Apr 22, 2025 · 34m · we-live-to-build
AI Can Write Code, But It Can't Do Architecture
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of We Live To Build, Red Hawk Technologies CEO Matt Struppelhoff explains why artificial intelligence can generate code but cannot replace human system architecture. He also details how founders can overcome business plateaus by implementing operational discipline, restructuring leadership, and transitioning away from day-to-day execution.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Sean holds 37.6% of the talking time here. How this is scored →
Sean as informed peer 3.5 Guest teaching 3.3 Guest disagreement 0.4 Sean pushing back 0.4
speaking balance: gold is Sean, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 18:08 Cautioning against relying on AI for legal answers
Matt politely refutes Sean's suggestion that AI can verify its own legal safety, noting generative inaccuracy means one must consult the written terms.
Hardest push from Sean ▶ 18:08 Challenging legal analysis needs with AI capabilitiesSean pushes back on the manual review of licensing terms by suggesting AI can simply ingest and analyze terms and conditions directly.
Biggest teaching moment ▶ 12:45 Architecture vs code generationMatt delivers an in-depth explanation of system architecture, showing why AI generating raw code cannot structure business logic layers or anticipate compliance needs.
Sean holds their own ▶ 28:36 Operational diagnosis via EOS frameworksSean demonstrates strong organizational expertise by detailing why founders must not occupy multiple seats under EOS and must limit quarterly rocks.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Sean as informed peer | Guest teaching | Guest disagreement | Sean pushing back | Why |
|---|---|---|---|---|---|---|
| Early Career Beginnings in Design and Technology | 3 | 1 | 0 | 0 | Matt recounts getting his start in 1994 building websites in Netscape Navigator, and Sean shares his parallel early experiences with HTML, C++, and school networking. | |
| The Evolution from WYSIWYG Editors to AI-Driven Development | 4 | 2 | 0 | 0 | Matt traces the evolution of WYSIWYG tools to prompt-based AI coding, while Sean provides technical color on modern tools like Figma and Lovable. | |
| Leveraging AI for Legacy Code and Documentation | 5 | 2 | 0 | 0 | Matt explains using AI to parse undocumented legacy code, while Sean relates his hands-on experience debugging database relational schemas using AI prompts. | |
| Why AI Generates Code but Cannot Architect Systems | 2 | 7 | 1 | 0 | Matt gives a deep breakdown on system architecture, emphasizing that AI creates code but lacks the foresight to separate business logic layers or evaluate scalability and compliance. | |
| Intellectual Property, Licensing Risks, and the Source of Truth | 4 | 5 | 2 | 2 | Matt explains IP leakage and copyleft licensing risks in open-source components. When Sean suggests asking AI directly, Matt gently clarifies the danger of relying on generative accuracy over explicit terms. | |
| Restructuring Leadership and Mindset to Scale Red Hawk | 1 | 3 | 0 | 0 | Matt explains how Red Hawk transitioned from a plateaued lifestyle business to an asset-building company by resolving the consultant's dilemma and bringing in executive partners. | |
| Operational Discipline and the Right Seats Under EOS | 6 | 2 | 0 | 1 | Sean demonstrates deep familiarity with the Entrepreneurial Operating System (EOS), explaining seat allocation and quarterly rocks using a peer's company as an example. | |
| Transitioning from Execution and De-risking Client Projects | 3 | 4 | 0 | 0 | Matt outlines his method of de-risking client software projects by probing technical blind spots like third-party API throttling and feasibility constraints. |
Statements from this episode (10)
Weisbrot: Non-Technical Founders Can Build Full Front/Back Ends Using Lovable
“Now I'm able to code entire front ends and back ends with natural language prompts using a tool called lovable, which I mentioned to you previously. And as a non-technical person who's not a designer, not a coder, but has visions can, I can imagine businesses,…”
Struppelhoff: Agile Often Worsens the Software Engineering Documentation Problem
“Documentation has always been a massive problem in the software engineering space. You know, agile comes in. People think that means you don't have to document things. True agile practitioners, scrum, scrum bond, they know documentation is important. They figu…”
Struppelhoff: AI Prompts Can Analyze Legacy Code and Generate Documentation
“So we can use AI prompts and actually understand the code and have it generate Comments for us to help us understand what's happening in somebody else's code. So that's really helpful.”
Struppelhoff: AI Can Generate Code, But Cannot Design Software Architecture
“It will take. Your natural language prompts varies, vary directly. It's not going to ask probing questions back to you. It's not, well, I understand what you're asking, but let's talk about the larger picture here. How frequently do you expect things to change…”
Struppelhoff: Avoid Low-Code AI Without Verifying IP Privacy Protections
“I wouldn't want to be pursuing a multimillion dollar idea using a low code AI solution without knowing if that big idea is now part of the ecosystem.”
Struppelhoff: Red Hawk Uses Google Gemini to Draft Project Backlogs
“For example, if a client gives me a what they would might consider a fairly detailed spec, I'm thinking about creating X. I'll submit that in through Gemini through our paid account so that it's all, you know, compartmentalized. It's not going out there into t…”
Struppelhoff: Red Hawk Plateaued Below $2M Top-Line Revenue for Years
“We were able to grow the business over the first five years, and then we just sort of plateaued and we stayed below a two million dollar a year, top line revenue company for a long time.”
Struppelhoff: Red Hawk on Track to Double in Size Again in 2025
“We've doubled in size each year the past three years, and we're on track to double again this year. Knock, I'm knocking on wood. I'm hoping it's going to go according to plan.”
Struppelhoff: Leaders Should Never Own Over Three Quarterly Projects
“You look at the way they handle quarterly rocks. You can't have more. You should never have more than three special projects per quarter.”
Struppelhoff: Third-Party APIs Rarely Expose Every Platform Feature
“They assume that that platform is gonna, everything you can do in that platform, you should be able to communicate back and forth for those features through an API. Oftentimes it's not true. The API can be very limited or an API might even have throttling.”