NNMQWeb Systems
Work

What we have already shipped live

Live systems — AI, real estate, auto, casino, tracker. No price list here. If a project interests you, write by email.

Featured work · Braincore Pro

Braincore Pro — a decentralized AI swarm

A swarm of specialized AI agents inside your company. There is no master node that can fail: the nodes find each other on their own, they share a single board, and the work is picked up by whichever one is the most available.

< 2 s

node discovery (42 ms across 3 nodes)

277 / 281

tests passing (4 not passing, named publicly)

0

npm packages — Node built-ins only

24 / 24

golden eval on a real model

No central orchestrator

A task handed to any node is executed by whichever node is freest. Proven in the test suite, not on a slide.

Survives a hard kill

Under SIGKILL and an hour of load: 0 lost, 0 double-executed tasks, node death detected in ~1.3 s.

Encrypted peer traffic

HMAC-signed frames and AES-256-GCM encryption. No libp2p, no brokers, no Zookeeper.

Only numbers travel

The registry accepts eight fitness fields and rejects content — we cannot leak data we never receive.

Technology

  • UDP gossip over dgram — probes every 300 ms, fanout 2
  • CRDT blackboard (last-write-wins + vector clocks)
  • Redis queue with a zero-dependency RESP client
  • Stigmergy: pheromone trails, 30 s TTL, true exponential decay
  • RSI levels R0–R5 — every change needs human approval

All figures come from braincore.pro — from their test suite, not from marketing.

Featured work · AI Command Center

AI Command Center — two jobs: visibility, and AI under control

Not a retainer, and not “a tool that only measures”. The first job is visibility with proof: what Google and AI answers actually show, the fix, then the measurement again. The second is AI under control: an EU AI Act risk map, agent evaluation with thresholds, autonomy levels and a log of every step.

2

jobs — visibility with proof · AI under control

46

checks per evaluation cycle

4

autonomy levels — what the agent may do alone

0

unexplained outcomes — every step is logged

Visibility with proof

Daily checks of index, citations and AI Overviews; the fix is shipped and measured again. The proof is the number before and after, not a feeling.

EU AI Act risk map

Risk assessment and documentation before the build, not after an incident.

Evaluation with thresholds

46 checks per cycle. An agent that misses a threshold does not go into production — and the report says so.

Autonomy levels and logs

It is explicit what the agent does alone, what needs approval and why. Every step stays on record, with a monthly report for management and legal.

How it is used

  • assessment → build → monitoring
  • machine-readable pricing (llms.txt)
  • desks: /seo-geo/ · /board/ · /commander/ · /mirror/
  • no retainer, no meetings

Snapshot 24 Oct 2026. Prices and terms live on aicommandcenter.pro — deliberately not copied here, so this text cannot go stale.

Contact

Interested in a project?

Fill in three fields and the message lands straight in our inbox. No mail app, no waiting.

Curious about a project?dejanswat@gmail.com