Keri Olson, VP of AI for Code at IBM and an expert in coding assistants, shares fascinating insights on enhancing developer productivity. She discusses the maturity of AI technologies that augment developers, highlighting their impact throughout the software development lifecycle. The conversation delves into the evolving role of coding assistants and the challenges developers face with existing tools. Keri also emphasizes the importance of best practices and transparency in AI, along with the exciting potential of purpose-built coding assistants to transform the development process.
32:12
forum Ask episode
web_stories AI Snips
view_agenda Chapters
auto_awesome Transcript
info_circle Episode notes
volunteer_activism ADVICE
AI Code Assistants as Tools
Think of AI code assistants as assistants or pair programmers, not replacements for developers.
Developers remain responsible for development work, leveraging AI for support.
question_answer ANECDOTE
Code Assistant Improves Documentation
A team at IBM used a code assistant to document a JavaScript project.
This reduced documentation time by ~90%, from three minutes to 12 seconds per file.
volunteer_activism ADVICE
Purpose-Built AI Solutions
Consider purpose-built AI solutions for specific coding use cases like IT automation or application modernization.
IBM offers tailored solutions for Ansible playbook creation and mainframe modernization.
Get the Snipd Podcast app to discover more snips from this episode
Keri Olson (@ksolson20, VP AI for Code at @IBM) talks about coding assistants across the software development lifecycle, the future of agents, and domain-specific assistants.
Topic 1 - Welcome to the show. Tell us about your background, and then give us a little bit of background on where you focus your time at IBM these days?
Topic 2 - Developer code assistants have become one of the most popular areas of GenAI usage. At a high level, how mature are the technologies that augment developers today?
Topic 3 - Software development has an entire lifecycle (Generate, Complete, Explain, Test, Transform, Document). It’s easy for developers to just plug in a service, but is that often the most effective way to start using GenAI in the software development lifecycle?
Topic 4 - Software developers are notoriously picky about what tools they use and how they use them. GenAI doesn’t “guarantee” outputs. Are there concerns that if different developers or groups use different coding assistants, that it could create more challenges than it helps?
Topic 5 - What is a holistic way to think about code assistants? How much should be actively engaged with developers, how much should be behind the scenes, how much will be automated or agentic in the future?
Topic 6 - In the past, we essentially had “real developers” (people who wrote code) and things like Low-Code for “citizen developers” on process tasks. Do you expect to see code assistants bringing more powerful skills to people that previously hadn’t identified as a real developer? (e.g. the great idea on a napkin that turns into a mobile app)
The Cloudcast