$ aemilius init --profile=modular --sandbox
✔ Core architecture initialized [zero-lock-in]
→ memory.attach(vector_bus) : ACTIVE
→ tools.mount([python_repl, web_eval]) : READY
⚡ Agent listening for task dispatch...
European community AI lab
Intelligence, in dialogue with life.
We build AI that helps us read the patterns, relationships, and systems that make the living world extraordinary.
Research-ledHuman-centeredLife sciences
LAB DIRECTORY //
01/02
Our projects / Project 001
Open sourceAgent baseEdge AIModular core
Aemilius agent
A copyable & ultra-modifiable agent base.
Modular architecture designed for strict separation of concerns — easily extensible, forkable, and adaptable to any workflow. Build, inspect, and deploy custom cognitive agents with zero vendor lock-in.
- TYPE
- Agent base
- FOCUS
- Copyable & ultra-modifiable
- PHASE
- Modular architecture
- DISPATCH
- Local & API workflows
AEMILIUS / ULTRA-MODIFIABLE AGENT BASE
Our projects / Project 002
Open researchDynamic runtimeAdaptive computeBio-inspired
Morph
An agent that can change how it thinks.
An experimental runtime for dynamic collaboration between specialized models. Rather than relying on a single static LLM, Morph orchestrates a population of capabilities that activate, sleep, or unload weights in real time.
- TYPE
- Open-source research
- FOCUS
- Dynamic agent runtime
- PHASE
- Early experimentation
- LIFECYCLE
- Active / Sleeping / Unloaded
MORPH / DYNAMIC MODEL POPULATION
01 / PRODUCT LOG / ARCHIVE
Latest release
Product launches, new capabilities, and milestones from the lab. Open the release archive to explore them.
> ACCESS /RELEASES
BROWSE RELEASES02 / RESEARCH NOTES / ARCHIVE
Insights
Research notes, perspectives, and ideas from the Phylogeny AI Lab. Enter the archive to read more.
> ACCESS /INSIGHTS
OPEN INSIGHTSHow we work / 02
A lab practice built around context.
Good AI research is more than a clever model. It asks who benefits, what the data leaves out, and whether the result holds up beyond the lab.
Research with people, not just datasets
Observe
Start with the people, evidence, and living systems behind the data.
Model
Build tools that reveal relationships while making their limits legible.
Validate
Test ideas in context, with domain experts and the communities affected.
OUR STORY // A EUROPEAN OPEN LAB
LAB MANIFESTO · INSPIRED BY NOUS RESEARCHAI should be free like the Internet. Not the private property of corporate labs.
Phylogeny is an independent European community AI lab. Inspired by the pioneering open-weights movement of Nous Research, we are building models and agents that are responsible, radically free, and adaptive. We reject the proprietary walled gardens pushed by Google, OpenAI, and Anthropic—because intelligence is humanity's next open protocol, not a corporate monopoly.
ACT 01THE ENCLOSURE // 2024
The corporate hijacking of intelligence
Google, OpenAI, and Anthropic are steering artificial intelligence toward extreme centralization: proprietary black boxes, monthly subscriptions, closed weights, and opaque safety theater engineered to lock in commercial monopolies.
Instead of empowering humanity, Big Tech is turning intelligence into a gated toll booth. They demand that researchers, developers, and citizens surrender their data and rent access to proprietary cloud servers. We refuse to accept a future where human cognitive progress is locked behind the walled gardens of three corporate boardrooms.
Phylogeny Manifesto // Section 01
ACT 02THE ALTERNATIVE // 2025
A European collective, inspired by Nous Research
Drawing direct inspiration from the decentralized open-source ethos of Nous Research, Phylogeny arose in Europe as an independent community lab for sovereign, responsible AI.
Nous Research proved that a distributed, passionate community can challenge centralized giants. Operating with European values of privacy, open science, and digital sovereignty, we build models and agent runtimes that are frugal, transparent, and adaptive. Rather than building frozen monoliths, we developed Aemilius Agent—a radically modular, copyable base—and Morph, our dynamic runtime where models sleep, wake, and free VRAM as tasks unfold.
Phylogeny Manifesto // Section 02
ACT 03THE OPEN COMMONS // 2026 & BEYOND
AI should be as free as the open Internet
The Internet transformed human civilization because no single corporation owned TCP/IP or the Web. Artificial intelligence must follow the exact same path.
The foundational cognitive infrastructure of the 21st century cannot remain the private property of a few corporate labs. We build for local laptops, university rigs, and decentralized edge devices—empowering anyone to run, inspect, and fork sovereign intelligence without permission. AI must be a public digital commons, owned by everyone.
Phylogeny Manifesto // Section 03
“The Internet was not created to be a cartel of three corporations, and artificial intelligence shouldn't be either. We stand with the open-source movement: building sovereign, adaptive models that belong to all of humanity.”
— Phylogeny AI Lab Collective · European Community Research
The Four Axioms
What we stand for.
Our principles are rooted in European digital sovereignty, open science, and sovereign agency over computation.
01 / DIGITAL COMMONS
Free Like the Internet
Just like Linux, TCP/IP, and the open Web, AI must remain an open protocol. Open weights, open codebases, and zero proprietary gatekeeping.
02 / NOUS SPIRIT
Inspired by Nous Research
We believe distributed, grassroots community research moves faster, thinks freer, and builds more faithfully for humanity than Big Tech bureaucracy.
03 / RESPONSIBLE & SOVEREIGN
Responsible by Design
European values of privacy, transparency, and energy efficiency. Verifiable models that respect human agency without hidden surveillance telemetry.
04 / ADAPTIVE & MODULAR
Adaptive Architectures
Rejecting rigid, frozen monoliths. We build decoupled, modular agents that adapt their compute, sleep when idle, and run frugally on local hardware.
100% OPEN
Inspectable & forkable weights
0 TOLLS
Zero Big Tech API lock-in
COMMUNITY
European independent collective
ADAPTIVE
Dynamic local agent runtimes
Join the open movement.
Inspect our code, fork the modular agent base, or build with our European research collective.