More Code with Jason episodes

333 - Kevin Dias, Author of The Problem-First Method thumbnail

333 - Kevin Dias, Author of The Problem-First Method

Published 9 Nov 2026

Duration: 00:57:05

Developers should prioritize understanding customer problems over quick feature implementation to avoid long-term maintenance burdens and maintain product vision.

Episode Description

In this episode, I talk with Kevin Dias about his journey from investment banking to programming in Japan, building Ambiki for speech language patholo...

Overview

Kevin Diaz, a developer living in Tochigi, Japan, describes his path from investment banking to software. Working in the asset-backed securities group in Charlotte, North Carolina, he struggled with "what value is this bringing to society?" and left to teach English in Japan, where he first taught himself to program using Ruby on Rails. A translation tool he built for a friend led to a development role at ProZ.com, where he spent "five or six years."

As he recalls, "back in 2018," a former banking colleague who had purchased a speech language pathology practice in Tennessee asked Diaz to build software to ease therapist documentation burden. Over three years they built a bespoke system that helped the practice grow "from 70 therapists up to 180 therapists." Diaz then spun out Ambiki as a SaaS product, acknowledging the first couple of years were "really rough" as they "rewrote from scratch a brand new Rails app" to support multiple organizations.

Diaz also discusses his book, The Problem First Method, emphasizing understanding problems before jumping to solutions. He contrasts vision-driven companies with feature-accumulation approaches and cautions that AI makes building feel cheap without changing the long-term cost of ownership.

What If

  • What if you designed your data model for multi-tenancy from day one, even when building for a single client?

    • Move: Before writing code for a single-client project, sketch organizational and user-relationship concepts (e.g., can a user belong to multiple organizations?) as part of your initial data model.
    • Why Now?: Diaz's three-year bespoke EMR had to be "rewritten from scratch" because it "fit this one practice like a glove" but lacked multi-organization concepts.
    • Expected Upside: You may avoid the "at least two years of building" gap Diaz faced after spin-out, potentially shortening your path to a sellable product. (Proposal)
  • What if you ran a problem-first survey before building any requested feature?

    • Move: When a customer or teammate requests a feature, pause and run a two-step survey: first, ask your team to name the top three underlying problems; second, have leadership rank them by customer-base impact and meter movement.
    • Why Now?: Diaz found his team was "jumping to solutions and kind of skipping past some of the discovery and the problem step."
    • Expected Upside: You may discover the requested feature isn't the right fix, as when the discount code request was resolved manually for "a hundred or a 200" customers instead of a multi-week build. (Possibility)
  • What if you treated every new feature as a 15-year "puppy" adoption?

    • Move: Before committing to any feature, write down both the build cost and your best estimate of indefinite ongoing maintenance cost, then compare the total against the value it delivers to customers.
    • Why Now?: Diaz notes "it's really easy for people to think about the upfront costs of adding a feature, but they don't think about the cost that stretches into the indefinite future of owning that feature."
    • Expected Upside: You may avoid building features that consume weeks of development for minimal return, protecting your long-term velocity. (Proposal)

Takeaway

  • Before building, invest time in getting the conceptual model right: "if you get the data model correct, everything will flow very nicely after that." Spend time on naming, tables, and relationships before writing application code.

  • Use the Canon TDD loop for every feature: list your specifications, pick exactly one, write a test that describes it, let it fail, write just enough code to pass, then move to the next.

  • When a customer requests a feature, ask "why" before building. Run a short survey to identify the underlying problem, then rank problems by customer-base impact and meter movement before committing to a solution.

  • Calculate the full cost of ownership before adding any feature, not just the build cost. As Diaz puts it, solving by addition means you're "stuck with this fricking dog for the next 15 years."

  • Go physically to your customers and watch them work rather than outsourcing product design to them. "Don't let your customers design your product for you" - gain a deep understanding of their world, then design.

Recent Episodes of Code with Jason

5 Sept 2026 327 - Will Carey, CTO of Brand New Box

"Explores how software agencies are shifting from coding to defining client needs, emphasizing AI's role in development while stressing the importance of conceptual design and real-world problem-solving."

More Code with Jason episodes