MakiLab is NYG's product studio. One accountable person directs AI agents that research, design, build, test and operate the software. We ship our own products first; this manual documents them.
Fig. 1.2 The best time of the day, then every hour
123
Fig. 1.3 Navegar: wind over the sea
Figures 1.1 to 1.3 show the real interface with demonstration data for Madrid and the Aegean, fixed at 22 October 2026.
A free, non-commercial weather app for the phone. It finds the best time of each day to go out, says how far to trust the forecast, and adds official warnings, air quality and pollen, a rain radar and alerts on the device.
Its second tab, Navegar, animates ECMWF wind over a sea map. Tap any point and the table below becomes the time bar for wind, gusts and waves.
Private notes and tasks, encrypted on your device.
In service
Notes, quick notes, tasks with reminders, projects, folders and a calendar. Content and structure are encrypted before they leave the device, so the server stores what it cannot read, and search runs locally.
A private assistant that operates, under human direction.
Private · daily
A conversational operations system used every day inside NYG. It connects durable context, specialist agents and real tools, and answers in the channels people already use. It is not offered publicly.
Channels
Telegram, Google Chat, web chat
Works with
Memory, documents, calendar, email, home
Control
A person directs; the system executes
ConversationWhat someone asks
ContextWhat MAKI already knows
AgentsWho does each part
ToolsWhere the work happens
Fig. 4.1 How a request travels through MAKI
AI changes how software is made. Judgement still decides what should exist.
More code is not the outcome; better products are. Every product in this manual is in use, deployed on infrastructure we run, and improved in the same loop that built it.
5
Operating procedure
Fast because the system is clear.
Direct
One accountable person defines the problem, the boundaries and the evidence that will count as done.
Build
Specialist AI agents work across product, design and engineering while shared context keeps the work coherent.
Verify
Tests, reviews, security checks and real-device validation turn rapid execution into dependable software.
Operate
Deployment, observation and iteration stay inside the product loop. Shipping is the beginning of evidence.
Bring the problem. We will bring the system.
We take on a small number of external product collaborations where the problem is meaningful, the decision path is direct and production, not theatre, is the goal.