Mycelia Research | Deterministic Artificial Life

Choose the audience that matches your question

The page contains three selectable readings of the same subject.

Coding agent and operator assistance. Demo metrics are illustrative.

Terminal, search, lint, test, git, and more.

Remembers your codebase and team context.

Specialised agents collaborate across concerns.

Review diffs, request changes, final sign-off.

Replay any session bit-for-bit when something needs review.

Autonomous agents that write code and keep receipts

to handle network timeouts, 5xx responses, and idempotent-safe conditions. Pure helper, fully unit-tested.

Cheaper habits, reserve clears the threshold.

High capability, no descendants this generation.

environmental memory, held outside the organism

inheritance, variation and death, priced in one world

one deterministic substrate, replayable to the turn

what a signal carries, measured not assumed

memory and anticipation, only where they pay

what would count as a group becoming a cause

Research cooperation, not software adoption

uncertainty carried forward from the link before it

measured difference between baseline and branch

does the macro description constrain the future better than the micro one

difference measured, interpretation still open

difference measured directly on one body

The null control is redrawn at every scale and stays inside the band in all four, which is what keeps the difference attributable to the declared change.

null control, same checkpoint, shuffled intervention

the macro description constrains the future beyond the micro description

the intervention changed resource use and signalling

the intervention changed which lineages persisted

the intervention changed individual condition

Rejected, requested more material than the region holds.

the first artefact is a question that can be wrong

method, uncertainty, every class of result

The structure outlasts its individual members.

Local interaction builds persistent structure.

Internal regulation defends a working range.

Learning within life, selection across lives.

Heredity and variation both appear in the record.

Structure develops over the course of a life.

The account closes at every causal boundary.

Organisms acquire and spend material and energy.

the reserve has to reach back across the line, and only the lower day does

rests, and does not have a descendant this time

One confidence axis, and one state that refuses to sit on it

That is the question a replay can answer.

Sometimes, even as its members come and go.

Often, and that on its own means very little.

This result speaks for this world and this regime, at the functional level, and no wider one.

that the finding generalises to another regime

the intervention was the only declared change

The sentence the result is allowed to support, and no wider one.

Method and uncertainty are fixed before the result is read.

One declared change, with the controls it was tested against.

The context is stated so a rerun can be compared honestly.

context of the run that produced the history

Anyone can reconstruct the conditions from this layer alone.

law, regime, population, observation window

The one declared change and its controls

The world changes, and only one of those turns out to be useful.

Starts a little bolder. Finds more, and takes more risk.

Starts a little more cautious. Finds less, and lasts longer.

Learned to keep away from one corner of the world.

the same difference could easily have gone the other way

the world changes here, and nothing announced it

All four hold before any organism has an opinion about the field.

Nobody writes one, and nothing scores one.

New lives start slightly different from the old ones.

The ways an organism can meet the world.

The things life needs, and they can run out.

The band order is part of the world identity. A branch may only leave the ring at the commit boundary, which is why the two rows below the ring are comparable at all.

every intention is capability bound and priced

no organism reads a value written this turn

No route from the organism to the law, or to the host. The population moves, the apparatus does not.

Changes nothing. It reconstructs the same three records from the checkpoint.

Changes what descendants start from, and opens a new lineage record.

Changes regulation and memory inside one life.

Changes what happened to it, not what it is or what it carries.

four cards, one board, equal weight by construction

Is the lower level description already sufficient?

branch and baseline stay inside the band

What was missing that would have sustained it?

members are replaced, the pattern is not

No reinforcing outcome, structure decays. The association leaves the machine and the energy it was costing is released with it.

A memory the world has invalidated can be released.

in this world, under these conditions, the group changed what happened next

completed public test, 65 000 000 agents

does a macro description constrain the future

many interacting groups over long histories

what did it do to resource use and coordination

Three regimes of observation, and the boundary past them

observed convention, with evidence and uncertainty

Hand on a whole life's worth of reserve.

Hold on to what happened, or change the world.

Cross ground, or leave a mark others may find.

edges mark shared method, not data flow between systems

conservation, causal order, bounded action

organism 4c1 reproduced, reserve above threshold

field b decayed below sensing floor at region 12

conservation check closed, turn committed

it remembers a resource here, and it has already gone

it can tell something is there, not what

It has faded, and another mark has arrived on top of it.

It has spread, and bent around what was in the way.

Only something standing right here would meet it.

emitted by another organism, source unverified

a current measurement inside the boundary

Role, group and semantic detectors read the history. Enabling one leaves the identity of that history exactly where it was.

these stay empty while the world on the left keeps running

terrain, resources that run out, and four small organisms nobody has named

The second lane carries the same identifier into every reading, so there is no handoff left to argue about.

See the other research projects and the bottom-up, deterministic questions around Dweve's research.

Mycelia is a Dweve research substrate for deterministic artificial-life experiments. Any transfer to Loom, Nexus, or Core remains a separate validation question.

Access the simulation substrate, run deterministic replay experiments, test hypotheses about emergent behaviour, and contribute observations to the research record.

Related research and resources that frame Mycelia's current questions.

The research design hashes population state at tick boundaries and proposes cross-target comparison. A working regression suite and release policy remain to be validated.

The proposed engine pipeline includes a conservation audit for mass, energy, charge, and information. Failure handling and release checks remain part of validation.

The design advances each substrate engine by one logical step per tick. Where parallelism is needed, deterministic reductions are a proposed implementation path rather than a completed cross-target result.

A run is described by its configuration (substrates, parameters, and initial conditions) plus a random seed. The report treats the same configuration plus seed as a deterministic replay condition to test.

Proposed population hash and cross-target comparison.

Conservation checks described; drift validation pending.

Substrate engines advance one step per tick in the design.

Initial state from configuration plus seed.

The report supports deterministic replay as a testable property under a declared configuration and seed. Cross-architecture equivalence, build-failure policies, and million-tick runs remain future validation work.

Replay target: same population (to validate)

Cross-platform determinism (to validate)

Information provenance (model assumption)

The model tracks signal propagation, memory, and agent-level decisions. Its information-provenance rule is a modelling assumption, not a result established by the current report.

Conservation hypothesis under evaluation

Agent-rule design; no top-down forcing in the model

The model represents predator-prey, mutualism, and competition as agent rules. Whether conservation behaviour holds across runs remains a validation question.

Bottom-up population dynamics (research hypothesis)

Cell biology on a proposed chemistry model

The design models cell-level biology on a chemistry layer. Whether population dynamics emerge from interactions remains a research question.

Newtonian and statistical primitives (model scope)

The model proposes Newtonian mechanics at macro scale with statistical-mechanics primitives where needed. Energy drift checks remain to be validated.

Reaction networks with declared rate laws (design)

The design describes reaction networks with declared rate laws and tracks mass and charge. Test coverage and enforcement remain validation work.

No creation ex nihilo (model assumption)