Retentions in Construction Contracts

3D Printed Concrete Houses

In this episode of The Architectural Podcast, guest Kirill from AMCreate explains how his company uses an on-site gantry system to 3D-print concrete houses in New Zealand, describing the process, equipment flexibility, and why printing in place reduces transport, crane, labor, and erection costs compared to factory precast approaches. He outlines typical material volumes (20–30 tons for a one-story house), weather limitations similar to masonry, and the company’s specialized mortar-like mix with additives and fiberglass for strength and shrinkage control. Kirill details seismic reinforcement using vertical and horizontal rebar with self-compacting concrete infill, waterproofing via thick concrete walls plus paint, and insulation using internal cavities filled with perlite to meet H1 requirements. The discussion also covers recyclability and life-cycle carbon claims, durability and comfort benefits (thermal mass, acoustics, fire and flood resilience), market adoption challenges, and the design freedom of curved, cost-neutral geometries in printed walls. Links: https://amcrete.co.nz/

Retentions in Construction Contracts

In this Architectural Technology podcast episode, Jonathan and Ethan explain retention money as a percentage withheld (often around 5%, with examples up to 10%) to ensure subcontractors and contractors return to remedy defects, with retentions commonly released 50% at practical completion and the remainder after the defects liability period. They outline how New Zealand’s Construction Contracts Act 2002 was introduced after a major company withheld retentions without setting them aside, then liquidated, leaving subcontractors as unsecured creditors; the Act now requires retentions to be held in trust, supports industry cashflow, outlaws “pay if paid/pay when paid,” and imposes payment timeframes and penalties. They discuss administration and reporting obligations on main contractors, contrast retentions with bonds, and clarify that the Act applies where none of the parties is a “residential occupier,” affecting some domestic-building scenarios, and they invite listeners to compare approaches in other countries such as England’s JCT contracts. Ethans's Youtube: https://youtu.be/fgsWdyekHBI?si=MGwf10l4Q4WGxBp7

Mastery Versus Proficiency

Meeting discussion clarified software proficiency timelines and distinguished technical mastery from foundational workplace competence.Defining Mastery Versus ProficiencyParticipants debated the distinction between software proficiency and mastery. Mastery requires managing complex functions, while proficiency involves understanding basics.Estimating Software Learning TimelinesFull-time study for 1 year is estimated for software mastery. Proficiency can be achieved within 1 week of focused effort.Contextualizing Professional Skill AcquisitionIndustry standards suggest long-term development is necessary for true expertise. The team decided that proficiency is the primary goal for employment readiness.

Comparing AI Model Capabilities

Participants evaluated various AI models for text and design workflows while debating architectural curriculum requirements.Comparing AI Model CapabilitiesDeepseek and Gemini offer distinct advantages for text and image tasks, necessitating a multi-tool approach for professional workflows. Participants decided to use different models for specific purposes, such as Deepseek for technical writing and Gemini for imagery.AI in Architectural RenderingAI tools effectively accelerate rendering and design, yet traditional software remains essential for high-quality, large-scale output. Professionals must provide specific prompts and references to achieve desired outcomes.Curriculum Needs for TechnologistsArchitectural education should integrate both coding and prompting to ensure graduates understand structural problem-solving and industry-specific app development. Mastery of code is critical for correcting errors in AI-generated scripts.

What Is The Best AI Platform

This episode of the Architectural Technologist podcast features a conversation between Jonathan and Ethan regarding their experiences with various AI tools, specifically comparing Google Gemini and DeepSeek. The following is a summary of the key takeaways from their discussion: * Technical Writing & Analysis: Jonathan found DeepSeek to be a superior tool for technical writing, noting its ability to analyze text for errors and provide multiple rewriting options. * Voice and Tone: DeepSeek offers unique "voices" for rewriting, such as "Kiwi Practical" or "Professional," which Jonathan found helpful for matching his natural speaking style. * Multimodal Capabilities: Gemini stands out for its ability to generate and read images, listen to music, and process audio files—features that DeepSeek currently lacks. * Privacy and Business Use: Jonathan prefers Gemini for business because it does not share user information or use it to train its large language models. * Open Source and Efficiency: Ethan highlighted that DeepSeek is an open-source, energy-efficient "thinking AI" that is currently free to use. * Newsletter Creation: Jonathan used both tools to organize course transcripts into newsletters. While Gemini suggested splitting the content, DeepSeek recommended keeping it as one cohesive topic, which Jonathan found more accurate for his needs. * Architectural Rendering: Ethan shared how he teaches students to use Gemini to render 3D models. By providing an address, the AI can use Google Maps data to create realistic day and night renders of buildings. * High-Fidelity Media: Jonathan described an architect friend using advanced Gemini models (referred to as Nano Banana) to generate incredibly detailed videos, such as a dog walking through a kitchen with realistic lighting and textures. The speakers concluded that there is no single "perfect" AI. Instead, they recommend a multi-tool approach: using DeepSeek for technical drafting and logic, while leveraging Gemini for creative tasks like image generation and audio analysis. DeepSeek vs. Google GeminiPractical AI ApplicationsFinal Conclusion