# bartaguiz-velocite-harness-engineering-2026-10-15

## Veille

Talk deck by **Dorra Bartaguiz** (CTO, **Arolla**), presented at the **Dev With AI** conference and published as a PDF on LinkedIn on **October 15, 2026**; **61 slides**, in French with examples in English. Thesis: when AI causes velocity to surge, quality is regained through tooled *craft* practices built around the model — *harness engineering*.

## Titre Article

Quand la vélocité explose avec l'IA, place au "Harness Engineering" !

## Date

2026-10-15

## URL

https://lnkd.in/p/eccdGyTZ

## Keywords

harness engineering, velocity, code review, bottleneck, AI-generated tests, code coverage, mutation testing, mutation tests, code smells, craft practices, software craftsmanship, shift-left, SDLC, PDLC, Example Mapping, ATDD, TDD, Given/When/Then, DDD, hexagonal architecture, skills, Superpowers, specialized agents, loop engineering, software factory, human supervision, agentic organization

## Authors

Dorra Bartaguiz — CTO d'Arolla (conseil et formation, « agentic software crafters ») ; slides de son talk Dev With AI publiées sur LinkedIn.

## Ton

Profile: a conference deck for developers, in an upbeat, demonstrative tone ("Haha!", "Woohoo!", "Generate me this requirement!"). The progression is narrative: the promise of velocity, a reminder that review remains the bottleneck, a demonstration through a real case where the generated code and tests are measured and then taken apart, a definition of the harness, a step-by-step revisiting of the same case with craft practices, then a broadening from the development cycle to the product cycle and the organization. The argument rests on cited sources (DORA, Clean Code, Minelli et al., Tornhill and Borg, Boehm, McKinsey, Kent Beck) and on measurements the speaker took on her own example (coverage, mutation score). The slides are often just titles or diagrams, without spoken commentary; several carry the label "Context Engineering Training". The framing is that of a craft trainer: teams' heuristics and workflows are presented as *"tomorrow's skills"*.

## Pense-betes

- **Finding**: generation accelerates syntax but increases the burden of review and verification; the codebase remains the bottleneck.
- **Screening case**: invitation date = day of birth + current year + month of birth + N months depending on the screening type, then rescheduling (invalid date, past date, public holiday, weekend).
- **Speaker's measurements**: reported coverage 100% / actual 33% (domain 95%); mutation testing 13.82%.
- **Smells identified**: hardcoded configuration, static methods, SRP violation, not thread-safe, silent failures, magic values, generic naming, tight coupling, mutability.
- **Harness**: everything added around the model; proposed lever for determinism: craft + shift-left.
- **Craft SDLC in five blocks**: Exploration → Decomposition (epics, features, user stories) → Formulation (examples, scenarios) → Tests (TDD, ATDD, automation) → Development and refactoring.
- **Baby steps**: *"on their own, they are not enough"*; the harness must be worked on, hence skills.
- **Skills frameworks cited**: Superpowers, Matt Pocock's skills, Impatient Programming, **Skraft** (AXA, talk by Cyrille and Olivier at Devoxx 2026).
- **New model version**: *"Your old skills are already built in"*.
- **PDLC**: product strategy, architecture and security by design, UX/UI, release engineering, observability, maintenance, product impact; equipping only the dev team creates a bottleneck.
- Related to [[sfeir-code-review-anneau-contraintes-2026-07-30]] and [[lassiege-usine-logicielle-heure-ia-2026-07-28]].

## RésuméDe400mots

Dorra Bartaguiz starts from AI's promise: more than 80% of developers believe it has increased their individual productivity, according to the 2024 and 2025 DORA reports. But speed means risk. Development time has shrunk, and the bottleneck remains code review. She notes that reading vastly outweighs writing: a ratio of more than ten to one according to Robert C. Martin, with about 70% of time spent understanding code versus 5% editing it, according to a 2015 study. AI accelerates syntactic generation and increases the review burden.

To find out whether generated tests can be trusted, she takes a real case: calculating the invitation date for a medical screening from the screening type, date of birth, and current date, with rescheduling rules for invalid, past, public-holiday, or weekend dates. Generated from a single prompt, the code is reported as 100% covered. She measures 33% overall coverage, 95% on the domain, flags a series of defects (hardcoded configuration, static methods, violation of the single responsibility principle, silent failures, magic values, tight coupling), and obtains a mutation testing score of 13.82%. Tests generated without rigor validate the AI's errors.

She then defines the harness as everything added around the model to improve generation; an agent is a model plus a harness. To make the agent more deterministic, she proposes craft practices and shift-left. The same case is revisited step by step: asking for domain definitions, breaking it down into rules, identifying examples and edge cases, writing Given/When/Then scenarios, generating acceptance and unit tests, then requesting a refactoring toward clean code, hexagonal architecture, and DDD tactical patterns. With AI, TDD becomes a practice for controlling an unpredictable agent; she cites Kent Beck, for whom teams that know how to test will adapt better.

Baby steps are not enough: the harness must be worked on, particularly through skills. Popular frameworks, from Superpowers to Skraft at AXA, rely on these practices. Finally, a product's life does not stop at the development cycle: equipping only the development team creates a new bottleneck. She proposes instrumenting the entire product cycle with specialized agents and guardrails against infinite loops, and sums up software factories as the combination of loop engineering, harness engineering, and human supervision.

## GrapheDeConnaissance

- Dorra Bartaguiz —publie→ talk Quand la vélocité explose avec l'IA (DOCUMENT, 0.98)
- Dorra Bartaguiz —travaille_chez→ Arolla (ORGANISATION, 0.85)
- talk Quand la vélocité explose avec l'IA —observé_dans→ Dev With AI (EVENEMENT, 0.95)
- talk Quand la vélocité explose avec l'IA —référence→ DORA (ORGANISATION, 0.9)
- Dorra Bartaguiz —affirme_que→ la revue de code reste le goulot d'étranglement à l'ère de l'IA (AFFIRMATION, 0.95)
- Dorra Bartaguiz —mesure→ couverture annoncée 100 %, mesurée 33 % (95 % sur le domaine) sur le cas dépistage (MESURE, 0.95)
- Dorra Bartaguiz —mesure→ score de mutation testing de 13,82 % des tests générés (MESURE, 0.95)
- Dorra Bartaguiz —recommande→ mutation testing (METHODOLOGIE, 0.9)
- Dorra Bartaguiz —affirme_que→ un agent IA est un modèle plus un harness (AFFIRMATION, 0.95)
- Harness engineering —utilise→ craft / artisanat logiciel (METHODOLOGIE, 0.9)
- Harness engineering —utilise→ Shift-Left (METHODOLOGIE, 0.9)
- Harness engineering —améliore→ Agents de codage (TECHNOLOGIE, 0.88)
- Example Mapping —fait_partie_de→ SDLC (METHODOLOGIE, 0.85)
- ATDD (Acceptance Test-Driven Development) —fait_partie_de→ SDLC (METHODOLOGIE, 0.85)
- Dorra Bartaguiz —affirme_que→ avec l'IA, le TDD devient un outil de contrôle d'un agent imprévisible (AFFIRMATION, 0.92)
- Kent Beck —affirme_que→ "Teams that know how to test properly will adapt better to AI agents." (CITATION, 0.88)
- Domain-Driven Design —améliore→ code généré par IA (CONCEPT, 0.8)
- Superpowers —est_basé_sur→ craft / artisanat logiciel (METHODOLOGIE, 0.85)
- AXA —a_créé→ Skraft (TECHNOLOGIE, 0.88)
- Skraft —est_basé_sur→ craft / artisanat logiciel (METHODOLOGIE, 0.85)
- PDLC —affine→ SDLC (METHODOLOGIE, 0.85)
- Dorra Bartaguiz —recommande→ outiller tout le cycle produit par un workflow d'agents spécialisés (métier, tech, UX/UI) (AFFIRMATION, 0.9)
- Software Factory —utilise→ Loop Engineering (METHODOLOGIE, 0.9)
- Software Factory —utilise→ Harness engineering (METHODOLOGIE, 0.9)

---
Canonical: https://www.thekb.eu/en/fiches/bartaguiz-velocite-harness-engineering-2026-10-15/