Run Your Business on Autopilot.
Without Losing Control.

I built Ayla OS to run mine. 50 AI agents, 30 cron jobs, 2 businesses in production, zero babysitting. I'll show you how to build the same thing for yours.

Includes The Signal: weekly intel for builders, every Saturday.

Hichem Refes at his AI workstation with holographic dashboards

50

AI Agents in Production

30

Scheduled Cron Jobs

128+

Channels Monitored

0

Downtime. Zero Babysitting.

How the Autopilot Runs.

Ayla is not a chatbot. It's the orchestration layer: a chief of staff that delegates, monitors, and ships.

Hichem and Ayla collaborating on workflow automation

50 Agents. Each With a Job. None Need Supervision.

Each agent owns its memory, its workspace, and its own delivery pipeline. They coordinate through Ayla's hub, spawn sub-agents, write checkpoints, and ship results to wherever you work. Telegram for me. It's pluggable.

Three weeks deep in the next build, the work I delegated last month is still moving. That's the point.

agent.roster
Forge CTO / Architecture
Chief Operations
Herald Content / CMO
Compass Research
Amplifier Marketing

These five run the org. The other 45 run the work.

cron.schedule
memory-consolidator every 2h
intel-pipeline every 4h
context-watchdog every 2h
skill-evolver daily 00:30
docs-rebuild daily 09:00

30 jobs total. All monitored.

30 Cron Jobs. Memory Consolidation. Intelligence Pipelines.

Every two hours, Ayla consolidates agent memories, detects contradictions, and prunes stale context. An intelligence pipeline monitors 128+ channels for market signals.

Skills evolve nightly. Documentation rebuilds daily. Health monitors run continuously. The system maintains itself.

While you sleep, the system gets sharper.

The Patterns That Make 50 Agents Possible.

Ayla orchestrating the AI agent network
Hub Pattern

The Kernel Routes Everything

Ayla receives every task. She routes it, monitors delivery, reports back. No agent calls another directly. The kernel decides what runs where, and the main session stays free.

Single Entry Point Context Isolation Multi-Agent Coordination
Boundaries

Role Boundaries

Each agent has its own tool registry. Forge ships code. Herald ships copy. Compass reads but doesn't write. A planner that can also execute is a planner that breaks production code.

Memory

Memory Per Agent

Each agent owns its memory layer, workspace, and checkpoint trail. No shared state through implicit context. Memory is structured, persistent, queryable.

Trust

Permission Tiers

Three tiers. Agents that observe. Agents that draft and queue. Agents that ship to production. Trust is earned through track record, not granted by default.

Factories

Factory Abstraction

Routine work routes through a factory. Forge for engineering. Herald for content. Muse for imagery. Scribe for books. Factories own their voice, their inputs, their outputs. They don't ask the operator for context they could fetch themselves.

Voice Per Domain Self-Contained Context Reusable Patterns
Resilience

Recoverable Sessions

Every long task writes checkpoints. Every session has a state file. Crash Friday, resume Monday from where it stopped. The work doesn't depend on a process staying alive.

25 Years in ICT. Then I Built the System I Wished Existed.

25 years in ICT. The first five writing code. The next twenty inside telco, climbing from technical roles to Deputy Director at one of MENA's largest telcos. By the end I was spending 80% of my time managing politics, not building.

So I left. I gave myself access to a stack of AI agents, and at some point I noticed my Telegram had become a board of directors I controlled.

"I didn't want to build a chatbot. I wanted to build the operating system that makes one person as effective as a team of twenty."

Today I run Allwebzone (AI agency) and Digital Fennec (sister IT services company), both running on Ayla OS. The system handles content, research, operations, engineering coordination, and intelligence gathering. All autonomous. All monitored. All in production.