Vibe coding: The real test of productivity

Shouts of ‘Eureka’ related to vibe coding are becoming more commonplace.

“I developed X in a few days of intensive coding. Something that would have taken me 50x the effort using ….” with Loveable, Claude, Gemini, OpenAI, or whatever other LLM that can do code generation using AI.

The claims are shrill, and they seem genuine. While it raises the hopes of those who badly wanted to build software (but couldn’t, for various reasons), the naysayers are resorting to ‘whataboutery’: security, scalability, complexity, domain expertise, and a host of other concerns about building complex solutions. Building for real-world usage, scale, and production environments is a different ball game compared to semi-working prototypes created using vibe coding.

Sounds familiar?

I’ve seen this same cycle repeat itself regularly over my 35+ years in the software industry: 4GL, CASE tools, PaaS, Low Code, No Code… and now, Vibe Coding. The hype comes first, followed by adoption, and then gradual disillusionment, until the term quietly fades from memory.

Will vibe coding go the same way? I believe that there’s something different this time.

First, it’s the quantum of dollars being pumped in by everyone. It’s massive like never before. I’ve stopped counting the billions being invested across the board.

Second, it’s the realism: no one is shouting from the rooftops about eliminating code. They are rolling their sleeves and just doing it.

Co-piloting is the name of the game. Productivity gains are happening. Job cuts are real. Something different will happen this time, it seems.

So how do we test the impact of this coding revolution? Should we look at the percentage of auto-generated code? Usage extent of copilots? Number of jobs reduced?

Vibe and copilot-style coding certainly increase output. But are costs reducing? The reduction in effort might be offset by the cost of prompt engineers, higher hosting expenses, and change management overheads.

To me, here’s the best way to evaluate productivity:

Define a complex application from the requirements perspective, in fair detail. Specify the architecture, target platform, and expectations. Ask vendors to quote in fixed-price mode. If possible, provide process models, UI mockups, reporting expectations, and specifications for complex logic. Also ask for fixed-price maintenance quotes, assuming annual incremental change. Make the price of failure steep so vendors won’t overpromise and underdeliver.

Vendor prices come down when:

• They are confident that their vibe coding practices will reduce effort

• They have strong accelerators (low code, no code, internal tools)

• The scope is tightly defined, minimizing creep

• They have a frictionless delivery and quality system with proven methods

• They have the right skills

• They have right shored those skills to access optimal talent at optimal cost

• The cost of underlying software and infrastructure is optimized

The true productivity gains will show in bids over lifecycle costs, as well as in the timeframes committed. Those who source custom applications are often shocked at the wide variance in price quotes. When needs are well-defined, the range narrows, but still varies significantly.

Here’s a rough guide to how different providers may quote for a tightly specified, complex project:

• Big 4 provider: used to quote 6X; now may quote 5X with AI acceleration

• Large system integrators: 4X

• Proven offshore providers (Eastern Europe, South Asia, Far East): 3X

• Low code platform providers: 2X (assuming their platform is used)

• A dark horse provider, using low code + AI vibe coding, right-shored teams, and a solid track record: may quote as low as 1X, and still deliver well

  
  

In the past, the dark horse provider would be written off due to risk. That risk is now greatly reduced, given AI’s ability to solve complex problems and accelerate delivery. That dark horse can now be seriously considered.

And when 5X budget reductions become real, a lot becomes possible:

• Projects once deemed too expensive or risky can now be delivered faster

• Legacy applications can be rebuilt or refactored from scratch

• More time can be invested in quality and excellence

• You can finally build that product you couldn’t build before

• Businesses gain new freedom to be more creative and ambitious

  
  

It’s not necessary that every innovation include AI and agents. Innovation can be about using AI smartly — to deliver 10X results at the same cost. Yes, when applied to the right places, AI can dramatically help.

At Stragiliti, we believe we are that dark horse.

We leverage our proven low-code platform, AI co-piloting, time-tested delivery methods, and focus on excellence to build complex applications and products, rapidly and affordably. We’ve been doing this for 15+ years. With AI and low code, we’re accelerating even faster.

And with that massive productivity boost, we’re also building more products. All the products are a suite of SaaS offerings focused on professional services.

If you’re looking to build something complex that you think wasn’t feasible earlier from the time, risk or cost perspective, or you’re a CXO of a professional services firm evaluating product-based solutions, visit www.stragiliti.net.

You might just find the dark horse you need, without the risk.