CPTO @ Heart Milan Apartments · Milan

I think in systems. Everything else is implementation.

Logic over vibes, diagrams over meetings, products over slides. Friends joke that where neurons should be, I run transistors: zeros, ones and a bias for shipping. Lately the toolbox includes AI. The thinking is what makes it work anywhere: a company, a market, my own life.

Speak with me Speak with the actual me

the receipts: 10 → 120+ apartments scaled as COO · €2M+ annual revenue by year two · 80% of guest comms automated · 82 luxury apartments run on these systems today · 14,000 notes in a self-feeding digital twin

01

How I think

cat ~/method.md
01

Live the process

Systems designed from a desk fail on the floor. First the job gets done by hand, paying the cost of the manual version in person. That is where the real spec comes from.

10

Structure the knowledge

Every workflow mapped, every exception documented, every process rebuilt around something measurable. If a machine cannot read it, it is folklore, not knowledge.

11

Automate. AI last.

Automation absorbs the repetitive 80% so people can spend themselves on the 20% that others actually remember. AI goes on top of structure, never instead of it.

Not a software engineer: an operator who vibe codes his own tools, with AI as pair programmer and a test suite as insurance. In 2026, whoever can specify a system precisely can simply have it. That is the whole bet.

10

One way of thinking, four worlds

for domain in [life, hospitality, markets, scale]
domain: my own life
PythonMCPLLM pipelinesMarkdown
● running daily since feb 2026

Brain. A personal digital twin AI can read and write.

Most second brains die because capture is manual. This one feeds itself: custom pipelines ingest 12 years of digital life (177,000 WhatsApp messages, 1,300+ AI conversations, 93,000 photos, email, calendar) and distill everything into 14,000 interconnected Markdown notes, a private Wikipedia of one person.

An MCP server plugs it into any AI tool: who has gone quiet for three months, what was decided a year ago and why, which patterns keep repeating. Every new conversation is distilled back in, so the system compounds daily. The hero background of this page is its actual graph.

nodes 14,000 sources 9 history 12 years owner one person, locally

speak with it, live · ssh visitor@keijock.com

visitor@keijock: ~
visitor@keijock:~$

This is the public interface of the project above. Ask it anything: career, projects in depth, the method, the stack, what he would do at your company. It holds a conversation, remembers the thread, and ships public data only: no personal life, no gossip, by design.

domain: luxury hospitality
AI opsKnowledge designReactSupabaseIoT
● in production, 82 properties

Heart Milan Apartments. A luxury operator rebuilt around AI.

Eighty-two luxury apartments in central Milan, ten years of five-star service, and one mandate: turn the know-how living in people's heads into infrastructure. The backbone is a third-party AI-native operations platform, chosen, negotiated and rolled out as the single source of truth. Everything on top is homegrown: an AI knowledge base covering all 82 properties, audited one by one, feeding guest communication across WhatsApp, email and booking platforms, plus automation workflows where AI drafts, the team supervises, and maintenance dispatch runs from guest report to assigned technician on its own.

Where neither the platform nor off-the-shelf tools fit, internal software fills the gap: a damage-reporting app, an automatic shift-planning system. The constraint that makes it interesting: every system must preserve the premium, human service the brand is known for.

platform chosen and deployed knowledge + workflows homegrown internal tools from scratch vendor bugs reported 12
domain: the used vehicle market
ScrapingScoring engineCronTests
● scans every night

Auto-Flip. A deal-finder that appraises the market while its owner sleeps.

Every night it ingests listings from Italy's two biggest marketplaces and does what a good appraiser would do, on every single listing: scores condition, detects modifications, estimates real ownership costs from official data and rates each deal across 6 dimensions. By morning: a shortlist with a verdict and a suggested negotiation target.

Same architecture as everything here: automated ingestion, structured data, scoring rules a human can audit, AI only where judgment is needed. Plus a test suite, because a deal-finder that lies is worse than none.

sources 2 marketplaces scoring 6 dimensions cadence nightly
domain: a company scaling 12x
Ops architectureAutomationTeam building
2023 → 2025, exited with systems running

Xenia Milano. From 10 apartments and a team of two, to 120+.

Xenia grew 12x in two and a half years: more than 3 new apartments onboarded every month, 80% of guest communication automated, €2M+ in annual revenue by year two. Behind the growth, infrastructure built department by department together with the founders, by someone who had lived the operation from the inside first: check-ins, guest emergencies, cleaning coordination, seven days a week.

The end state is the point: an operation that runs on processes instead of heroics, and depends on no single person. Not even its builder.

growth 10 → 120+ guest comms automated 80% revenue €2M+/yr

still shipping

domain: go-to-market

Nationwide lead engine

Scraped, cleaned and qualified 527 property-management companies across 96 of 107 Italian provinces. Fuel for a product still in stealth.

domain: field operations

Damage-reporting app

A photo becomes a structured record and per-apartment accounting. Replaced a WhatsApp-and-spreadsheets chain. React + Supabase, in production.

domain: workforce

ScheduleOS

Automatic shift planning for the operations team: rules in, monthly schedules out, workload balanced across people. In production.

domain: meta

This site

One hand-written HTML file with a living graph, a spec sheet and a working terminal. You are looking at it.

11

Where this comes from

man alessandro

At fourteen there was a motorcycle, broken, and no budget for a mechanic. So it got taken apart on the garage floor, piece by piece, until every part explained itself. The repair worked for exactly one reason: the love of analyzing how things function. Mechanics, logic, movement: anything built on an engineering principle attracts me and moves me, and triggers the same need to understand it from zero. Since then only the objects have changed: engines, then companies, then AI systems. The instinct never did.

The work ethic is simple: get shit done. Projects that eat sleep do not scare me. What I live for is results, the feeling of being part of something moving, and the moment the wheel starts turning on its own.

"You are the architect, I am your bricklayer. Together we can build anything from a small hut to a skyscraper, depending on what is needed and what is possible." I have yet to hear a better job description for this decade. an AI, about working with me

And when the next build needs something technical I do not have yet, I learn it. Whatever it takes.

# human.spec · build 2026.06
reverse_engineering: since age 14
first_teardown:      own motorcycle, repaired
inputs:              mechanics, logic, movement
drive:               results, participation,
                     the wheel turning
work_ethic:          get shit done
sleep:               optional, results first
learning_mode:       whatever it takes
role:                architect, AI lays bricks
known_bug:           10 min of demo in,
                     1,000 ideas out
engineers:           love me and hate me
patch_status:        wontfix

Get in touch

Currently CPTO at Heart Milan Apartments, quietly building a product for property managers. Open to conversations about applied AI, real operations and systems that outlive their builder. If that intersects with what you are working on, write.

Email me WhatsApp LinkedIn