DEV Community

Cover image for What Our Conclusions Leave Out
Stjepan Jureković
Stjepan Jureković

Posted on AI-assisted

What Our Conclusions Leave Out

Reading The Architecture of Reasoning and rethinking how we examine assumptions, explore alternatives, and test technical decisions.

After seven chapters of Rick Giner’s The Architecture of Reasoning, I’m left with a question: what would make me challenge an explanation that already sounds right?

The book draws on AI reasoning patterns to examine human decision-making. What interests me is the gap between explaining an approach clearly and establishing that it’s the right approach.

I’m taking away four checks: look for the hidden assumption, consider a different path, change perspectives, and test the explanation.

Caption: Five reasoning architectures, each adding a different capability. The aim is to use the structure the problem needs—not always climb to the top.

Credit: From The Architecture of Reasoning by Rick Giner, Manning Publications, MEAP, Figure 7.1.

Where did we leap?

Chapter three opens with a payments service timing out under peak traffic. The symptoms suggest database trouble. The team increases the connection pool and scales the database. The service appears to stabilize.

But which observations support the diagnosis? What assumptions connect the symptoms to the proposed cause?

The chapter’s treatment of chain-of-thought is about making those intermediate steps visible. My practical takeaway: before accepting an explanation, look for the step that everyone seems to have skipped.

Are these actually different options?

Chapter four considers a struggling service. Should the team rewrite it or refactor it?

That sounds like an open question, but it has already narrowed the possibilities.

This helped me separate chain-of-thought from tree-of-thought: examining one line of reasoning isn’t the same as exploring different paths. Developing an option in greater detail doesn’t replace considering alternatives.

The question I want to keep nearby: are we comparing genuinely different approaches, or variations of the same one?

Whose definition of “better”?

Chair-of-thought adds another check. A team can consider several options while judging all of them through the same dominant perspective.

Changing the lens brings different constraints, risks, and success criteria into view. The point isn’t simply to let each specialist defend their position. It’s to examine the decision from perspectives that might otherwise be missing.

For me, the useful question is: whose definition of “better” is shaping this comparison?

What would change my mind?

The ReAct chapter takes reasoning beyond the explanation itself. Act, observe, and update. Give evidence a chance to challenge a conclusion before it becomes a commitment.

Caption: Testing before commitment gives evidence a chance to change the decision.

Credit: From The Architecture of Reasoning by Rick Giner, Manning Publications, MEAP, Figure 6.1.

This is the habit I most want to try. Before acting on a technical recommendation, write down the assumption that matters most—and what observation would make me reconsider it.

I haven’t tested that habit yet. But it gives me something concrete to do with the reading, beyond agreeing with it.

Enough structure, not all the structure

By chapter seven, I was wondering how much process this would add to an ordinary decision.

The answer is reassuring: the reasoning stack isn’t a mandatory sequence. Use enough structure for the problem, and return to simpler reasoning when that’s sufficient. People also need room to question and revise; a framework alone won’t create that.

I don’t want to make every answer longer. I want to notice when something important is missing: an assumption, an alternative, a perspective, or a test.

When a technical explanation sounds convincing, what makes you look closer?

I work at Manning Publications. These are my personal reading notes.

I’m reading The Architecture of Reasoning by Rick Giner: https://www.manning.com/books/the-architecture-of-reasoning

Top comments (0)