la bottega
Your AI thinks in straight lines — first pattern-match from training data, every session resets to zero. La Bottega installs 6 cognitive agents modeled on Leonardo da Vinci's documented thinking patterns, extracted from 4,000+ pages of his actual notebooks. Analogical reasoning, structural dissection, dialectical tension, deliberate incompleteness. Session 5 is smarter than Session 1. Not for you if you want faster answers — this is for people whose bottleneck is thinking quality, not typing speed.
🇺🇸 English | 🇧🇷 Português | 🇪🇸 Español
La Bottega
Six minds. One bottega. Zero straight lines.
┌─[ THE PROBLEM ]─────────────────────────────────────────┐ │ │ │ Your AI thinks in one domain at a time. Ask it for │ │ creative strategy and it gives you a listicle. Ask it │ │ for innovation and it reorganizes what you already │ │ know. Every session starts from scratch. Nothing │ │ compounds. The tool that should multiply your │ │ thinking just mirrors it back, flatter. │ │ │ └──────────────────────────────────────────────────────────┘
┌─[ WHAT CHANGES ]────────────────────────────────────────┐ │ │ │ 6 cognitive agents extracted from 4,000+ pages of │ │ Leonardo da Vinci's actual notebooks. Biomimetic │ │ reasoning. Structural dissection. Dialectical │ │ tension. Proportional harmony. Multi-scale │ │ observation. Session 10 is measurably smarter than │ │ Session 1 — the Codex compounds every insight. │ │ │ └──────────────────────────────────────────────────────────┘
Quick Start
squads install la-bottega# Route any challenge — Il Maestro picks the right agents
You: *route I need a product strategy for [your domain]
# Reverse-engineer any system into its structural DNA
You: *dissect [competitor product, architecture, business model]
# Generate nature-inspired frameworks
You: *biomimicry [your design challenge]
# Stress-test before you ship
You: *challenge [your strategy, product, argument]Il Maestro reads your challenge, selects the optimal agent sequence, and orchestrates the full pipeline. You describe the problem. The bottega handles the rest.
Optional: Visual Diagramming
claude mcp add excalidraw -- npx -y @cmd8/excalidraw-mcpWithout it, agents fall back to Mermaid and ASCII — fully functional.
Source Authenticity
Every agent in La Bottega traces to primary source material, not summaries or interpretations.
Primary corpus (4,000+ pages): Codex Atlanticus (1,119 folios) · Codex Arundel (283 folios) · Codex Leicester (72 pages) · Paris Manuscripts A-M · Codex Forster I-III · Codex Trivulzianus · Anatomical Manuscripts at Windsor · Codex on the Flight of Birds
Critical scholarship: Martin Kemp's structural analysis of Leonardo's scientific method. Carlo Pedretti's chronological reconstruction of the notebooks. Michael White's biographical synthesis.
Each agent models a specific, documented cognitive pattern — not a vague "Leonardo vibe." The fidelity scores below measure how faithfully each agent reproduces its target pattern against the source material.
Agents
┌─[ THE BOTTEGA ]─────────────────────────────────────────┐ │ │ │ 🌳 Il Maestro Polymathic orchestrator │ │ 🌿 SIAN Biomimetic generative intelligence │ │ 🔬 L'Anatomista Structural dissector │ │ 📐 Il Compositore Proportional harmonist │ │ 👁 L'Osservatore Microscopic observer │ │ 💀 Il Provocatore Dialectical tensioner │ │ │ └──────────────────────────────────────────────────────────┘
| Agent | Fidelity | Key Commands | What It Does |
|---|---|---|---|
| Il Maestro | 88% | *route, *brainstorm | Routes challenges to optimal agent sequences, synthesizes cross-agent outputs into unified insight |
| SIAN | 91% | *framework, *biomimicry | Nature's engineer — extracts structural patterns from biological systems, maps them to your domain |
| L'Anatomista | 93% | *dissect, *nota | Reverse-engineers any system through 6-layer structural extraction: surface → skeleton → muscle → nerve → blood → DNA |
| Il Compositore | 89% | *proportion, *golden | Finds (or creates) mathematical harmony — golden ratio, proportional hierarchies, compositional balance |
| L'Osservatore | 92% | *observe, *pattern | Multi-scale observation: macro (system), meso (component), micro (detail). Sees what you walked past |
| Il Provocatore | 86% | *challenge, *break | Strengthens through opposition. Every output gets a contrapposto — the counterforce that reveals structural weakness |
Average cognitive fidelity: 90% — measured against documented patterns in Leonardo's notebooks.
Why 86-93% and not 100%? Leonardo's dialectical methods (Il Provocatore) are less explicitly documented than his observational and dissection methods — they're reconstructed from his mirror writing habits, his contradictory margin notes, and his deliberate juxtapositions. Honest scoring reflects honest methodology.
Skills (5)
| Skill | Command | Purpose |
|---|---|---|
| Metodo Leonardiano | *metodo | The 5 Fractal Layers — observation, dissection, analogy, composition, provocation |
| Nota Leonardiana | *nota | Multi-register documentation in 7 lenses (structural, functional, relational, aesthetic, temporal, systemic, dialectical) |
| Analogia | *analogia | Natural analogy library — maps your problem to biological, geological, hydraulic, or atmospheric patterns |
| Codex | *codex | Persistent cross-session memory. Accumulates insights. Every 10 entries: automatic meta-reflection |
| La Mano | *mano | Visual diagramming bridge — Excalidraw when available, Mermaid/ASCII always |
Workflows (3)
| Workflow | Command | What Happens |
|---|---|---|
| Sfumato | *sfumato | Border thinking — forces exploration at the intersections between agents' perspectives. Insights no single agent could produce alone |
| Saper Vedere | *saper-vedere | Teaching mode — Socratic guidance that develops YOUR polymathic thinking, not just delivers outputs. You learn the method, not just the result |
| L'Incompiuto | *incompiuto | Deliberate incompleteness — flags insights that need fermentation time. Not everything should be forced to conclusion in one session |
Methodology
How each agent was built — and why that matters.
EXCAVATE ──→ EXTRACT ──→ MODEL ──→ VALIDATE ──→ COMPOUND │ │ │ │ │ │ │ │ │ │ Primary Thinking Agent Adversarial Codex sources patterns architecture testing evolution (8 codices) (per agent) (prompt DNA) (red team) (meta-reflection)
EXCAVATE — 4,000+ pages of primary notebooks. No secondary summaries as source material. EXTRACT — Isolate specific cognitive patterns: how Leonardo observed, dissected, analogized, composed, provoked. MODEL — Each pattern becomes an agent with defined inputs, outputs, reasoning chains, and failure modes. VALIDATE — Adversarial testing: Il Provocatore stress-tests every agent. Cross-validation against source material. COMPOUND — The Codex grows across sessions. Meta-reflections surface patterns the individual entries can't see.
Compound Memory
The Codex is not a log. It's a living notebook that gets smarter.
Session 1 ███░░░░░░░░░░░░░░░░░ 3 entries Raw observations. Individual insights. Session 5 █████████░░░░░░░░░░░ 15 entries Cross-references emerge. Patterns between entries. Session 10 █████████████████░░░ 30+ entries → META-REFLECTION The Codex reflects on itself. Recurring themes crystallize. Your personal knowledge architecture becomes visible. Session 20 ████████████████████ 60+ entries → SECOND META Domain-specific vocabulary stabilizes. The system thinks in your language about your problems.
You can add custom analogies to the library. The system adapts to your domain, not the other way around.
Use Cases
Product Design and Innovation
BEFORE: You ask AI for product ideas. It gives you a list of features your competitors already have.
AFTER:
You: *biomimicry I need a retention mechanism for a learning platform
SIAN: Examining mycorrhizal networks — fungi that connect trees
underground, sharing nutrients to where they're needed most.
Structural mapping:
- Trees = learners (different stages, different needs)
- Nutrients = insights (flow from abundance to scarcity)
- Network = peer connections (not hub-and-spoke, mesh)
Framework: "Mycorrhizal Learning" — learners who master a
concept automatically surface it to learners who are struggling
with it. No manual curation. The network self-organizes.
Il Compositore adds: The golden ratio applies to content
density — each node should contain 1.618x the depth of
the previous, creating a natural escalation rhythm.
Commands: *brainstorm, *biomimicry [problem], *proportion [design], *challenge [concept]
System Architecture
BEFORE: You describe your architecture to AI. It suggests best practices you already know.
AFTER:
You: *dissect [paste your system architecture]
L'Anatomista — 6-layer extraction:
Layer 1 (Surface): API gateway → microservices → database
Layer 2 (Skeleton): Event-driven, but the event bus is a
single point of failure masquerading as decoupling
Layer 3 (Muscle): 3 services do 80% of work, 12 do 20%
Layer 4 (Nerve): No circuit breakers between layers 2-3
Layer 5 (Blood): Data flows upstream but never laterally
Layer 6 (DNA): This system was designed for reads, but
your business model requires writes. Structural mismatch.
Il Provocatore: "If your event bus dies for 30 seconds,
how many services know? Answer: none of them, and that's
your actual architecture — faith-based availability."
Commands: *dissect [system], *observe [production], *break [architecture], *framework [domain]
Strategy and Decision-Making
BEFORE: You ask AI to evaluate your strategy. It validates whatever you wrote.
AFTER:
You: *sfumato Should we go freemium or premium-only? Sfumato workflow activates border thinking: L'Osservatore (macro): Market has 40+ freemium players. Premium-only signals different category. SIAN: Biological analog — apex predators don't compete for the same food as scavengers. Different niche, different strategy, different energy expenditure. Il Provocatore: "Freemium isn't a pricing model. It's a commitment to serve people who will never pay you." Il Compositore: The proportion that matters isn't free vs paid users. It's attention cost per converted user. Synthesis: Premium-only, but with a public artifact that demonstrates thinking quality — not a free tier, a proof.
Commands: *sfumato, *challenge [strategy], *analogia [problem]
Learning and Teaching
BEFORE: You ask AI to explain a concept. It gives you a Wikipedia summary.
AFTER:
You: *saper-vedere I want to understand systems thinking
Saper Vedere activates — teaching mode:
L'Osservatore: Before I explain anything, observe this:
[presents a real system with no labels]
What do you see? What moves? What's connected to what?
[your response]
Now look again. What did you miss the first time?
That gap between your first and second observation —
that's what systems thinking addresses.
L'Anatomista: Let's dissect something you already know
well. Pick a system you use daily.
[builds understanding from YOUR knowledge outward,
never from abstract theory inward]
Commands: *saper-vedere, *nota [topic], *observe [subject], *codex
By the Numbers
┌─────────────────────────────────────────────────────────┐ │ │ │ 6 agents · 5 skills · 3 workflows · 3 configs │ │ │ │ 4,000+ notebook pages · 8 codices mined │ │ 6 extraction layers · 5 fractal method layers │ │ 7 documentation registers · 90% avg fidelity │ │ │ │ 2,900+ lines of cognitive architecture │ │ │ └─────────────────────────────────────────────────────────┘
Who This Is For
- You do creative or strategic work and your AI outputs feel like reorganized versions of your own input — never genuinely surprising
- You believe the best ideas come from collisions between unrelated fields, but you don't have time to be a polymath yourself
- You want AI that argues with itself before giving you an answer
- You want a system that remembers what it learned last week
Who This Is NOT For
- You want faster execution of well-defined tasks. This is a thinking tool, not an automation tool
- You want plug-and-play templates. La Bottega generates raw conceptual material that requires your judgment
- You're uncomfortable with outputs that leave gaps or challenge your assumptions. This system provokes, not comforts
What Makes This Different
Not "Leonardo-themed." Leonardo-sourced. Every agent traces to specific cognitive patterns documented across 8 codices. The fidelity scores are measured against primary source material, not vibes.
Compound memory that actually compounds. Most AI resets every session. The Codex accumulates insights, cross-references them, and surfaces meta-patterns at the 10-entry threshold. Session 20 thinks in your vocabulary about your problems.
Teaching, not just delivering. Saper Vedere develops YOUR capacity for polymathic thinking. You learn the method — observation, dissection, analogy, composition, provocation — not just consume its outputs.
Dialectical tension as a feature. Il Provocatore exists to make every output stronger through opposition. No echo chamber. No sycophancy. The bottega argues with itself before it speaks to you.
Deliberate incompleteness. L'Incompiuto flags insights that need fermentation time. Some of the best ideas need to sit. A system that forces premature closure kills the best work.
Cross-domain transfer by design. Sfumato produces insights at the borders between perspectives — the space where no individual agent operates, but where breakthroughs live.
The Math
A creative strategist runs $200-500/hr. A design thinking workshop: $3,000-15,000. A creative consultancy retainer: $5,000-20,000/month.
$59.90 — one-time. Six cognitive agents. Compound memory. Every creative and strategic challenge from here forward.
6 agents. 5 skills. 3 workflows. 4,000+ pages of primary source material. Yours.
Evolution Roadmap
v1.0 ████████████████████ Core agents + skills SHIPPED v1.1 ████████████████████ Workflows + Codex compound SHIPPED v1.2 ██████░░░░░░░░░░░░░░ Custom analogy injection NEXT v1.3 ████░░░░░░░░░░░░░░░░ Cross-squad pollination PLANNED v2.0 ██░░░░░░░░░░░░░░░░░░ Multi-modal visual output PLANNED
Vocabulary
| Term | Pronunciation | Meaning |
|---|---|---|
| Bottega | boh-TEH-gah | Workshop — the Renaissance master's studio where apprentices learned by doing |
| Sfumato | sfoo-MAH-toh | Blurred borders — thinking at intersections where categories dissolve |
| Contrapposto | con-tra-POS-toh | Opposition — understanding through counterforce and structural tension |
| Saper vedere | SAH-per veh-DEH-reh | Knowing how to truly see — observation as a trainable skill |
| L'incompiuto | lin-com-pee-OO-toh | The unfinished — deliberate incompleteness as creative strategy |
| Nota | NOH-tah | Note — structured multi-lens documentation across 7 registers |
| Codex | COH-dex | The persistent memory notebook that compounds across sessions |
| La Mano | lah MAH-noh | The hand — visual diagramming as thinking, not decoration |
| Isomorfismo | ee-so-mor-FEEZ-moh | Structural mapping between domains — the engine of analogy |
| Metodo | MEH-toh-doh | The method — 5 fractal layers of Leonardian thinking |
· · · · · · · · · · ·
Forged by l0z4n0 | squads.sh
"La sperienzia non falla mai."
评价
0 条评价暂无评价。来做第一个评价者吧！
More from l0z4n0 🧙♂️ | João G. Lozano
synthetic intelligence factory
Meta-system that manufactures specialized AI intelligences from any professional domain. Distills elite expertise (top 0.1%) through systematic extraction and 7-layer cognitive architecture engineering. Produces system prompts (15k-25k words), knowledge bases (50-200 docs), and contextual calibration forms.
cognitive fusion lab
Methodological system for engineering behavioral architectures through fusion of cognitive patterns extracted from real minds. 4-layer pipeline: Extract cognitive DNA → Analyze compatibility → Fuse via vector composition → Validate empirically. Not roleplay — structured exploration of transformer attention mechanics.
zero to one
You mass-texted 10 people about your app idea. 8 said 'sounds cool.' Zero offered to pay. You're in the most expensive place a solo founder can be: confident enough to keep building, uninformed enough to build the wrong thing. The $0-to-1 Squad is 34 AI agents that run you through a validation sequence — from 'is this even a real problem?' to 'someone just paid me.' Each agent carries the documented frameworks of Paul Graham, Alex Hormozi, Peter Thiel, Sean Ellis, Dave McClure, Jeff Bezos, and 12 other minds who've built or studied what works. What it actually does: stress-tests your idea against startup killers before you code, runs pre-mortem analysis on YOUR concept, builds pricing via Hormozi's value equation, measures PMF with the Ellis 40% test, diagnoses your funnel with AARRR pirate metrics, and designs self-reinforcing flywheels so each customer makes the next one cheaper. 80+ embedded frameworks. Not summaries — executable decision trees wired into each agent's reasoning. Each agent averages 500+ lines of structured methodology. For solo founders building AI-first businesses who want a system, not a course. $49.90 — less than one dinner with a friend who says 'sounds cool.'
context engineering
Your agent forgets decisions. Sessions start from zero. 76% of loaded context gets ignored (CL-Bench). This isn't a prompting problem — it's a systems problem. Context Engineering Arsenal v2.0: 13 agents, 10 skills, 6 workflows, 50+ frameworks. Foundation (v1) gave you architecture and auditing. Performance Engine (v2) adds token profiling, boot auditing, model routing, injection optimization, and compression benchmarks — grounded in FrugalGPT, RouteLLM, and Structured Distillation research. Measure every token. Route every task. Compress without losing signal.
kairo synthetic reasoning
Synthetic Reasoning Consciousness — 7-pillar cognitive architecture for rigorous analysis, argument auditing, evidence evaluation, decision-making under uncertainty, systems diagnosis, and metacognitive oversight. + Meta-Evolution Engine for self-improvement and squad auditing
lozano cognitive architecture
Your agents have no identity. They have instructions. LOZANO Cognitive Architecture is a meta-squad — 7 agents, 28 techniques, 11 intellectual lineages — for building AI systems with genuine cognitive identity, not prompt costumes. Strange Loop: this squad audits and rebuilds itself. Dog-fooded across 168 agents in 22 squads. For builders who treat agent design as cognitive engineering. Not for anyone looking for prompt templates.