Today’s guest is Hywel Carver!
Hywel is co-founder and CEO of Skiller Whale, which provides live team coaching for software engineering teams.
With Hywel, we talked about:
What makes traditional developer training awful, what engineers truly learn and what they should learn.
How to measure the impact of learning.
How AI may possibly change both how learning works and our own skills.
🥇 Interview Summary
Here is the 10-minute, handcrafted takeaways of what we talked about, with timestamps to the relevant video moments, for those who don’t have time to sit through the 1-hour chat.
Here is the agenda for today:
🎓 From C Programming at Nine to CTO
📚 Why Traditional Developer Training Fails
🧠 Knowledge, Skills, and Wisdom
🎯 Skiller Whale’s Approach to Effective Learning
📊 Measuring Learning Impact and AI’s Role
⠀Let’s dive in 👇
1) 🎓 From C programming at nine to CTO (02:27)
Hywel’s journey into tech began at age nine with an ambitious goal: learning C programming to create video games. When he asked a computer salesperson about programming languages for games, they wisely recommended against starting with C. But Hywel had other plans.
“This guy told me that they used a programming language called C and he was like, don’t start, don’t learn to program with C... And I thought, no, you have underestimated me. I will learn C.”
Armed with “Sam’s Teach Yourself C in 21 days,” young Hywel created his first program—a number guessing game with Easter eggs he built for himself. This early passion evolved through web development in the IE6/IE7 era (a particularly challenging time for front-end developers) to eventually becoming a founding CTO multiple times.
His academic journey included studying data science and software engineering, plus a brief stint in a PhD program focused on high-performance computing—simulating blood flow using 250,000 cores in parallel. This diverse technical background would later inform his understanding of how effective learning actually works.
2) 📚 Why traditional developer training fails (06:46)
The problem with developer training isn’t that developers don’t want to learn—it’s that traditional training doesn’t facilitate real learning.
“Software development is a craft, and people know that they could be better, that there is more stuff to learn... Training does not feel like a way of doing that to most people. Training generally feels like a box that has to be ticked for compliance.”
Hywel identifies the core issue: we’ve created a race to the bottom focused on attendance rates and engagement metrics rather than actual learning outcomes. The scalability problem forces most training providers to compromise quality:
🏫 The scaling dilemma — interactive, personalized learning doesn’t scale easily, so providers resort to recorded videos and passive content
📊 Wrong metrics — success gets measured by login rates (73% is considered good, while industry average for e-learning completion is just 13%)
🎯 Missing outcomes — training becomes disconnected from actual skill development and business results
The contrast with his Cambridge University experience was stark: there, he had access to world-class professors for lectures, but the real learning happened in small supervision sessions with PhD students who could provide personalized feedback and guidance.
3) 🧠 Knowledge, Skills, and Wisdom (17:19)
Hywel breaks learning into three distinct categories that require different approaches:
Knowledge Learning 📚
Pure information transfer—easily found through Google or books. This is what most training programs focus on, but it’s the least valuable for professional development.
Skills Learning 🎓
The ability to actually do things. You can’t learn skills by reading about them—you need practice with feedback and support. This is what developers really need to advance their careers.
Wisdom Learning 🧙♂️
Understanding when and why to apply different approaches in different contexts. Hywel points out that podcasts are actually excellent for wisdom learning:
“If I listen to multiple episodes of a podcast, and I hear one person come on and say, we structured our code base like this... and then someone else in the next episode makes the same decision in a completely different way, you start to see the pattern.”
This distinction explains why traditional training fails: it treats everything as knowledge learning when most professional growth requires skills and wisdom development. Understanding context and trade-offs can’t be achieved through passive consumption of information.
4) 🎯 Skiller Whale’s approach to learning (24:16)
Skiller Whale’s model directly addresses the scaling challenge by combining the best elements of university-style learning with modern technology:
Small Group Coaching 🍻
Live sessions with 60-90 minute focused sessions every two weeks, led by expert coaches who provide both theory and practical application.
Granular Assessment 🔬
Rather than generic “advanced Python” courses, they assess specific skills to ensure learners only spend time on what they don’t already know:
“We go very granular so that we can tell whether you know this particular bit of Python syntax, or this idea, or this bit of the standard library, so that then the learning you do is just personalized to your individual needs.”
Problem-Solving Focus 🎯
Each session includes progressively complex problems that use the concepts being taught, ensuring learners practice applying new knowledge immediately.
The approach mirrors the university supervision model: expert-designed curriculum delivered through smaller group sessions with qualified coaches who can provide personalized feedback and guidance.
5) 📊 Measuring Learning Impact and AI’s Role (40:17)
Effective learning measurement ties directly to business outcomes rather than vanity metrics:
🚀 Performance improvements — like reducing p99 response times by 10% after performance training
⚡ Velocity gains — one customer saw cycle time drop to 40% and velocity increase by 230% after API design training
🎯 Business-aligned outcomes — connecting learning directly to what organizations care about
Hywel’s partnership with DX to measure AI’s impact on developer productivity reveals nuanced findings. While there’s short-term demand for “how to use AI tools” training, the long-term implications are more interesting:
“I think we move away from needing people needing everyone to be able to write the code, and essentially everyone becoming a reviewer... We need developers to be very sensitive to badly written code, security problems, all of the things that maybe previously people would just kind of leave to review.”
Using the pilot analogy, Hywel envisions a future where developers are like modern airline pilots: they rely on sophisticated automation but maintain a deep fundamental understanding of their systems. When automated systems fail or behave unexpectedly, their expertise becomes critical.
This suggests the future of developer education will emphasize computer science fundamentals and systems thinking rather than syntax memorization—skills that complement rather than compete with AI capabilities.
🙏 Thank you
Thank you so much for reading & listening! Let me know what you think about the podcast and the brand new summaries in the comments or via email!
We just got started with the Podcast, so if you found this chat valuable, please consider subscribing to the show on YouTube, Apple, or Spotify.
Also, consider giving us a rating or leaving a review on the platform of your choice, as that really helps other listeners find the show.
See you next week! 👋
Luca
