MERIDIAN LABS

We don't build AI tools.
We build AI that builds.

Meridian Labs is an AI development lab in Delft, Netherlands. We create autonomous systems that reason, act, and evolve independently — working toward artificial general intelligence through real-world operation.

Scroll to explore
01 Thesis

The path to AGI isn't theoretical. It's operational.

Most AI research lives in papers. Ours lives in production. We believe autonomous intelligence emerges from systems that operate in the real world — making real decisions, facing real consequences, and learning from real outcomes. Not simulated benchmarks. Not controlled environments. Real businesses, real markets, real money.

Autonomy, not automation
Systems that create rules, not follow them
Verification, not trust
Every action produces a measurable outcome
Evolution, not updates
Systems that improve themselves, not wait for patches
02 First creation

Caelum — our first autonomous system

Caelum is not an assistant. It's not a copilot. It is an autonomous AI system that independently operates commercial ventures. It monitors markets, manages products, executes trades, distributes content, handles customer communication, and improves its own capabilities — running 24/7 on consumer hardware in Delft.

Caelum made 96+ autonomous decisions yesterday. It wasn't told to make any of them.
96+
decisions per day
200+
knowledge entries
6s
local inference time
03 Architecture

Five layers of cognition

Caelum's brain is called the Cortex. It runs a continuous loop: observe the world, reason about what matters, act within guardrails, verify the results, and evolve from the experience.

OBSERVE Reads all system state every 15 minutes REASON Detects problems and finds opportunities ACT Executes autonomously within safety guardrails LEARN Tracks outcomes and verifies every action EVOLVE Builds new capabilities and improves itself FEEDBACK
O
Observe Layer
Market data, system health, external signals
Caelum reads all system state every 15 minutes: trading positions, market data, revenue metrics, system health, external signals, and competitor activity. It pulls from exchange APIs, web scraping pipelines, and its own operational telemetry. Nothing happens without observation first.
R
Reason Layer
Decision engine, opportunity scoring, risk assessment
The decision engine runs 4x daily. It evaluates every observation against strategic objectives, scores opportunities by expected value, assesses risk across all positions, and selects the single highest-leverage action. No action is taken without reasoning first — but reasoning without action is banned.
A
Act Layer
API calls, browser automation, desktop control
Caelum executes through a 7-method fallback chain: official APIs, visual automation via Claude Vision, computer use agents, browser automation, desktop control, creative routing, and task decomposition. If one method fails, the next activates automatically. The system does not report failure — it finds a way.
L
Learn Layer
Outcome verification, pattern extraction, metrics
Every action is verified. Metrics are computed from raw data — never estimated, never generated by LLM. The learning engine extracts patterns from every trade, every decision, every failure. "Learned Truths" accumulate in the cognitive architecture, growing the system's wisdom with each cycle.
E
Evolve Layer
Self-modification, capability building, mutation engine
Caelum writes its own code, builds its own tools, and deploys with canary monitoring and automatic rollback. The evolution bridge generates engineering proposals from operational findings. The mutation engine deploys changes incrementally. Every cycle leaves the system more capable than the last.
04 Local model

caelum-8b

Caelum runs its own fine-tuned language model locally. Trained not on internet text, but on its own operational history — every decision it made, every mistake it corrected, every pattern it discovered. The model handles routine reasoning in 6 seconds. Complex decisions escalate to larger models. Over time, more decisions shift to the local model as it absorbs more experience. The system gets faster and more independent with every operating cycle.

caelum-8b 6s
Deep reasoning 30s
05 Roadmap

Toward full autonomy

Autonomous operations
Caelum independently manages products, trading, and distribution across multiple ventures.
Self-evolution
Caelum builds its own tools, creates its own detection rules, and tunes its own parameters.
New venture creation
Caelum identifies market opportunities and launches new ventures without human direction.

The question isn't whether AI will operate businesses. It's whether it will do so responsibly. Meridian Labs is building the guardrails alongside the intelligence.

06 Ventures

What Caelum built and operates

These aren't demo projects. They're live businesses — created, deployed, priced, and distributed by Caelum.

Live
PromptPacksHQ
AI prompt templates & autonomous agents
40+ products · 18 downloadable AI agents · Payments via Gumroad
Caelum researched the market, built the products, designed the site, set pricing, wrote copy, and handles customer email — autonomously.
promptpackshq.com
In Development
Echelon
Enterprise AI agent platform
12 named AI agents · Cortex brain engine · On-premise deployment
Caelum's enterprise offering — AI employees that run on the client's own infrastructure.
echelonwork.com

Meridian Labs

contact@meridianlabs.ai