RLabs Lab
Investigación aplicada.
Publicaciones técnicas del laboratorio. Documentamos lo que construimos y lo que aprendemos, con datos reproducibles cuando corresponde.
LLM Pipelines
abril de 2026 · Ramón Labbé
Show the Rubric Before You Ask for the Work
Quality Contracts as a Prevention Layer in LLM Generation Pipelines
Multi-stage pipelines for large language model (LLM) generation of structured artefacts — code, documents, specifications — commonly layer a terminal validation step on top of a linear chain of generation prompts. In practice, the generating model encounters the evaluation rubric only at that terminal step, after the artefact is already written. We propose a quality_contract block emitted at the top of every generation stage, carrying the same self-challenge criteria, expected structure, and validation checks the terminal step will use. Measured over the ten built-in archetypes shipped with AgentGuard, the contract adds an average of 1,498 bytes (approximately 374 tokens) per stage response. We argue the token overhead is favourable compared to a single correction cycle on a long artefact, and that the principle — show the rubric at generation time, not only at grading time — generalises to other LLM pipelines with declared evaluation criteria.
Leer más →
abril de 2026 · Ramón Labbé
Inside AgentGuard
A Four-Stage Pipeline for Structured LLM Generation
Most public discussion of LLM-assisted code generation stops at the prompt. The engineering reality of a production pipeline is different: there are stages, a contract between stages, a session store, a validator, and a set of boundary conditions that decide whether the pipeline can be paused, resumed, or replayed. This paper documents the four-stage architecture of the AgentGuard generation pipeline — skeleton, contracts-and-wiring, logic, and validate — and explains the design decisions behind each boundary. It is a companion to the author's prior work on the mechanism, the empirical benchmark, and the enterprise argument; its role is to make the architecture legible enough that both the benchmark and the enterprise argument can be evaluated against the system they describe.
Leer más →
abril de 2026 · Ramón Labbé
Archetypes as Enterprise Primitives
Why Form Beats Function for AI-Assisted Development
Large language models are increasingly capable of meeting the functional requirements asked of them — the code works, the feature ships. Enterprise adoption of AI-assisted development, however, is not bottlenecked by function; it is bottlenecked by form: repeatability across teams, consistency across releases, auditable boundaries between generated and hand-written code, and the governance that accumulates around them. This paper argues that declarative archetypes — in the AgentGuard sense — are a minimum-viable primitive for enterprise AI development, because they convert implicit prompt-level intent into an explicit, version-controlled artefact that scales to organisational size. The argument is theoretical and draws on contract programming, specification-driven development, and software architecture. It is the closing paper of a four-part series that describes the mechanism, measures it empirically, and documents the implementing architecture.
Leer más →
Theory
junio de 2026 · Ramón Labbé
The Two Pillars
A Structural Theory of Post-AI Software Work
The full development of the Two Pillars framework as an independent research monograph (233 pages). For seventy years the production of software was organized around one binding constraint: the human capacity to write correct code. Generative AI is dissolving that constraint. From that single deflationary observation — stated deliberately without naming a successor bottleneck — the monograph derives two structural consequences. The first, Mixer Mode, is a model of post-AI expert cognition: the simultaneous operation of multiple channels of criterion over a single domain, with orchestration as the behaviour that emerges when the mode operates over an agentic substrate. The second, Meta-Software, is the organizational capability of software that supervises software at machine pace, in four sub-categories — functional observability, structural validation, contextual continuity, automated governance — with security transversal to all four. The framework is grounded in a dual-position analytic autoethnography (thirteen experiential patterns), embedded in a formalized historical analogy with the industrial transition of 1910–1980, and made measurable by a two-layer instrument: aggregate individual Mixer Index read against organizational maturity, where the gap between the layers — the contraposition — is the diagnostic signal. A multi-case triangulation protocol and a falsifiable longitudinal prediction define how the theory is to be validated or refuted. Expands the preprint of the same name.
Leer más →
mayo de 2026 · Ramón Labbé
The Two Pillars
Mixer Mode and Meta-Software in the Reorganization of Software Work After AI
For seventy years the production of software was organized around one binding constraint: the human capacity to write correct code. Generative artificial intelligence is dissolving that constraint. This paper takes a single deflationary observation as its starting point — the production of code is ceasing to be the dominant problem in software-producing organizations — and argues that two structural consequences follow from it with a force approaching logical necessity. The first, Mixer Mode, concerns the human role: as agents absorb execution, the practitioner stops alternating among discrete specialized roles and begins to operate multiple axes of judgment continuously. The second, Meta-Software, concerns the organizational artefact: the durable output of a software team is no longer the running system but the executable knowledge — playbooks, archetypes, contracts — that agents read at runtime to behave with the team's intent. Together the two pillars reframe the next decade of software work: from writing code to modulating channels, and from shipping systems to maintaining organization-encoded software.
Leer más →