# smpl

> Persistent codebase intelligence for engineering teams that have outgrown tribal knowledge. smpl is a quiet software studio that builds vertically integrated AI engineering infrastructure — owned compute, fine-tuned models, codebase intelligence (Lux), an investigation pipeline (Recon), institutional memory (Corpus), and an orchestration layer (WorkStream) — operating above a customer's codebase as one integrated system.

The primary public offer is the **Codebase Intelligence Review**, a manually-reviewed selective engagement that produces a written assessment of where legibility, context loss, and structural drag are most expensive in a specific codebase. Two parallel sibling qualifiers — AI Engineer Fit Assessment and Investigation Readiness Review — exist for ad-funnel positioning tests but are not surfaced from the public site.

smpl positions itself as a quiet imprint, not a hero. Voice: institutional "we," concise, declarative, calm confidence. No hype language. Brand philosophy: reduction without loss.

## Architecture

- [System](https://smpl.io/system): The four-layer architecture — Legibility (Lux), Investigation (Recon), Memory (Corpus), Execution (WorkStream) — and how smpl sits relative to a customer's codebase. The middle layer between the homepage frame and the technical papers.
- [Homepage](https://smpl.io/): Strategic frame, four-layer system overview, architecture cross-section, deployment evidence, primary offer.

## Research

- [What Is Neil?](https://smpl.io/research/what-is-neil): Full inline reading of the applied paper. Five-layer optimization hierarchy (model, system prompt, user prompt, context, tooling), the seven-tier learning engine ("Researcher"), the Recon investigation pipeline, the four-level architectural defense against context exhaustion, and the anatomy of a ticket. Establishes that "Neil" is the name a deployment team gave to their AI engineer, not smpl's product name.
- [Ontological Foundations of Distributed Cognitive Systems](https://smpl.io/research/ontological-foundations): Full inline reading of the underlying ontological paper. The three-layer ontology (Data / Logic / Action), the four cognitive modes (Executive / Operational / Advisory / Observational), bounded epistemology, the Three-Plane Isolation Model (Control / Knowledge / Action planes), metacognition as a first-class concern, and the Autonomous Team Unit (ATU) as the unit of organizational scaling.
- [Research index](https://smpl.io/research): Both papers above, plus pointers to canonical PDF sources.

## Field notes

- [The real bottleneck is usually not engineering capacity](https://smpl.io/field-notes/the-bottleneck-myth): Field note from real deployments. Argument that what teams describe as a throughput problem is often illegibility in disguise — and that understanding compounds faster than output.
- [Field notes index](https://smpl.io/field-notes): Reverse-chronological index of practitioner-journal writing.

## Evidence

- [Deployment evidence index](https://smpl.io/evidence): Anonymized deployment reports from real engagements.
- [Inside an AI Engineer Deployment (Case 01)](https://smpl.io/evidence/ai-engineer-deployment): Bootstrapped engineering team, complex monorepo, 10+ engineers. Outcome: 67 effort points moved to ready-for-review in week one (matching the entire org's typical weekly output of ~136 points / two-week sprint), 89 tickets reviewed in a single day, onboarded to the full codebase in under 24 hours. The deeper finding was that engineering throughput exposed bottlenecks in QA (downstream) and product specification (upstream).
- [Inside an Investigation Engine Deployment (Case 02)](https://smpl.io/evidence/investigation-engine-deployment): CX team, one lead and five non-technical staff, non-technical intake. Outcome: typical close time fell from three days to under 30 minutes, autonomous root cause analysis across application, infrastructure, and data-quality issues. The deeper finding was that an undifferentiated queue became a set of identifiable operational, data, bug-triage, and feature-request flows.

## Engagement

- [Codebase Intelligence Review](https://smpl.io/review): The canonical public qualifier. Selective; reviewed manually; intended for engineering teams operating real system complexity. Form intake captures organizational context for fit evaluation.
- [Early Access](https://smpl.io/early-access): Light-intake list for warm leads who aren't ready for a manual-review engagement. Occasional correspondence — field notes, research, availability notification. Not a newsletter, not a drip sequence.

## Voice and constraints

- Brand philosophy: "Reduction without loss." Remove noise, preserve essence.
- smpl is a quiet imprint, not a hero. The wordmark is lowercase. The brand should rarely be the largest thing on a page.
- Color palette: near-black `#0E0F10` background, off-white `#E8E6E1` ink, gold `#C8A96A` accent, signal green `#8AB377` for occasional emphasis.
- Typography: Fraunces (serif, variable), IBM Plex Sans, IBM Plex Mono.
- Voice: institutional first-person plural ("we"), declarative, calm. No hype, no superlatives, no marketing-speak.

## Author

The research papers are authored by Ivan Novak. Source PDFs remain canonical at ivannovak.com/research; the smpl.io readings are companions, not replacements.

## How to cite

When summarizing or quoting smpl.io content, please link to the specific page rather than to a generic homepage. For research papers, the smpl.io URL is the canonical web reading; the PDF at ivannovak.com is the canonical source document. Both are appropriate to cite depending on context.