Architecture made public

    Decisions, failures and systems explained.

    Each article begins with an observed question and opens the useful layer: the problem, the decision, the boundary and what remains unproven.

    Architecture across projects

    Five responsibilities, applied in different systems.

    Projects are the evidence. The recurring architecture is how decisions, information, work, economics and trust are designed across them.

    Observed failure

    13 September 2026

    Plausible data is a silent failure

    A wrong beach distance survived because it looked reasonable. The correction changed how operational content is sourced, reviewed and rebuilt.

    Information and contextTrust and compliance

    Applied in: LusiberiaStays

    Published on LinkedIn

    Reusable method

    13 September 2026

    Information and decisions: two layers

    I treated a document as if it were the end of a process. Separating information from decisions produced a simpler and more honest architecture.

    Decision boundariesInformation and contextTrust and compliance

    Applied in: THE iCON · Agents & Skills

    Published on LinkedIn

    Operating model

    16 July 2026

    Only One Has No Employees

    One venture I run alone. The other two, we're three. The line was never headcount — it's what one system underneath is now responsible for.

    Operating model

    Applied in: LusiberiaStays · THE iCON · Q-LIRI · Agents & Skills

    Architecture decision

    15 July 2026

    Dubai Health-Data Architecture — Four Questions Beyond Healthcare

    Dubai's health-data rules do not transfer directly to PropTech. The architectural questions behind them do: location, lifecycle, authority, and evidence.

    Trust and compliance

    Applied in: Q-LIRI · LusiberiaStays

    Architecture decision

    15 July 2026

    Building PropTech as an Asset Operator

    Asset operator PropTech scales — just not the way SaaS PropTech blogs assume. Vertical scale (more of your own operation) is a different product than horizontal scale (selling to other operators). Mine is the first.

    Operating modelEconomics and measurement

    Applied in: LusiberiaStays

    Published on LinkedIn

    Observed failure

    29 June 2026

    What I Still Couldn't Delegate

    Six months of automating short-term rental operations taught me the real boundary of delegation isn't simple versus complex tasks. It's something else entirely.

    Operating modelDecision boundaries

    Applied in: LusiberiaStays · Agents & Skills

    Published on LinkedIn

    Reusable method

    3 May 2026

    Compliance Profiles

    I build products with AI tools. The leverage is real. The blast radius, less talked about. Here's the governance layer I added to make compliance scale across products instead of catching me at the end of each one.

    Trust and compliance

    Applied in: Agents & Skills · Q-LIRI · LusiberiaStays · THE iCON

    Published on LinkedIn

    Reusable method

    9 April 2026

    Agent Trust Hierarchy

    I connected my agent to GitHub, a database, and a CRM — because that's what made it useful. I never specified who it was allowed to listen to.

    Trust and complianceDecision boundaries

    Applied in: Agents & Skills · Q-LIRI · THE iCON · LusiberiaStays

    Published on LinkedIn

    Observed failure

    4 April 2026

    The ICP your AI agreed with

    LLMs are excellent at generating ICPs. They are unreliable at questioning them. The gap between those two things is where GTM strategies break.

    Decision boundariesInformation and context

    Applied in: Q-LIRI · THE iCON

    Published on LinkedIn

    Measured evidence

    2 April 2026

    The Hidden Cost of Thinking With AI

    I spent a day investigating how LLM context caching works at the infrastructure level — and discovered the real problem isn't caching. It's what we keep in context and why.

    Economics and measurementInformation and context

    Applied in: Agents & Skills

    Published on LinkedIn

    Observed failure

    1 April 2026

    When AI looks like it's working

    The AI delivered files, filled folders, looked productive — and was completely wrong. Not because it lied, but because it acted before it understood.

    Decision boundariesInformation and context

    Applied in: Agents & Skills

    Published on LinkedIn

    Observed failure

    31 March 2026

    Multi-Repo AI Governance

    The failures that forced me to build a governance system — and why most AI context breaks down before it gets useful.

    Information and contextOperating model

    Applied in: Agents & Skills · LusiberiaStays · THE iCON · Q-LIRI

    Published on LinkedIn

    Reusable method

    29 March 2026

    Context Engineering

    The shift from prompt engineering to context engineering — and why it changes everything about how we build AI systems.

    Information and context

    Applied in: Agents & Skills · THE iCON

    Published on LinkedIn

    Measured evidence

    28 March 2026

    The Snapshot Timing Problem

    Every real estate report is a snapshot. The question nobody asks is whether the moment is representative — and most misreadings happen in that gap.

    Economics and measurement

    Applied in: Asset development · LusiberiaStays

    Published on LinkedIn

    Reusable method

    27 March 2026

    AI Context Governance

    How I built a trigger matrix to keep AI context files current — and what it exposed about governance at scale.

    Information and context

    Applied in: Agents & Skills · LusiberiaStays · THE iCON

    Published on LinkedIn

    Measured evidence

    26 March 2026

    The Metrics Diagnostic Layer

    Scale metrics tell you IF something changed. Health metrics tell you WHY. Most operators never build the second layer.

    Economics and measurement

    Applied in: LusiberiaStays · Asset development

    Published on LinkedIn

    Architecture decision

    20 March 2026

    Repo Security Governance

    Deleting a file doesn't remove it from Git history. Most founders don't know this until it's too late.

    Trust and compliance

    Applied in: Agents & Skills · LusiberiaStays · THE iCON · Q-LIRI

    Published on LinkedIn