I did a podcast! I recently joined my JUXT colleagues Malcolm Sparks, Joe Littlejohn and Denis Lobanov on JUXTCast to discuss a project I’ve been working on over the past several months: the JUXT AI Radar.

As leader of JUXT’s AI chapter, I’ve looked to find ways to help our team understand the rapidly evolving AI ecosystem, and now we’re sharing that knowledge with the wider tech community. As with the Thoughtworks Radar which provided the original inspiration, the AI Radar maps frameworks, tools & techniques, languages, and platforms, providing structured insights and recommendations for navigating this complex landscape.

The conversation was an opportunity to cut through the hype and explore some more nuanced perspectives on AI. We went off on lots of tangents, including my thoughts on why the results of studies into the impact of AI on software engineering are highly inconsistent, and why I’m watching interface developments in the Agentic IDE space closely. At the time of recording, Amazon’s Kiro IDE has just been launched.

I also shared my view that while LLMs represent a genuine paradigm shift for natural language processing, classical machine learning still offers some major advantages that shouldn’t be overlooked. However, if you do need to fine-tune or self-host LLMs, I strongly recommended taking a look at Mistral’s offerings because their models offer an economical balance of power and runtime efficiency.

Hopefully this episode captures the thoughtful spirit that drives our approach to understanding AI at JUXT. We’re committed to providing collaborative, evidence-based insights that help both our team and the broader tech community make informed decisions.

You can listen to the full episode on your preferred podcast platform or watch it on YouTube. I hope you enjoy the discussion as much as I did.