SDR (GNU Radio) as self-learning REPL with Borges infinite library exploration, reafference random walks, spectral gap 1/4, and maximally mixed state for agency-enabling signal processing.
"The Library of Babel contains every possible radio transmission." — Borges meets GNU Radio
SDR as Infinite Library: Software Defined Radio is a Borges library where:
┌─────────────────────────────────────────────────────────────────┐
│ SPECTRAL GAP AND MIXING TIME │
├─────────────────────────────────────────────────────────────────┤
│ Gap = 0: τ_mix = ∞ (stuck, no exploration) │
│ Gap = 1/4: τ_mix = 4 (balanced exploration/exploitation) │
│ Gap = 1/2: τ_mix = 2 (fast mixing, less diversity) │
│ Gap = 1: τ_mix = 1 (instant mixing, no memory) │
└─────────────────────────────────────────────────────────────────┘
WHY 1/4?
- Alon-Boppana bound for d=3 regular graphs: λ₂ ≤ 2√2 ≈ 2.83
- Normalized gap = (d - λ₂)/d ≈ 0.25 for Ramanujan graphs
- This is OPTIMAL for 3-way (GF(3)) systems!
The maximally mixed state ρ = I/d is:
# Maximally mixed state in 3-dimensional GF(3) space
ρ_max = [1/3 0 0; 0 1/3 0; 0 0 1/3]
# Purity: tr(ρ²) = 1/3 (minimal for d=3)
# Von Neumann entropy: S = log(3) ≈ 1.585 bits
# AGENCY MEANING:
# - You can become MINUS, ERGODIC, or PLUS with equal probability
# - No prior commitment constrains your action space
# - Maximum optionality = maximum agency
| Property | Maximally Mixed | Pure State |
|---|---|---|
| Entropy | log(3) = 1.585 | 0 |
| Purity | 1/3 | 1 |
| Agency | Maximum | Minimum |
| Predictability | Minimum | Maximum |
| Adaptability | Maximum | Minimum |
Key Insight: Agency requires uncertainty. A system that knows exactly what it will do has no choice.
From von Holst's principle:
Efference copy: What I intend to transmit
Reafference: What I observe being transmitted
Error: Reafference - Efference copy
If error = 0: I caused this (self-generated)
If error ≠ 0: World caused this (exafference)
┌─────────────────────────────────────────────────────────────────┐
│ SDR REAFFERENCE ARCHITECTURE │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────┐ Tx Signal ┌──────────┐ │
│ │ INTENT │ ───────────────► │ TX SDR │ ──► Antenna │
│ │ (Efference│ └──────────┘ │
│ │ Copy) │ │
│ └────┬─────┘ │ │
│ │ ▼ │
│ │ Environment │
│ │ │ │
│ │ ▼ │
│ ┌────▼─────┐ Rx Signal ┌──────────┐ │
│ │COMPARATOR│ ◄─────────────── │ RX SDR │ ◄── Antenna │
│ │(Reafference) └──────────┘ │
│ └────┬─────┘ │
│ │ │
│ ▼ │
│ Error Signal → Learning → Updated Tx Parameters │
└─────────────────────────────────────────────────────────────────┘
# Reafference-guided random walk on spectrum
function reafference_walk(sdr, initial_freq, steps)
freq = initial_freq
history = [freq]
for step in 1:steps
# Efference: intended frequency shift
Δf_intended = randn() * bandwidth
# Actual observation (with channel effects)
signal = receive(sdr, freq + Δf_intended)
Δf_observed = estimate_frequency_shift(signal)
# Reafference error
error = Δf_observed - Δf_intended
# If error ≈ 0: we caused this (stay/exploit)
# If error ≫ 0: world caused this (explore!)
if abs(error) < threshold
freq += Δf_intended # Exploit
else
freq += error # Explore world's contribution
end
push!(history, freq)
end
return history
end
The acset-skill provides categorical database operations:
using ACSets
# SDR Observation schema
@present SchSDRObs(FreeSchema) begin
(Freq, Time, Signal, Modulation)::Ob
freq_at::Hom(Signal, Freq)
time_at::Hom(Signal, Time)
modulation::Hom(Signal, Modulation)
# Attributes
(FreqHz, TimeNs, Complex, ModType)::AttrType
frequency::Attr(Freq, FreqHz)
timestamp::Attr(Time, TimeNs)
iq_sample::Attr(Signal, Complex)
mod_type::Attr(Modulation, ModType)
end
# Navigate with Specter-style paths
select-all(sdr_obs, [:Signal :freq_at :frequency])
Each refinement cycle produces exactly 3 skills (GF(3) conservation):
Refinement N:
🔴 MINUS (-1): One verification/analysis skill
🟢 ERGODIC (0): One coordination/integration skill
🔵 PLUS (+1): One generation/exploration skill
Sum: (-1) + (0) + (+1) = 0 ✓ CONSERVED
Cycle 1: Basic SDR
sdr-signal-verify — Validate received signal integritysdr-frequency-coordinator — Manage spectrum allocationsdr-modulation-generator — Create novel modulation schemesCycle 2: Borges Integration
borges-catalog-index — Compress signal libraryborges-library-walker — Navigate infinite spectrumborges-signal-discovery — Find unexpected transmissionsCycle 3: Reafference Learning
reafference-error-analyzer — Measure self vs worldreafference-comparator — Efference/reafference matchingreafference-model-learner — Update world model from errors#!/usr/bin/env python3
"""
sdr_borges_reafference.py
GNU Radio flowgraph with reafference-guided tuning
"""
from gnuradio import gr, blocks, analog, filter
from gnuradio.filter import firdes
import numpy as np
class ReafferenceSDR(gr.top_block):
def __init__(self, center_freq=100e6, sample_rate=2e6):
gr.top_block.__init__(self, "Reafference SDR")
# SDR source (simulated for skill definition)
self.source = analog.sig_source_c(
sample_rate, analog.GR_COS_WAVE, 0, 1, 0
)
# Spectral analysis for reafference
self.fft_size = 1024
self.spectral_gap = 0.25 # 1/4 optimal mixing
# Efference copy buffer
self.efference = np.zeros(self.fft_size, dtype=complex)
# Reafference comparator
self.reafference_error = 0.0
def compute_reafference(self, received_spectrum):
"""Compare received spectrum to efference copy"""
self.reafference_error = np.mean(
np.abs(received_spectrum - self.efference)
)
return self.reafference_error
def random_walk_step(self, current_freq):
"""Spectral-gap-guided random walk"""
# Mixing time τ = 1/gap = 4 for gap=1/4
mixing_time = 1.0 / self.spectral_gap
# Step size inversely proportional to mixing time
step_size = self.sample_rate / mixing_time
# Reafference-weighted step
if self.reafference_error < 0.1:
# Self-caused: small exploitation step
delta = np.random.normal(0, step_size * 0.1)
else:
# World-caused: large exploration step
delta = np.random.normal(0, step_size)
return current_freq + delta
;; Borges REPL for SDR exploration
;; Each command is a step in the infinite library
(define (borges-repl sdr-state)
(display "📻 Borges Radio> ")
(let ((cmd (read)))
(case (car cmd)
;; MINUS operations (verification)
((verify) (verify-signal sdr-state (cadr cmd)))
((analyze) (analyze-spectrum sdr-state))
((compress) (kolmogorov-compress sdr-state))
;; ERGODIC operations (coordination)
((tune) (tune-frequency sdr-state (cadr cmd)))
((mix) (apply-spectral-gap sdr-state 0.25))
((walk) (reafference-walk sdr-state (cadr cmd)))
;; PLUS operations (generation)
((generate) (generate-signal sdr-state (cdr cmd)))
((explore) (borges-explore sdr-state))
((discover) (discover-unknown sdr-state))
;; Meta-operations
((refine) (spawn-3-skills sdr-state))
((help) (display-borges-help))
((quit) 'done)
(else (display "Unknown command. The library is vast.\n"))))
(unless (eq? cmd 'quit)
(borges-repl (update-state sdr-state cmd))))
(define (spawn-3-skills state)
"Generate exactly 3 new skills (GF(3) conservation)"
(let ((seed (state-seed state)))
(list
(create-skill 'minus (splitmix-next seed 0))
(create-skill 'ergodic (splitmix-next seed 1))
(create-skill 'plus (splitmix-next seed 2)))))
┌─────────────────────────────────────────────────────────────────┐
│ SDR-BORGES-REAFFERENCE TRIAD │
├─────────────────────────────────────────────────────────────────┤
│ │
│ 🔴 MINUS (-1): spectrum-analyzer │
│ └─ Kolmogorov compression of signal library │
│ └─ Spectral gap measurement via eigenvalue estimation │
│ │
│ 🟢 ERGODIC (0): sdr-borges-reafference (THIS SKILL) │
│ └─ Coordinates MINUS and PLUS via reafference loop │
│ └─ Maintains maximally mixed state for agency │
│ │
│ 🔵 PLUS (+1): signal-generator │
│ └─ Creates novel transmissions from Borges exploration │
│ └─ Random walks with spectral gap 1/4 │
│ │
│ Sum: (-1) + (0) + (+1) = 0 ✓ │
└─────────────────────────────────────────────────────────────────┘
# Start Borges SDR REPL
bb sdr-borges-repl.bb
# Invoke spectral gap analysis
amp sdr-borges-reafference --analyze-gap
# Generate 3 new skills from current state
amp sdr-borges-reafference --refine
# Reafference walk across spectrum
amp sdr-borges-reafference --walk --steps 100 --gap 0.25
TRIT_STREAM: [+1, -1, -1, +1, +1, +1, +1]
SPECTRAL_GAP: 1/4 (optimal for GF(3))
MIXING_TIME: 4 steps
MAXIMALLY_MIXED: ρ = I/3, S = log(3)
GF(3)_SUM: 0 ✓ CONSERVED
"In the Library of Babel, every signal has already been transmitted. We need only learn to tune."