CasaSol
Marbella · Costa del Sol
v3.6 · 2026-05-10

Sovereign Intelligence for Local Real Estate

A model with full access to your private context outperforms a more powerful model that lacks it.

In many cases, it already does. The asymmetry that matters isn't compute — it's information. CasaSol is the operational instantiation of that thesis, on a Mac Mini, inside an estate agent's office in Marbella. Zero cloud. GDPR-by-design.

After The Sovereign Individual & Every Company Can Be a Palantir Now · Why CasaSol, localfirstai.eu

The Asymmetry

The information moat. Not the compute moat.

A frontier cloud model reasoning about a Marbella property has zero access to the agency's listing histories, viewer-aggregate patterns, comparable-sale deltas, and authored neighbourhood context. The local Coach has all of it.

This is the version of the claim that survives sceptical engineering scrutiny. "A smaller model wins" would be a weak claim. "A smaller model with full context beats a bigger model with no context" is the claim. The corollary the booth makes physical: a Mac Mini with all of an agency's lawfully-held context will out-answer Claude Opus 5 on questions about that agency's market, every time, because Claude Opus 5 cannot see the data.

General intelligence is commoditised at the API level. The portals own flat data — price, beds, baths, postcode. The only remaining moat for a Costa del Sol agency is hyper-local context: which streets flood, which schools have waitlists, which sub-areas turned over in Q1 and at what discount to asking. Today that context is wasted.

The Real Constraint

What CasaSol cannot do.

Not all knowledge is digital. Much of it is tacit, fragmented, or poorly structured. Local models do not solve this automatically. But where data is rich, consistent, and private, it has a structural advantage.

The Sovereign Individual, localfirstai.eu

CasaSol turns structured private data into intelligence. It does not extract tacit knowledge that hasn't been captured. An agency whose CRM is rich, consistent, and lawful gets the asymmetry advantage immediately. An agency whose CRM is mostly contact data gets a Coach with very little to mine.

01

Commercial precondition

Data richness is required for commercial value, separate from technical and legal preconditions. Every prospective alpha agency gets a 30-minute audit of CRM shape before we commit to a pilot.

02

Honest pitch surface

At the booth and in any agency conversation, what CasaSol cannot do is part of the pitch. Most booths claim everything; CasaSol claims a limit, on purpose. The honesty filters for the right kind of seriousness.

03

Adjacent service surface

Data-poor agencies are not lost prospects; they are consulting prospects of a different shape — help us capture tacit knowledge into structured form first. We do not sell that today. Whether we should is a v3.6 question.

04

The audio horizon — solving input friction

Tacit knowledge doesn't extract itself, and agents hate typing. The v3.6 architecture bypasses the keyboard entirely. An agent dictates a 30-second voice note from the car after a viewing. An on-device Whisper model transcribes it inside the hardware perimeter. The ontological router splits the transcript — private leverage goes into the vault, clean physical attributes surface as public tags. Zero cloud APIs. Zero typing. Live at OLÉ.

The Engine

Three layers. One physical boundary. The firewall is filesystem, not policy.

CasaSol is the architect of the boundary between what the agency lawfully knows internally and what the cloud is allowed to see externally. The boundary is enforced by hardware and process permissions — not by prompt instructions.

INSIDE THE AGENCY PERIMETER Mac Mini · M4 Pro · FileVault · no egress TIER 1 — THE COACH Gemma 4 · 26B listing patterns · viewer aggregates comparable sales · property attributes → agent_intel.json PASS 2 · redaction schema TIER 2 — THE LAUNDERER Same model · second pass forbidden-fields check · code-enforced enriched with property-attribute tags → public_listings.json FILESYSTEM FIREWALL no read access to ~/.casasol/private/ TIER 3 — THE BOUNCER Gemma 4 · E4B (Router) + MCP routing-plan schema · ADR-001 reads only public_listings.json three tools · live since 2026-03-29 MCP CLOUD buyer's Claude buyer is the controller of their own query data No personal data crosses any boundary. Article 22 out of scope by construction. runs nightly on every change on every query A misconfigured prompt cannot leak the intel layer because the intel layer is not in the process's filesystem view.
Fig. 1 — The Engine
TIER 1 The Coach

Internal intelligence. Never leaves the building.

Gemma 4 26B reads lawful agency data — listing histories, price-change patterns, days-on-market, aggregated viewer feedback, comparable-sale data, the agency's own neighbourhood guides. Emits agent_intel.json to the agency dashboard. Property and market patterns only — never inferences about identifiable natural persons.

TIER 2 The Launderer

Schema-enforced redaction. Richer than handwritten copy.

Same Reducer, second pass, different system prompt. Generates public_listings.json against a versioned redaction schema enforced in code, not prompt. Inherits property-attribute tags from Tier 1 (commute time, school catchment, flood-extent overlap) — a richer surface than the agency could write by hand, with zero personal-data exposure.

TIER 3 The Bouncer

Edge connector. The boundary is physical.

Gemma 4 E4B fronts the MCP server, emitting routing plans against an ADR-001 discriminated-union schema. Plans dispatch to three tools — search_properties, get_property, market_summary — already proven against Claude Desktop. The MCP process has no read permission on the private layer. A prompt-injection attack fails at the OS, not the prompt.

The Sovereignty Pair

Two sovereign parties. One negotiated boundary.

The agency is sovereign over its lawfully-held listing data. The buyer is sovereign over their own query data, through their own contract with their AI provider. MCP is the protocol over which they transact. Neither surrenders data to the other; neither surrenders data to a third Leviathan.

buyer@claude ~ ❯
"I'm moving to Marbella from London, working at PTA Málaga, budget €400k. I want a quiet street that doesn't flood, walkable to a primary school in English, and a 50-minute morning commute is the maximum I'll accept. Show me three options and tell me what you'd worry about with each."
// router routes →
plan: search_properties
budget_max: 400000
  commute_to: "PTA Málaga"
  commute_max_min: 50
  flood_overlap: false
  school_walkable_english: true


bouncer 3 listings matched, enriched with property-attribute tags
reads from public/ only · no personal data in scope · plan validated against ADR-001

Three places that hit the brief, with the worry attached to each.

Nueva Andalucía — three-bed townhouse, €385k. 42 min A-7 to PTA walkable · Aloha International outside 2024 SAIH flood layer Quiet cul-de-sac. Worry: ground-floor unit faces north, the courtyard sees light only in summer afternoons.

Cancelada — semi-detached with garden, €398k. 38 min A-7 to PTA walkable · St. George's Cancelada outside flood extent Worry: A-7 sound-bleed if the bedroom faces east; a viewing should happen during morning rush.

Sierra Blanca lower — two-bed apartment, €399k. 48 min A-7 to PTA walkable · Aloha Worry: at the upper end of your budget for the metres you get, and the comparable-sale layer shows the sub-area closed 11% under asking in Q1 2026 — there is room to negotiate.

Because the buyer is the controller of their query data and the agency is the controller of its listing data, the controller / processor mapping is unusually clean. This is the consequence of the architecture, not its purpose. The architecture is the sovereignty pair; clean compliance falls out of it.

The Compliance Pack

In Marbella luxury real estate, the lawyer is in the room before the IT director is. The pack is a sales artifact.

CasaSol ships with a compliance binder the agency hands to its DPO on day one. Eight artifacts, written by an EU GDPR practitioner, reviewed by the agency's own DPO before any corpus is generated. Bilingual (English canonical, Spanish executive summary) for OLÉ Marbella.

Records of Processing Activities

Pre-filled RoPA template, Article 30 aligned, ready for the agency to adapt and sign.

Data Processing Agreement

Vendor ↔ controller DPA between CasaSol and the agency.

Legitimate Interests Assessment

LIA documenting the Article 6(1)(f) basis for the Coach layer.

Privacy Notice Template

For agencies surfacing Bouncer queries to public buyers.

Data Subject Request Procedure

Access, rectification, erasure, restriction, portability — with documented appliance commands.

Breach Notification Procedure

72-hour template aligned with AEPD reporting requirements.

Retention Schedule

Tier 1 outputs retained on a documented schedule. Raw CRM never imported beyond the pipeline pass.

Audit Log Specification

What the appliance logs, where, and for how long. Bounded to the physical machine.

For Four Audiences

The booth signage, transcribed. Pick your row.

For Engineers

Ask me about the 22K cliff and how the Tier 1 → Tier 3 firewall survives a prompt injection.

Prefill cost on Gemma 4 26B-A4B grows super-quadratically past ~25K tokens of on-the-wire prompt. The bottleneck is memory bandwidth, not VRAM. Mitigations: cliff-aware coarsening, an 18K hard prompt budget, a streaming watchdog, and one-key restart of ollama serve.

Chronos Incident 003-Alpha · ADR-001 schema · OpenClaw cascade · Apple Silicon UMA

For CTOs & Founders

Ask me how your CRM's "useless" notes become an AI moat — without sending one byte to OpenAI or Anthropic.

The agency's listing histories, viewer aggregates, comparable-sale deltas and authored neighbourhood guides are exactly the dataset that converts a generic LLM into local intelligence. None of it leaves the appliance. The buyer's Claude pattern means the cloud's discovery surface works for you, on your terms.

Public portals own static snapshots — beds, baths, asking price. True alpha lives in velocity. The v3.6 Temporal Delta Engine parses historical listing decay, viewing-frequency drops, and agent note sequences to surface real-time pricing leverage. Static facts belong to Idealista; market velocity belongs entirely to you.

moat economics · MCP integration · Temporal Delta Engine · vendor independence · what an alpha pilot looks like

For Lawyers & DPOs

Ask me about Article 22 and the Continuous Compliance CI/CD watchdog.

Architecture forbids unlawful processing at the data-shape level, not at the policy level. Forbidden inputs and forbidden outputs are enforced in code via the redaction schema. Article 22 is out of scope by construction. The compliance pack — RoPA, DPA, LIA, DSR procedure, breach procedure — is part of the deliverable.

Static compliance binders are snapshots. The v3.6 Programmatic Compliance CI/CD layer sweeps every listing against 8 pre-registered GDPR liability categories before it reaches the public corpus — generating a daily record of zero leaks. The methodology is pre-registered and the evidence is public before the claim was made.

controller / processor mapping · lawful basis · CI/CD compliance gate · Exp 006 evidence · DPO walkthrough

For Sceptics

Ask me what CasaSol can't do. The honest answers are part of the demo.

Tacit knowledge does not extract itself. Agencies whose CRM is mostly contact data do not get a useful Coach. The asymmetry advantage requires data that is rich, consistent, and private. Where it isn't, no model — local or frontier — substitutes for the work of capturing it. Most booths claim everything; this one claims a limit, on purpose.

data-richness audit · what disqualifies a pilot · the Real Constraint · failure modes

The Substrate

The hardware sits on a desk. The repository is in Europe.

No hidden infrastructure. The development substrate matches the deployment substrate — both inside the perimeter we ask buyers to trust.

Hardware
Apple Silicon · M4 Pro
Unified memory architecture. ~273 GB/s memory bandwidth. The bottleneck for local inference, named honestly.
Models
Gemma 4 · 26B + E4B
Open weights, served by Ollama. Reducer (26B) does the work. Router (E4B) does the routing.
Orchestration
OpenClaw cascade
Chronos Experiment 005 — Reducer + Router pattern. ADR-001 routing-plan schema as the wire format.
Protocol
Model Context Protocol
MCP ↔ Claude Desktop bridge proven 2026-03-29. Three tools, live, against the 58-listing corpus.
Forge
Codeberg, not GitHub
European, non-profit, FOSS-aligned. The sovereignty pitch shouldn't be hosted on Microsoft-controlled US infrastructure.
Operating envelope
22K-token prefill ceiling
Hard prompt budget at 18K. Cliff-aware coarsening. Streaming watchdog. Documented in Incident 003-Alpha.
Where to Find Us

No pitch deck. No QR-coded slides. A Mac Mini on a table, and a question for you.

The left screen runs two demos back to back. First: an agent voice note — recorded on a phone, AirDropped to the appliance — transcribed on-device by Whisper, split into private vault and public surface by the router, and redacted to a clean GDPR-safe listing in under 30 seconds. No cloud. No typing. Second: a synthetic toxic agent note redacted live by the same model, the same prompt that passed a pre-registered 20-fixture fidelity sweep with 0 leaks across 8 GDPR categories.

The right screen runs Claude Desktop against the live 58-listing corpus via MCP — real questions, real local data, real answers. Next to it: the B200 postcard, full-size, on heavy stock.

What you see was already running on that machine before the conference. The compliance evidence is pre-registered and public — run before the system was claimed to work.

Startup OLÉ · Marbella
June 17–18, 2026
A local product, at the local event. Look for the Mac Mini — there is only one.
No LinkedIn connection requests. No newsletter signup. The capture mechanism is a heavy-stock card you'll find inconvenient to throw away. → tickets & venue at startupolemarbella.eu