Make It Work, Make It Right, Make It Fast” in the Age of AI

The well-worn mantra “Make It Work, Make It Right, Make It Fast” has long served as a pragmatic guide for software development. Yet, in this era where Artificial Intelligence offers the seeming ability to “Make It Work” nigh-on instantly, one must ponder whether we risk short-circuiting crucial steps on the path to building something truly “Right”. Popularised on Cunningham’s Wiki, this sensible progression suggests:

1. Make It Work - first, establish technical feasibility.
2. Make It Right - then, construct it properly, with an eye to quality and maintainability.
3. Make It Fast - finally, optimise for performance once correctness and quality are assured.

Today, the initial stride—“Make It Work”—is undergoing a profound transformation, driven by Artificial Intelligence and the burgeoning practice of vibe coding.

AI & Vibe Coding: A Boon for “Making It Work”

To be clear, AI assistants and vibe coding—where developers employ natural-language prompts to conjure functional code—are remarkably effective tools for this initial exploratory phase. Need to sketch out a user interface, integrate with a basic API, or demonstrate a core user journey? AI can often deliver a rudimentary solution in a matter of hours, rather than the customary days or weeks.

This aligns neatly with the principles articulated in Ash Maurya’s Running Lean, which underscores the importance of minimising the time and cost associated with experimentation, particularly for startups and those pushing the envelope. Rather than investing weeks in crafting elaborate prototypes to test a hypothesis, teams can leverage AI to swiftly generate a Minimum Viable Product (MVP). The objective at this stage isn't polished perfection; it's about achieving speed and facilitating learning. You're essentially placing a small bet to quickly validate or invalidate an assumption with minimal outlay.

In this nascent “Make It Work” phase, expending excessive effort on anticipating every edge case, hardening security protocols, or architecting for immense scale is often counter-productive—a potential waste of precious resources if the fundamental premise proves flawed. From the outset, striving for production-grade code runs counter to the agile spirit of rapid feedback loops. AI, in this context, serves to uphold that spirit by delivering something tangible, and delivering it swiftly.

Key benefits of AI-assisted coding:

• Substantial productivity uplift: Developers employing AI assistance have reported productivity gains of up to 126% (Forbes).
• Intuitive prototyping capabilities: Techniques like voice-driven coding and natural-language prompts enable the creation of initial structures without requiring deep mastery of intricate syntax (Glide Apps, Wikipedia).
• Accelerated market validation: Shorter iteration cycles significantly reduce the time required to garner real-world feedback (McKinsey).

This dramatic acceleration resonates strongly with lean methodologies—build to learn, learn quickly, and adapt accordingly, either by pivoting or persevering based on the evidence gathered.

The Pivot Point: Why “Making It Right” Demands More Than AI

While the allure of rapid prototyping is understandable, and the initial progress can feel substantial, a critical shift in mindset is essential once that initial hypothesis shows promise. The journey towards a production-ready application—one that is reliable, secure, scalable, and, crucially, maintainable—requires more than just code generated by an AI.

“Making It Right” necessitates human intuition, critical judgement, and a qualitative understanding of the problem domain. It involves stepping back from the generated code and posing more fundamental questions:

• Did the initial experiment genuinely validate the most critical underlying assumption?
• Have we uncovered unexpected insights that hold greater value than our original objective?
• Can the initial, rapidly assembled approach fundamentally support the evolved needs of the system?

This phase demands nuanced discussions and considered decisions that current AI cannot adequately replicate. It often involves interpreting subtle user feedback, anticipating future requirements, and making architectural trade-offs informed by experience and context.

Example: Restaurant Inventory Management - Moving Beyond Basic Tracking

Make It Work (AI-assisted): One could swiftly scaffold a rudimentary system to record incoming ingredients, track usage during service, and issue alerts for low stock levels. This demonstrates the basic feasibility of inventory tracking.

Make It Right (Human Insight): Through initial trials and conversations, one discovers that restaurant owners and chefs are less concerned with simple stock counts and more interested in the interplay between inventory and profitability—menu item margins, the impact of seasonal ingredient costs, and food wastage. They require insights that connect inventory movements with menu engineering and sales patterns to inform smarter purchasing and pricing strategies.

The Pivot: “Making It Right” now entails evolving from a basic tracking tool into a comprehensive business intelligence platform, requiring:

• Robust data models that establish clear relationships between ingredients and menu items (including those used across multiple dishes).
• Integration with point-of-sale (POS) systems to capture real-time sales data.
• Sophisticated mechanisms for tracking and analysing food waste.
• Analytical interfaces focused on providing actionable insights into profitability, rather than just raw quantities.

This represents a significantly more complex undertaking than the initial AI-generated prototype, demanding careful consideration of established restaurant workflows and underlying business needs.

The Crossroads: To Stick or To Twist?

Drawing upon my experiences with startups and rescuing Minimum Viable Products, once the core concept has been validated, stakeholders invariably face a critical “stick or twist” decision:

• Stick: Empower the development team to refactor and iteratively enhance the initial prototype. Often, a significant portion—perhaps up to 80%—of the initial codebase can be salvaged and refined.
• Twist: When the learnings gleaned from the initial phase reveal fundamental shifts in scope or complexity, a more deliberate, ground-up rebuild may prove to be the more sustainable path, even if key conceptual patterns from the prototype are retained.

Making the right call at this juncture is crucial to avoid accumulating crippling technical debt and to establish a solid foundation for a truly scalable product.

Building the Bridge to “Make It Fast”

The human-guided “Make It Right” phase is what ultimately enables the “Make It Fast” stage:

1. Well-defined architectural patterns (such as Dependency Injection and modular components) create clear extension points and facilitate targeted optimisation.
2. Comprehensive automated tests, particularly regression suites, provide the necessary safety net to ensure that performance enhancements do not inadvertently introduce regressions or break existing functionality.
3. Focused profiling and tuning can then be applied to address genuine performance bottlenecks, rather than engaging in premature or misguided micro-optimisations.

Investing in building a solid, well-tested system first is the key to unlocking faster, and crucially, safer performance gains down the line.

Conclusion: From Fleeting Sparks to Enduring Systems

AI and vibe coding are undoubtedly powerful allies in the early stages of software development, accelerating the feedback loop, improving stakeholder engagement through tangible demonstrations, and facilitating lean experimentation. However, the output of this initial burst of speed is frequently an experiment, rather than a polished, production-ready artefact.

The true craftsmanship emerges in the “Make It Right” phase—where human judgement, strategic architectural thinking, and rigorous attention to quality transform those fleeting initial sparks into enduring, valuable systems. Only then can one confidently proceed to “Make It Fast” and deliver software that is not only performant but also reliable and fit for purpose in the long term.