RACE Programming
Framework · Origin

First Line Software: the pioneer

RACE Programming did not originate in an academic paper, a conference keynote, or a consultancy framework. It was developed inside a software engineering organization that needed to answer a practical question: how do you restructure a delivery business around AI systems that are becoming genuinely capable of production engineering work?

That organization was First Line Software AI Lab, the R&D center and methodology think tank of First Line Software — a global software engineering company with delivery teams across multiple continents and decades of production engineering practice behind it. The AI Lab was given the mandate to develop both the methodology and the tooling required to operate AI-native at scale, not as a research exercise but as the delivery model for the parent company's future.

What "pioneer" means here

The word is used precisely. RACE Programming was not adapted from an existing methodology. It was built from the axioms up — informed by practice at First Line Software, validated against production outcomes across multiple projects, and codified only after the underlying patterns stabilized.

Three specific pioneer contributions:

Organizational

First Line Software built the first production role model of the three-tier team structure. The Team Principal / Pit Wall / Pit Crew model — with the ACE Technical Lead and AI Product roles forming the Pit Wall pair — was staffed, trained, evaluated, and iterated against real client outcomes before the methodology was named. The transition patterns between traditional Scrum teams and RACE Programming teams, the criteria for identifying engineers who can operate as ACE Technical Leads, the coaching process for AI Product role — all of this was developed in practice inside First Line Software.

Process

The Stint as a one-week iteration ending in a Pit Stop (production deployment), the specific minimal ceremony set, the economic model benchmarking RACE Programming against Scrum at equivalent headcount — these were not theoretical. They were arrived at through a documented transformation program inside First Line Software's delivery organization, with before/after measurement against prior Scrum baselines. The 4× throughput figure is not projected; it is measured: 800 SP/month (classic Scrum) → 3,200 SP/month (Race Programming with two Pit Crews) at the same 7–9 person headcount.

Technological

The Executable User Story format — including the decision to combine traditional user story structure with embedded Gherkin acceptance criteria consumable by AI coding systems — was developed, tested, and standardized at First Line Software. The upstream tooling to process, version, and regenerate downstream artifacts from Executable User Stories was built to support the methodology. RACE Programming's extension of industry-standard Everything as Code to place Executable User Stories as the originating artifact is a First Line Software contribution to the broader field.

Why this matters for teams considering the methodology

Many methodologies are formulated first and implemented later. The gap between formulation and implementation is where most methodological promises fail. RACE Programming was implemented first and formalized second — which means every prescription in the methodology was chosen because it survived contact with production delivery, not because it fit a conceptual schema.

For organizations considering adoption, this has practical consequences:

  • Benchmarks are real. The throughput numbers, the role compression ratios, the cost model — these come from delivered projects, not pitch decks.
  • Tooling exists. The technological stack that supports RACE Programming is not hypothetical. It runs in production at First Line Software and supports client engagements today.
  • Trained practitioners exist. ACE Technical Leads and AI Product Managers trained inside First Line Software are the current bench of accredited practitioners. There is no shortage of the methodology; there is a finite supply of practitioners at any given time.
  • Failure modes are documented. The patterns that do not work — attempts to adopt RACE Programming piecewise, misaligned incentive structures, insufficient context discipline — are documented because First Line Software encountered them first.

The relationship between RACE Programming and First Line Software

RACE Programming is an open methodology. This site exists precisely to make it accessible to the broader engineering industry — to document the axioms, theorems, practices, and artifacts so that any organization can understand, critique, and adopt it.

First Line Software's role is implementer, not owner. The methodology is published, critiqued, and evolved in the open. The tooling to run it at production grade is available to the industry through First Line Software engagements until equivalent tooling emerges elsewhere. Long term, RACE Programming is intended to be a standard of practice in AI-native delivery, not a product.

For teams that want to implement RACE Programming themselves, this site is the specification. For teams that want help from the organization that built it, First Line Software's AI-Native Development practice is the direct path.

RACE Programming is authored by Pavel Khodalev, CTO of First Line Software. The methodology was developed with contributions from the First Line Software engineering leadership team and informed by production delivery outcomes across multiple client engagements between 2024 and 2026.