The podcast Invest with the Best, Scott Wu discusses his company Cognition and AI.
2024-12-17
https://open.spotify.com/episode/3hCL9ZcSmM8pzEPPO3UnVB
My guest today is Scott Wu. Scott is the co-founder and CEO of Cognition, which is an applied AI lab that has created the first AI software engineer, which they call Devin.
In just a year since founding Cognition, Devin functions at the level of a junior software engineer, capable of handling complete engineering workflows from bug fixing to submitting pull requests. He is a former competitive programming champion and describes the field as simply “the art of telling the computer what you want it to do." Scott predicts AI will surpass the world's best competitive programmer within 1-2 years and sees this technology not as replacing programmers, but as democratizing software creation.
We discuss the bottleneck in software development, the future of AI in various industries, and the challenges of leveling up Devin. Towards the end, you’ll also hear Scott do an insane card trick on me. You can find the video on our X and to grasp the madness fully. Please enjoy my conversation with Scott Wu.
###
Chat GPT 4o (prose)
Scott Wu, co-founder and CEO of Cognition, is reshaping the landscape of software engineering with Devin, the world’s first fully autonomous AI software engineer. Within just one year, Cognition has elevated Devin’s capabilities from those of a high school-level coder to the sophistication of a junior engineer. Embedded seamlessly into tools like Slack, GitHub, and other development environments, Devin manages entire engineering workflows—including bug reproduction, code debugging, unit testing, and pull request generation—offering teams a transformative, asynchronous approach to software development. Designed to complement human engineers, Devin not only automates repetitive tasks but also multiplies productivity, allowing engineering teams to tackle complex problems while Devin handles the heavy lifting. Organizations using Devin report dramatic efficiency gains, with every hour of Devin’s work equivalent to 8–12 hours of human engineering time.
Devin’s impact is particularly evident in industries with vast legacy systems, such as banking and healthcare, where decades-old codebases often impede innovation. Devin’s ability to automate modernization efforts and version upgrades has provided a clear path to greater efficiency, security, and scalability. Wu sees this as a harbinger of the future: a democratization of software creation, where even individuals without coding expertise can use AI to translate natural language instructions into functional, efficient code. He predicts that within the next five to ten years, AI will evolve to the point where anyone can articulate their desired software solutions in plain language and receive polished, high-performance outputs—opening the door to previously unimaginable levels of customization and personalization.
Beyond its technical contributions, Cognition’s work raises profound questions about the economics and future of software development. Wu observes that software creation has long been constrained by supply, with a chronic shortage of engineers limiting what organizations can achieve. With AI agents like Devin dramatically reducing the cost and time required for software development, the industry will likely shift focus to demand generation, distribution strategies, and new business models. Wu notes that traditional software companies reliant on high switching costs may face disruption, as AI-driven automation significantly lowers barriers to migration and implementation. Conversely, businesses that leverage personalization, network effects, and data-driven insights stand to thrive in this new environment, where efficiency and adaptability will be paramount.
Wu also reflects on the philosophical implications of this transformation, particularly in relation to Artificial General Intelligence (AGI). While he remains cautious about declaring AGI a present reality, he highlights how AI is already closing gaps in human capability, task by task, through iterative improvement. This recursive trajectory mirrors the evolution of software engineering itself, where each innovation—from assembly languages to the cloud—has abstracted complexity, enabling engineers to focus on higher-order challenges. Devin exemplifies this progression, serving as both a tool and collaborator that elevates the creative potential of human engineers.
Cognition’s approach to AI agents further distinguishes it from the competition. Unlike simple code assistants that autocomplete text, Devin operates as a dynamic agent capable of making autonomous decisions, executing complex workflows, and adapting to specific organizational contexts over time. Wu emphasizes that Devin learns and improves as it integrates into a company’s ecosystem, gaining a level of familiarity with the codebase that rivals even long-tenured engineers. This deep contextual understanding, paired with Devin’s ability to interact asynchronously, enables developers to offload not just routine maintenance but also the iterative problem-solving tasks that typically consume the majority of their time.
Scott Wu’s vision extends beyond the immediate impact of Devin to the broader potential of AI in society. By enabling software to be built faster, cheaper, and more effectively, Cognition aims to unlock an unprecedented wave of innovation. Wu compares this to past technological shifts, such as the advent of mobile computing, but with the added complexity of continuous, incremental advances in AI models and capabilities. Far from being a threat, Wu views AI as a catalyst for greater human creativity, enabling individuals and organizations to build solutions to problems that were once out of reach. Cognition, with its trailblazing work on Devin, stands at the forefront of this transformation, offering a glimpse into a future where the boundaries between human ingenuity and machine efficiency blur to redefine what’s possible.
###
Chat GPT (bullets)
The podcast episode features Scott Wu, co-founder and CEO of Cognition, an applied AI lab that has developed "Devin," the first fully autonomous AI software engineer. Wu describes Devin as capable of handling entire engineering workflows, such as debugging, unit testing, and submitting pull requests. Positioned at the level of a junior engineer, Devin represents a breakthrough in software automation, offering efficiency gains equivalent to eight to twelve hours of human engineering work for every hour it operates.
Key Highlights:
-
Cognition's Focus and Devin's Capabilities: Cognition aims to automate software engineering through Devin, which operates seamlessly within tools like Slack and GitHub. Devin's core strength lies in its asynchronous workflow, enabling teams to manage multiple tasks simultaneously. This helps engineers focus on creative problem-solving while delegating routine tasks.
-
Devin's Evolution and Usage:
- Devin has progressed from performing at the level of a high school computer science student to functioning as a junior engineer in just a year.
- Its primary applications include code modernization, version upgrades, and replatforming for industries like banking and healthcare. For example, Devin excels in updating large, decades-old codebases efficiently.
-
AI’s Role in Democratizing Software Development:
- Wu envisions AI reducing barriers to entry in programming by translating natural language instructions into functional, efficient code. This could democratize software creation, making it accessible to non-programmers.
- He predicts a paradigm shift where AI outperforms top programmers, akin to AlphaGo's landmark achievement in AI.
-
Economic and Business Impacts:
- The widespread adoption of AI agents like Devin could significantly reduce the cost of software development and shift the industry's focus from supply constraints to demand and distribution challenges.
- Companies relying on high switching costs may face disruption as AI agents lower implementation and migration costs.
-
Future of AI and Agents:
- Wu discusses the growing importance of AI agents that act and make decisions in real-world contexts, such as handling multi-step tasks autonomously.
- He highlights opportunities for businesses to prepare their products for "agent customers" alongside human users, emphasizing the integration of agents into collaborative tools like Slack and GitHub.
-
Philosophical and Practical Considerations:
- Wu contemplates whether AI advancements have reached Artificial General Intelligence (AGI), noting that the definition and scope of AGI remain fluid.
- He stresses that while AI will tackle many routine tasks, humans will continue to play a critical role in defining problems and designing solutions.
This discussion offers a glimpse into the transformative potential of AI in software engineering and beyond, with implications for productivity, creativity, and industry-wide disruption.
