It has been a busy month for the EfficientEther research and editorial work. Our founder Ryan Mangan has published two articles with Computer Weekly and contributed to a podcast on the challenges AI faces with hallucinations, the role of green coding in modern software, and how emerging standards like ISO 42001 keep the discipline honest.
Enhancing AI precision
Advances in large language models have pushed AI's capabilities and ethical applications forward in parallel. Ryan's first Computer Weekly piece looks at how new standards, including ISO 42001 for AI management systems, are giving teams a framework to deploy AI responsibly. The piece also examines failure modes such as "model autophagy disorder," a feedback loop in which models trained on AI-generated outputs degrade in quality over time, and the operational practices needed to avoid it.
Read the full article on Computer Weekly: Advancing LLM precision and reliability.
Green coding practices
The second article looks at the environmental side of software development. Choices made at the language and runtime level have measurable effects on energy use, and over the lifetime of a service those effects compound. Languages like C and Rust tend to be more energy-efficient than Python in many workloads, and the piece walks through where the differences come from and where the trade-offs lie.
Read the full article on Computer Weekly: Green coding: the role of energy efficiency in development.
Addressing hallucinations in AI
Hallucinations remain one of the harder operational problems in AI deployment. The podcast discusses how techniques like Retrieval-Augmented Generation help anchor model outputs in verifiable data, and what is still missing for high-stakes domains like healthcare and finance.
Listen on YouTube: AI hallucinations podcast.
Why this matters
The thread running through all three pieces is the same one underneath the EfficientEther product portfolio. AI is most valuable when it is grounded in the realities of the systems it operates in, energy-aware in how it runs, and built against frameworks that keep its behaviour predictable. That is the foundation EtherAssist is built on, and it is the lens we apply to every new product on the platform.
Further reading
If you want to see how this work lands in our products, sign up for EtherAssist at app.etherassist.ai or register your interest in the wider EfficientEther Platform.