these are NOT visual effects. each frame is computed by python on the server.
PHYSICS — energy flows between alive neighbours via Ohm's law (R=0.2Ω), phase decisions are local-only (each node sees ~6 neighbours), supercap charge consumed proportionally to phase² · SELF-REPAIR — each node tracks neighbour heartbeats (150ms timeout). On loss, it removes routes through dead neighbour and learns new routes from surviving neighbours. No central controller. · ADAPTIVE COUPLING — phase 0=dormant, 1=sensing, 2=grip, 3=load-bearing, 4=rotation lock, 5=fused. Each node picks its own phase based on charge + load + damage. No global state. · DAMAGE — sustained over-load (load > 5N at phase < 3) increments damage. At damage=1.0 the node dies and the mesh routes around it.
OPEN TELEMETRY · LIVE OBSERVABILITY
researchers love observable systems · every metric here is computed by python from the live mesh · refreshes 4× per second
UPTIME
—
seconds physics has run
ENERGY
—
% of max capacity
HEARTBEAT
—
% live links responding
AVG ROUTES KNOWN
—
per node
BROWNED OUT
—
low-charge live nodes
FUSED · PHASE 5
—
at peak coupling
STIFFNESS / WATT · M2+ DOMINANT METRIC
—
Newton-resistance per Watt drawn · proxy until force-gauge calibrated
PHASE DISTRIBUTION · live histogram
0 dormant1 sense2 grip3 load4 lock5 fused
ENERGY · 60-second rolling
supercap recharge wave fires every 4s · watch the saw-tooth
JSON ENDPOINTS · curl them yourself
GET /api/dent/state · full snapshot
GET /api/dent/telemetry · metrics summary
GET /api/dent/routes/<id> · one node's routing table
POST /api/dent/damage/<id> · kill a node
POST /api/dent/heal/<id> · revive
POST /api/dent/load/<id>?n=6 · apply force
POST /api/dent/recharge · top up
FOUR-PHASE PATH · WHERE WE ARE · WHERE WE'RE GOING
paradigm-by-paradigm. each phase has explicit success criteria. each unlocks the next.
LIVE NOW
PHASE 1 · PROVE THE LATTICE
forget nanites. forget AGI claims. forget full self-healing biology. build the substrate first.
REQUIRES:
20–50 (we have 100) physical nodes
reversible magnetic coupling
variable stiffness states
local communication only
dynamic topology awareness
SUCCESS CRITERIA:
✓ node failure tolerated
✓ topology re-routes
✓ structure changes shape (phase 0–5)
✓ force distribution adapts
STATUS: simulated · running on python (this page) · all 4 criteria met in software · hardware milestones M1–M3 next
NEXT
PHASE 2 · MORPHOLOGICAL COMPUTATION
instead of "the brain controls the body" — "the body computes with the brain."
PRIMITIVES:
tension fields = information
vibration patterns = state
morphology = memory
coupling stiffness = cognition
core abstraction layer: two-way vibrational touch. not cameras. not symbolic models. field-state perception.
UNLOCKED BY:
working hardware mesh from Phase 1 · piezo sensors at each node · coupling-stiffness reading via current draw
RESEARCH
PHASE 3 · MATERIAL METABOLISM
the shard concept. transport and structure are the same system.
FUNCTIONS COMBINED IN ONE FLOW:
vascular logistics
self-routing material transport
mobile reinforcement
energy delivery
biology already does this. blood transports fuel + repair + signaling + immunity + temperature, all in one fluid. we recreate the concept computationally.
HARDEST PART · MOST REVOLUTIONARY IF IT WORKS
HORIZON
PHASE 4 · AUTONOMOUS MORPHOGENESIS
a robot today: has a predefined body. this direction: body emerges from task.
A SYSTEM THAT:
reshapes itself around force demands
grows support dynamically
hardens under load · softens for dexterity
repairs damage
routes power adaptively
this is no longer a robot in the classical sense. it is closer to programmable matter infrastructure.
THE DECISION · ROBOTICS PLATFORM OR UNIVERSAL SUBSTRATE?
the most important decision the dent matrix faces is whether it is a robotics platform (build a hand, sell hands), or a universal substrate where robotics, biology, cognition, metabolism, and manufacturing all converge into one medium.
these are different ambitions. the first scales like a product company. the second is a paradigm shift. the lattice we just proved supports either path — but the first hardware build chooses for us.
THE REAL EDGE · ABSTRACTION UNIFICATION
the edge is not AGI. not patents. not consciousness theory.
the edge is abstraction unification — naturally collapsing the boundaries between:
software ↔ structure
body ↔ compute
energy ↔ material
sensing ↔ morphology
repair ↔ growth
most researchers stay inside disciplines. this work doesn't. that produces either nonsense, or paradigm shifts. the determining factor is whether the physical primitives work.
WHAT'S NEXT · NOT MORE THEORY · REAL EXPERIMENTS
where the future of this system gets decided:
FIELD EXPERIMENTS tactile mesh on real fingers · grip force sweep
LATTICE MECHANICS FEM analysis · which lattice geometry maximises adaptive stiffness range
all parts are off-the-shelf, available from common suppliers. the simulation parameters match real-world specs.
PART
SPEC
SUPPLIER
£/UNIT
Microcontroller
ESP32-C3 (Wi-Fi + ESP-NOW mesh)
Mouser / Aliexpress
1.50
Supercapacitor
5F · 2.7V · 18 J energy store
RS · Farnell · Eaton HV series
0.30
Hall-effect sensor
A1324 (5mT analog magnetic field)
Allegro · Mouser
0.40
Coil + ferrite
1mm wire, 30 turns, 6mm OD ferrite
Aliexpress wholesale
0.20
H-bridge driver
DRV8837 (1.8A polarisation flip)
TI · Digi-Key
0.60
Boost regulator
TPS61023 supercap → 3.3V rail
TI · Mouser
0.50
PCB
2-layer, 15mm hex, JLCPCB Q100
JLCPCB
0.18
SMD passives
resistors / caps · 0402 reel cost
LCSC
0.15
PER NODE TOTAL
£3.83
100 NODES (matches this simulation)
£383
+ assembly (JLCPCB SMT, 100 panels)
£250
DESKTOP DEMONSTRATOR · TOTAL
~£633
HOW TO REPLICATE THIS IN HARDWARE · 5 STEPS
1. Layout 100 hex PCBs on a single panel · ESP32-C3 + supercap + coil per tile 2. Flash each ESP32-C3 with mesh firmware: ESP-NOW broadcasts a 1-byte heartbeat to all 6 neighbours every 16ms 3. Phase logic: each MCU runs the same `_decide_phase()` rule from dent_mesh.py — local context only 4. Self-repair: routing table is built from heartbeat receipts. Stop receiving = remove from routes. Standard mesh-net practice. 5. Power: aluminium-foil shards delivered via the polarisation flip act as both fuel + mass transit. Implementation: pre-charged supercap shards passed pad-to-pad via timed magnetic gradient.
NEXT FEASIBILITY MILESTONES
M1 (week 1-2) 7-node single-PCB demo · firmware ports the python rules · prove ESP-NOW heartbeat + reroute on physical kill M2 (week 3-4) 19-node hex panel · prove magnetic phase actually grips when current applied (force gauge measurement) M3 (month 2) 100-node full panel matching this simulation 1:1 · video proof of damage + reroute + grip transfer M4 (month 3) aluminium-foil shard delivery test · prove the metabolic loop closes (energy in via shard transit) M5 (month 4-6) 3D folded scaffold · origami locomotion · the body that walks
WHY THIS MATTERS
The page above is not a mock-up. dent_mesh.py is running on the server right now, computing real physics for 100 nodes at 60 Hz. Click a node to damage it — watch python's bellman-ford reroute the mesh. Recharge the panel — watch the supercap energies climb. The simulation parameters match real off-the-shelf parts (ESP32-C3, 5F supercap, A1324 hall sensor). The whole desktop demonstrator is <£700.
The Dent Matrix is not a deck. It's a buildable substrate. This page proves it computes. Next milestone proves it grips.