When LLMs Act as Translators

What happens when coding is no longer the bottleneck in software development?

In this episode, Alberto Brandolini - creator of EventStorming and pioneer in domain-driven design - joins software engineer and Kandddinsky founder Marco Heimeshoff to explore how AI is transforming the practice of building software, and what remains fundamentally human in the process.

Together, they reflect on the growing importance of collaborative modelling, domain language, organisational coherence, and feedback loops in a world where software can increasingly be generated through interaction rather than deterministic programming.

This episode offers a grounded yet provocative perspective on what it means to be human, in an increasingly agentic world. Tune in.

Alberto and Marco also speak about how AI is reshaping their day-to-day development practices - from using Claude Code and Obsidian-based memory systems to designing “harnesses” that constrain and guide increasingly capable agents.

The conversation explores the rise of transient software, the limits of “vibe coding,” and why bounded contexts, modular architectures, and shared language become essential when working with probabilistic systems.

Together, they offer a practical glimpse into how software engineering is evolving from writing deterministic code toward orchestrating learning, context, and collaboration between humans and AI systems.

Episode co-hosted by Eugenio Battaglia.

Key Highlights
👉 Coding is no longer the primary bottleneck in software development; the real challenge is shaping context, boundaries, and shared understanding for AI systems.
👉 Collaborative modelling becomes even more important in an AI-native world, because humans still need to align on purpose, trade-offs, and organisational intent.
👉 “Harness engineering” is emerging as a new discipline focused on constraining, guiding, and coordinating AI systems through workflows, memory, tests, and domain context.
👉 Large language models can accelerate software production dramatically, but ambiguity in language and organisational misalignment still create major risks.
👉 Faster feedback loops may expose organisational incoherence more quickly, forcing companies to confront outdated structures, unclear responsibilities, and low-value work.
👉 Human conversations, organisational politics, and qualitative understanding remain irreplaceable because people rarely know — or communicate — exactly what they truly need.
👉 The rise of “vibe coding” may increase speed in the short term, but without deep understanding and modular boundaries, systems can quickly become fragile and unmanageable.

Topics /chapters

(00:00) What happens when Coding Stops Being the Bottleneck - INTRO

(01:31) Introducing Alberto Brandolini and Marco Heimeshoff

(04:18) The AGI Debate and the Coding Shift: Early Observations from the Frontier

(10:25) How do we reimagine modeling?

(17:19) The Real Shift in AI Work

(28:54) AI - From Modeling to Co-Creation

(37:31) From Human Alignment to Agent Alignment

(46:54) Mapping, Ontologies, and the Limits of Controlling AI

(56:07) What’s Next?

Remember that you can always find transcripts and key highlights of the episode on our website: https://www.boundaryless.io/podcast/alberto-marco

Episode recorded on Apr 20, 26

Find out more about the show and the research at Boundaryless at https://boundaryless.io/resources/podcast/

Get in touch with Boundaryless:

• Twitter: https://twitter.com/boundaryless_
• Website: https://boundaryless.io/contacts
• LinkedIn: https://www.linkedin.com/company/boundaryless-pdt-3eo

Music

Music from Liosound / Walter Mobilio. Find his portfolio here: https://blss.io/Podcast-Music