind it.

"The main way I learn these days is by working with people who are writing articles that um are going on onto my site because my primary effort these days is getting good articles onto my site."

• Engaging with experts and reputable sources is essential for staying informed about AI developments and understanding their implications.

• Junior engineers should seek mentorship from experienced professionals to guide their learning and development.
• AI tools can be useful, but it's crucial to critically evaluate their outputs and understand their limitations.
• Building a solid foundation in communication and collaboration skills is essential for long-term success.

"Find some good senior engineers who will mentor you because that's the best way that you're going to learn this stuff and a a good experienced mentor is worth their weight in gold."

• Mentorship from experienced professionals is invaluable for junior engineers, providing guidance and fostering growth.

• The tech industry continues to offer significant potential, although current economic conditions present challenges.
• AI is expected to bring transformative changes, but its long-term impact remains uncertain.
• Despite economic fluctuations, the tech industry remains a promising field with ongoing demand for skilled professionals.

"I still feel there's a hu so many huge things that can be done with technology and software. Um, and we are on, you know, we're still in a situation where demand is way more than we can imagine."

• The tech industry holds significant potential, with ongoing demand for skilled professionals, despite current economic challenges.

• Recommended reading includes "Thinking Fast and Slow" by Daniel Kahneman for insights into probabilistic reasoning.
• "The Power Broker" by Robert Caro provides an in-depth look at power dynamics and is praised for its exceptional writing.
• Engaging with well-written works enhances one's writing skills and broadens understanding of complex topics.

"A book I I do particularly like to recommend is Thinking Fast and Slow by Daniel Kaneman. I like it because um he does a really good job of trying to give you an intuition about numbers in and spotting some of the many mistakes and fallacies we make when we're thinking in terms of probability and statistics."

• Recommended readings provide valuable insights into probabilistic reasoning and power dynamics, enhancing understanding and writing skills.