Echo

One adaptive intelligence across every scientific frontier.

Echo is an Adaptive Structural Intelligence (ASI): a governed cognitive system that turns hard problems into mathematical models and testable hypotheses, across disease, physics, materials, climate and the governance of AI itself.

Built by one engineer in Adelaide over six months, on a desktop.

See the results Partner with Echo
stable unstable ρ = 0 1.0 Sepsis 0.77 Alzheimer’s 0.86 Glioblastoma 0.91
Spectral abscissa of each disease model before and after a modelled intervention. Negative means the model’s governance system is stable again. These are outputs of a mathematical model, not patient outcomes.

Six functions keep a living system in order. Disease is what happens when they fail together.

Echo’s Multi-Cell Governance Framework (MCGF) represents each disease as a collapse across six interacting regulatory functions. Drugs become perturbations to that system, cell lines become failure patterns, and the question “what would fix this” becomes something you can compute.

Echo can also reverse-engineer the mathematics back into biology. That makes a personalised oncology framework possible, and the next step is an immunologist to build it with.

Identity

What a cell is, and what it is allowed to become.

Surveillance

Noticing when something has started to go wrong.

Calibration

Keeping each response in proportion to the threat.

Effector

Actually acting on the problem once it is found.

Memory

Retaining what was learned for next time.

Tolerance

Knowing what not to attack.

Echo builds a model for each problem instead of forcing every problem through one workflow.

That is what ASI means here: Adaptive Structural Intelligence. Adaptive in how it approaches a problem, structural in how it carries what it learns across domains.

Adaptive

Echo assembles the reasoning, models, tools and compute a problem needs, with persistent memory across sessions and several models working together.

Structural

Its MFQL layer represents problems by their structure, so a pattern learned in one domain, like governance failure in sepsis, can be tested against another, like grid instability.

Governed

Echo learns and improves itself, but inside explicit authority limits, with provenance on every claim and a way to roll back any change. 56,913 claims and 69 locked predictions are tracked so far.

How Echo changes itself

No part of Echo is allowed to be the only judge of a change to itself.

  1. 1Propose a change
  2. 2Run it in a sandbox
  3. 3Independent and adversarial tests
  4. 4Signed approval
  5. 5Canary release
  6. 6Monitor
  7. 7Roll back if it fails

What Echo has produced

A selection of results so far.

Disease coverage
5,485 governance taxonomies generated, one for each MeSH disease descriptor.
Curated disease models
42 hand-curated models covering about 450 subtypes, with 375 hypotheses.
Treatment optimisation
Modelled interventions return sepsis, Alzheimer’s and glioblastoma models to stability.
Drug synergy (NCI-ALMANAC)
On held-out cell lines with real Bliss ComboScores, Echo’s pipeline reached Pearson r = 0.643, up from a 0.429 baseline. Built and run by Echo in about four hours. More compute takes it further.
Cross-domain transfer
Echo turns a problem in one domain into mathematics, then carries that structure into another. In one example it took cancer, made it a math problem, and converted it into a fisheries management problem: on a synthetic 60-species set with 30 unseen species, cold-start r = 0.786 and warm r = 0.973.
Carbon Growth Material (CGM)
A building material that is grown, not manufactured. Sand, seawater and a co-culture of cyanobacteria and calcifying bacteria bind into stone at room temperature in 5 to 7 days, at 5 to 15 MPa and about 5 kWh/m³, against roughly 500 kWh/m³ for concrete. Designed to scale from a ~$195 first brick to panel factories and ocean platforms.
Room-temperature superconductivity (RTSC)
Hypotheses for superconductivity at room temperature, including light-driven (Floquet) approaches.
Engineering prototypes
14 designs, including algae photobioreactors, biochar kilns and biogas systems.
Governance controls
Recursion governance, mutation watch and adversarial tests run internally.
150+ further discoveries
Across physics, oncology, climate, materials and more.

Cancer was where Echo started. It is not where it stops.

Ecology

The same structural approach that models disease is being tested on species co-management, starting with the fisheries benchmark.

Climate and energy

Echo has produced design concepts for carbon removal, biogas and energy-transition systems that are ready for engineering review.

Governing AI

The six governance functions apply to autonomous AI as much as to cells: identity, monitoring, calibration, bounded action, audit and safe shutdown.

Echo needs what a desktop can’t give it.

Ideas are no longer the bottleneck. Testing them is. If you can help move a result one column to the right, get in touch.

Email Kyle
  • An immunologist
    to build Echo’s personalised oncology framework.
  • Research labs
    to run non-clinical assays and try to reproduce Echo’s predictions.
  • Data holders
    with real drug-combination or ecological datasets to test transfer beyond synthetic benchmarks.
  • Compute and engineering partners
    to finish the memory-bound experiments and build the first prototypes.