Looking for an IT solution that's reliable, cheap, and easy to maintain?

Atlas: a multi-project orchestrator that runs my whole build pipeline.

Atlas is my own multi-project controller. It dispatches headless Claude Code workers across a fleet of web projects running from a single Linux machine, so one developer can ship like a small studio. The internals stay private, but here is the shape of it: a controller, a queue of tasks, autonomous workers, and a human in the loop for the decisions that matter.

Controller
plans & splits
Queue
tasks wait
Workers
headless & autonomous
Deploy
human approves
atlas — task run (illustrative)
$ atlas dispatch --project example-site --task "add contact form"
controller task queued  #0042
worker-3   branch atlas/contact-form
worker-3   tests passed
✔ ready for review — deploy awaits human approval
~15projects
1machine
0agencies

An IT solution that's reliable, cheap, and easy to maintain.

Send us your project brief. Our own build pipeline — Atlas, the orchestrator powering everything on this page — breaks it into tasks and starts building. You get a working demo within days, not weeks. Every piece of work stays reviewed by a human before it ships: fast doesn't mean unsupervised.

Reliable

Human-reviewed before anything ships, and monitored after launch — not a demo that breaks the week after.

Cheap

An AI-assisted pipeline cuts the hours an agency would bill for — the saving passes straight to you.

Easy to maintain

Simple, well-documented builds — no framework sprawl, no vendor lock-in, no mystery code to inherit later.

Professional, fast, and properly maintained after launch — not a discount shop, a small studio built to move quicker than an agency twice its size.

The AI Lab: local models, private by design.

A personal lab for running local LLMs with Ollama, comparing models side by side, and building privacy-first AI tooling. This is where I test ideas before they go into production, and where the agent orchestration work started.

qwen2.5-coder:1.5b

Smallest, fastest — the default for cheap routed tasks.

Ollamalocal

qwen2.5-coder:7b

Stronger coding model, same family, heavier.

Ollamalocal

dolphin3:8b

A different model family, used for comparison.

Ollamalocal
ollama — local only (illustrative)
$ ollama run [model] "compare these two answers"
no data leaves this machine
🔒 Privacy-first

Live projects

mezquita.life

Retreat centre platform with booking flow and bilingual content.

BookingsEN/ES

Visit →

La Sociedad Psicodélica

Bilingual community platform with membership and events.

MembershipEventsEN/ES

Visit →

Natural Life Style

Platform for intentional communities and shared land.

Next.jsCommunity

Visit →

HempHuntersMap

Map-based directory of cannabis clubs and shops in Spain.

MapsDirectory

Visit →

TradingBit

An algorithmic trading platform — research and execution tooling, kept intentionally private.

PythonAutomation

Skills

Full-stack web
Linux and self-hosting
AI and local LLMs
Agent orchestration
Automation
Payments integration
Bilingual (English and Spanish) delivery
Design

About me

Berriedale Johnson

Autónomo developer in El Campello, Spain, running around 15 projects solo. I come from a background in activism and social movements, with ties to the UK.

Languages: English and Spanish, a little Italian, learning German.

Need a site that doesn't look like everyone else's? Let's talk.

Prefer email? Use the form above and we'll reply from there.

Phone / WhatsApp: +34 604 850 742

Instagram · LinkedIn · YouTube