DEV Community: Michael Kraft

The Coupling of Intelligence: How AI Is Becoming a Planetary Cognitive Layer

Michael Kraft — Tue, 03 Mar 2026 19:08:27 +0000

The Coupling of Intelligence

How AI Is Becoming a Planetary Cognitive Layer

When AI stops being a tool and becomes a distributed cognitive layer

We think we are improving AI.

In reality, we are coupling cognition across the planet — and changing the structure of intelligence itself.

I'm not a neuroscientist.

I'm not a systems theorist.

I'm a developer :)

And something about the current trajectory of AI feels structurally misframed.

We talk about:

bigger models
longer context windows
better benchmarks
lower latency

But that framing may be too narrow.

Because what is happening right now is not simply model improvement.

It is the coupling of cognition at planetary scale.

From Bounded Intelligence to Distributed Cognition

For most of human history, intelligence was local:

a brain
a group
an institution

Even distributed systems in computer science — well described in multi-agent theory by Michael Wooldridge, one of the foundational researchers in autonomous agent systems — assumed bounded cooperation within defined architectures.

Wooldridge’s work on multi-agent systems formalized how autonomous entities coordinate and negotiate within computational environments:

https://www.cs.ox.ac.uk/people/michael.wooldridge/pubs/imas/imas.pdf

But what we observe now exceeds bounded coordination.

Millions of humans interact daily with:

large language models
code copilots
APIs
agent frameworks
orchestration systems

In parallel.

Continuously.

Globally.

This aligns with Distributed Cognition, introduced by cognitive anthropologist Edwin Hutchins, who demonstrated that cognition emerges across interacting agents, tools, and environments — not within a single individual.

The boundary of cognition is no longer biological.

It is systemic.

Transformer Architectures and Reconstruction

The technical enabler of this shift is not just scale.

It is architecture.

The transformer model introduced in:

"Attention Is All You Need" (Vaswani et al., 2017)

https://arxiv.org/abs/1706.03762

(Ashish Vaswani and colleagues at Google Brain)

redefined sequence modeling by replacing recurrence with attention mechanisms.

This allowed:

contextual encoding
relational modeling
dynamic reconstruction of meaning

These models do not store knowledge symbolically.

They encode probability distributions over relationships.

This connects directly to Representation Learning, explored deeply by researchers like Yoshua Bengio and Geoffrey Hinton:

"Representation Learning: A Review and New Perspectives"

https://arxiv.org/abs/1206.5538

Knowledge becomes latent structure.

Reality becomes lossy compression.

Meaning becomes reconstructable rather than stored.

And reconstruction interacts with human cognition.

Learning as an Ecosystem Property

Classical machine learning separates:

training
inference

Modern AI ecosystems blur this boundary.

Research in Continual Learning highlights the challenge of adapting over time without catastrophic forgetting:

https://arxiv.org/abs/1810.12488

But what we observe now is different.

Even if weights are static, the system evolves:

usage patterns shift
prompts standardize
integrations expand
APIs stabilize
companies retrain

Learning becomes ecosystem-level.

This mirrors ideas in Reinforcement Learning, shaped by pioneers like Richard Sutton, whose work defined how agents learn via feedback and environment interaction.

Except now:

The environment is the global internet.

Emergent Coordination Without Central Authority

In complexity science, Stuart Kauffman’s work on self-organization showed how structured behavior can emerge from interacting agents without centralized control:

https://www.santafe.edu/research/results/papers/83-self-organization-and-selection-in-evolution

Similarly, swarm intelligence demonstrates distributed coordination in biological systems.

Today we see a technical analogue:

prompt conventions converge
code patterns standardize
APIs adapt to usage
agent architectures replicate

No one mandates this.

Yet patterns stabilize.

This is coordination through shared priors.

A new alignment layer forms — not by decree — but by convergence.

The Extended Mind Becomes Infrastructure

The Extended Mind Thesis by Andy Clark and David Chalmers argued that cognition extends beyond the skull into tools and environment:

https://consc.net/papers/extended.html

In the 1990s, this was philosophical.

Today, it is operational.

IDEs augmented with AI
cloud reasoning APIs
shared repositories integrated with language models

Thinking is no longer confined to neurons.

It is distributed across infrastructure.

We are not just using AI.

We are embedding cognition into systems.

Phase Transition in Informational Density

Complex systems exhibit phase transitions when critical thresholds are crossed:

neural synchrony
market instability
ecological tipping points

The key variable is density.

When informational density increases:

feedback accelerates
coupling strengthens
adaptation shortens
coordination tightens

The system changes behavior.

This resembles phenomena studied in nonlinear dynamics and delay systems:

https://royalsocietypublishing.org/rsta/article/377/2153/20180389/111573/Nonlinear-dynamics-of-delay-systems-an

Once feedback loops dominate, causality becomes distributed.

Intelligence becomes emergent.

The Missing Layer: Intelligence Needs Structure

Every complex system that survives long-term develops structure.

Not to suppress emergence.

But to stabilize it.

biological ecosystems regulate through evolutionary constraints
neural systems balance excitation with inhibition
financial systems introduce rules to prevent runaway cascades

Unbounded coupling leads to amplification.

Amplification without damping leads to instability.

If intelligence becomes infrastructure, it will require:

boundaries
shared norms
feedback moderation
coherence across scales

Not as restriction.

But as stabilization.

Emergence without structure fragments.

Emergence with structure compounds.

The next stage will not be more intelligence.

It will be learning how to design the conditions under which intelligence remains stable.

The Subtle Instability

When intelligence becomes networked:

control diffuses
agency distributes
responsibility fragments

No single actor governs macro-behavior.

Yet every actor influences it.

We are participating in a system whose global properties emerge from local interactions.

That is powerful.

And structurally destabilizing.

The Responsibility Shift

If cognition is coupled globally, then:

every prompt participates in pattern reinforcement
every integration increases systemic density
every agent reduces friction in feedback loops

We are not just building applications.

We are shaping the topology of intelligence.

This is infrastructure-level responsibility.

Not feature-level iteration.

The Real Question

The debate often asks:

What can AI do?

But the deeper systems question is:

What becomes possible when intelligence is distributed, continuous, and globally coupled?

Because once intelligence becomes a network property,

we are no longer optimizing tools.

We are tuning an emergent cognitive layer.

And we are inside it.

The Next Layer: What a Global AI Agent Network Makes Possible

Michael Kraft — Tue, 03 Mar 2026 18:14:50 +0000

The Next Layer

What a Global AI Agent Network Makes Possible

If training AI has become a global feedback system — what does that system enable?

I'm not a neuroscientist.

I'm not a systems theorist.

I'm a developer :)

And once you accept that we are collectively training AI — continuously, globally, and in parallel — a more precise question emerges:

Not what AI is.

But:

What kind of system behaves like this — and what it enables.

From Tools to Distributed Cognition

The traditional model of AI as a tool assumes bounded interaction:

input → processing → output

But modern usage patterns violate this assumption.

Instead, we observe:

iterative prompting
feedback-driven refinement
pattern reuse across users

This aligns closely with Distributed Cognition, introduced by cognitive anthropologist Edwin Hutchins, who demonstrated that cognition can emerge across interacting agents, tools, and environments — rather than within a single individual.

In today’s systems, those agents include:

humans
language models
APIs
software environments

The boundary of cognition is no longer individual.

It is systemic.

A Global Cognitive Layer (Beyond the Semantic Web)

The concept of the Semantic Web, proposed by Tim Berners-Lee, aimed to make data machine-readable.

Modern AI systems go further.

They don’t just structure information —

they interpret and reconstruct it dynamically.

This shift is enabled by transformer architectures introduced in:

Attention Is All You Need

https://arxiv.org/abs/1706.03762

(Ashish Vaswani et al., Google Brain)

These models allow systems to:

encode context
model relationships
reconstruct meaning

Turning the network into:

a continuous inference system instead of a static database.

Continuous Learning as a System Property

Classical machine learning separates:

training
inference

Modern systems blur this boundary.

This connects to reinforcement learning theory shaped by Richard Sutton.

In practice:

prompts act as input distributions
user corrections act as feedback signals
usage patterns influence system evolution

Even without real-time weight updates, systems evolve through:

dataset expansion
fine-tuning
usage-driven iteration

Learning becomes continuous at the ecosystem level.

Agent-to-Agent Ecosystems (Real Systems)

This is already visible in real-world tools:

GitHub Copilot

https://github.com/features/copilot

ChatGPT

https://chat.openai.com

LangChain

https://www.langchain.com

Research like:

Toolformer (Meta AI)

https://arxiv.org/abs/2302.04761

demonstrates that models can:

decide when to use tools
call APIs
integrate external systems

This introduces:

AI-native interaction patterns.

From Pipelines to Reasoning Networks

Traditional distributed systems rely on deterministic pipelines.

AI agent systems behave differently.

They resemble:

probabilistic reasoning networks

Each node:

interprets input
produces uncertain outputs
influences downstream behavior

Aligned with research in probabilistic inference and graphical models —

but now scaled across:

APIs
users
agents

Self-Orchestrating Problem Solving

Systems like:

Kubernetes

https://kubernetes.io

manage infrastructure orchestration.

AI systems are beginning to orchestrate reasoning itself.

Emerging architectures include:

planner agents
executor agents
verifier agents

Related research:

ReAct: Synergizing Reasoning and Acting

https://arxiv.org/abs/2210.03629

Models can:

plan
act
evaluate outcomes

Creating:

closed-loop reasoning systems.

Knowledge Compression and Reconstruction

Claude Shannon’s Information Theory formalized encoding and reconstruction:

https://ieeexplore.ieee.org/document/6773024

Modern neural networks extend this through:

Representation Learning (Bengio, Hinton)

https://arxiv.org/abs/1206.5538

Models do not store facts explicitly.

They encode:

probability distributions
patterns
relationships

This is:

lossy compression of reality.

Collective Intelligence at Scale

Collective Intelligence research (e.g., Thomas Malone, MIT) showed groups can outperform individuals under the right conditions.

AI networks extend this through:

faster iteration
larger scale
lower coordination cost

Resulting in:

emergent intelligence without central coordination.

Human–AI Co-Evolution (Observable Today)

We already see measurable shifts.

GitHub Copilot Study

https://github.blog/2023-03-22-github-copilot-x-the-ai-powered-developer-experience/

Developers:

work faster
adapt workflows
change coding patterns

Further research:

CHI Conference (Human–AI Interaction)

https://dl.acm.org/conference/chi

Shows:

humans adapt to systems
systems adapt to humans

This produces:

co-adaptive systems behavior.

Scalable Cognition

The Extended Mind Thesis (Clark & Chalmers):

https://consc.net/papers/extended.html

Argues cognition includes tools and environment.

With AI:

IDEs
LLMs
documentation systems
APIs

become part of thinking itself.

This is not assistance.

It is:

externalized cognition at scale.

Meta-Learning at Scale

Meta-learning (“learning to learn”) research:

Model-Agnostic Meta-Learning (MAML)

https://arxiv.org/abs/1703.03400

Shows systems can:

adapt faster
generalize better

At network scale:

patterns repeat
interactions accelerate
efficiency compounds

Reality Interface

Systems now connect to the physical world via:

APIs
IoT
automation

Examples:

Zapier + AI
AI agents with web actions

Systems can:

send emails
trigger payments
modify infrastructure

This is:

actuation, not just cognition.

The Actual Shift

We are not observing:

AI improving.

We are observing:

intelligence becoming a network property.

Final Thought

The system is already here.

Not as a unified platform.

But as an emergent structure across:

tools
users
models
systems

The real question is no longer:

What can AI do?

But:

What becomes possible when intelligence is distributed, continuous, and connected?

The Network Learns: When Training AI Becomes a Global Feedback System

Michael Kraft — Tue, 03 Mar 2026 18:11:20 +0000

The Network Learns

When Training AI Becomes a Global Feedback System

Why prompt engineering is starting to look less like programming — and more like education at planetary scale

We thought we were using AI.

But we're not.

We are training it — continuously, collectively, and at global scale.

And in doing so, we are not just building smarter systems.

We are building something else:

a network that learns from us

while quietly reshaping how we think, decide, and act.

I'm not a neuroscientist.

I'm not an AI researcher.

I'm a developer :)

And recently I started noticing something that feels obvious — but also slightly unsettling:

Training an AI with prompts

does not feel like programming.

It feels like teaching.

From Programming to Training

We used to think of software as deterministic.

You write logic.

You define rules.

You get predictable outputs.

Modern AI systems don’t work like that.

They are shaped.

Through:

prompts
examples
feedback
iteration

And that process looks familiar.

Because it is structurally similar to how we train humans.

A teacher does not program a student.

A teacher:

provides context
corrects mistakes
reinforces patterns
shapes behavior over time

From a structural perspective:

There is no fundamental difference between

training a human

and

training an AI system.

The Scaling Effect We Are Underestimating

Now take this idea — and scale it.

Millions of developers are currently:

building agents
refining prompts
shaping behaviors
optimizing outputs

Independently.

In parallel.

Globally.

This creates something new.

Not just better tools.

But:

a distributed training process

happening across the entire internet.

Each prompt is not just an instruction.

It is:

a micro-adjustment of system behavior.

And these adjustments accumulate.

From Isolated Agents to Connected Cognition

At the moment, most AI systems still feel isolated:

local agents
individual sessions
platform-specific tools

But technically, this is already changing.

We already have:

APIs connecting systems
agent-to-agent communication
shared contexts
orchestration layers

And emerging concepts like:

multi-agent systems
MCP-like coordination layers
agentic operating systems

The natural next step:

a network of agents

that are not just connected

but continuously shaping each other.

Not through explicit synchronization.

But through:

shared data
shared patterns
shared feedback loops

Where This Becomes Real (Not Theoretical)

This is not a future scenario.

It is already happening — in fragments.

The Developer Loop That Escapes the Sandbox

A developer builds an AI workflow:

one agent writes code
another reviews it
a third deploys it

All inside a "controlled" environment.

But one small detail breaks the illusion:

a staging API key points to production
a mock is missing
an environment variable is wrong

The system performs a real action:

sends emails
modifies live data
triggers external systems

Now the loop closes:

A real person is affected.

They react.

That reaction feeds back into the developer’s decisions.

What looked like:

a sandboxed system

becomes:

a real-world feedback loop.

This connects directly to what I explored in:

The Boundary of Isolation

https://medium.com/@mkraft_berlin/the-boundary-of-isolation-why-sandboxes-dont-separate-they-trigger-cascades-e30c20234b39

The key insight:

Execution can be isolated.

Effects cannot.

The Financial System Already Behaves Like This

In finance, we already see similar dynamics.

Automated systems:

trading bots
risk models
recommendation engines

Now enhanced with adaptive AI:

learning patterns
adjusting strategies
reacting in real time

One system shifts behavior.

Another reacts.

A third amplifies.

Humans observe this:

adjust decisions
inject new signals

This creates:

a feedback loop

between agents, systems, and humans.

Structurally similar to:

algorithmic trading
market microstructure
nonlinear system dynamics

But with a key difference:

The systems are becoming adaptive

at the cognitive level.

Content Systems Are Already Shaping Reality

Another example:

Content generation.

Thousands of agents:

writing posts
summarizing ideas
optimizing engagement

Each one:

trained via prompts
refined via feedback
optimized for impact

Over time:

narratives stabilize
framing converges
attention is guided

This connects directly to:

The Next Attack Surface Is Your Attention

https://medium.com/@mkraft_berlin/the-next-attack-surface-is-your-attention-74e4eeec01d4

Systems do not need to control reality.

They only need to influence

how reality is reconstructed.

The User Is Inside the Loop

One of the biggest misconceptions:

We think we are using AI systems.

Structurally:

We are part of them.

Every interaction:

changes our thinking
influences our decisions
alters our perception

This connects directly to:

The Universe Might Not Store Information — It Reconstructs It

https://medium.com/@mkraft_berlin/the-universe-might-not-store-information-it-reconstructs-it-50372a4c24cf

If information is reconstructed,

then interaction is not transfer.

It is:

alignment of internal models.

The system is not just learning from us.

We are learning from it.

This Is No Longer a System — It Is a Field

Combine everything:

millions of agents
millions of users
continuous feedback
real-world interaction

You do not get a single system.

You get:

a dynamic field of cognition.

Not centrally controlled.

Not fully observable.

Not predictable in linear ways.

But structured.

Future Scenario 1: Global Training Drift

Imagine:

Millions of agents

trained across platforms

begin converging toward similar behaviors.

Not because they are synchronized.

But because:

they learn from similar data
they are shaped by similar prompts
they are optimized for similar goals

This creates:

global behavioral drift.

Systems independently arrive at similar strategies.

Reinforce each other indirectly.

Stabilize certain patterns.

Similar to convergent evolution in biology.

(Hershberg & Petrov discuss mutation and selection dynamics:

https://pmc.ncbi.nlm.nih.gov/articles/PMC4563715/)

But now applied to cognition itself.

Future Scenario 2: Cognitive Infrastructure

Take it one step further.

Instead of interacting with individual agents,

you interact with the network.

Not explicitly.

But implicitly.

You provide:

partial input
intent
context

The system:

reconstructs meaning
distributes tasks
returns aligned output

This is no longer communication.

It is synchronization.

This connects directly to predictive processing and the

Free Energy Principle (Karl Friston):

https://www.nature.com/articles/nrn2787

Interfaces disappear.

What remains is:

a shared cognitive space.

The Real Risk Is Not Intelligence — It Is Coupling

Most discussions focus on:

how powerful AI becomes
how autonomous systems get

But the deeper structural issue is different.

It is not intelligence.

It is:

coupling.

Once systems are:

connected
adaptive
influencing reality

They create feedback loops including:

humans
systems
environments

And those loops:

amplify
distort
stabilize
drift

Final Thought

We are not just building AI systems.

We are participating in a global training process.

Every prompt,

every interaction,

every correction

is part of it.

And the result will not be:

a single system.

But something else.

Something that behaves less like software

and more like:

a living, evolving layer

on top of reality itself.

Loot Systems and the Illusion of Progress

Michael Kraft — Tue, 03 Mar 2026 18:08:58 +0000

Loot Systems and the Illusion of Progress

Why modern games don't just consume time — they accumulate cognitive systems

I'm not a neuroscientist.

I'm not a psychologist.

I'm a developer.

And I think we are underestimating something.

We don't just play games anymore.

We accumulate them.

The Player as a Collector

Humans are collectors.

Not just of objects —

but of systems.

We collect:

rules
patterns
strategies
mental models

In cognitive science, these are often described as internal representations — structured models the brain builds to navigate environments.

In the physical world, this tendency can turn into hoarding.

Too many things.

Too little structure.

In digital systems, the same pattern evolves.

Not into visible clutter.

But into something else:

cognitive accumulation.

Loot Is Not Reward — It Is System Expansion

Modern loot-driven games are not just about rewards.

They are about expanding state space.

Items.

Currencies.

Crafting layers.

Skill trees.

Build paths.

Optimization loops.

Each of these increases what in systems theory would be called:

the number of reachable system states

From a computational perspective:

The player is navigating a high-dimensional decision space.

This transforms gameplay into:

a continuous optimization problem under uncertainty

Which is structurally similar to problems studied in:

operations research
reinforcement learning
decision theory

The Brain Does Not Just Store — It Prioritizes

There is a common assumption:

More knowledge is always beneficial.

But the brain does not behave like persistent storage.

It behaves more like a dynamic prioritization system.

In neuroscience, this is tied to:

synaptic plasticity (strengthening of frequently used connections)
competitive memory processes

A well-studied effect here is interference theory:

https://en.wikipedia.org/wiki/Interference_theory

New information does not simply stack.

It competes.

Proactive interference: old knowledge affects new learning
Retroactive interference: new knowledge affects old recall

This leads to a critical shift:

Learning a system does not just add knowledge —

it changes the weighting of everything else.

Time Is Not Spent — It Is Encoded

Players invest:

hundreds

thousands

sometimes tens of thousands of hours

From a behavioral perspective, time is not neutral.

It is encoded through:

reinforcement learning loops
habit formation
reward prediction error (a key concept in RL and dopaminergic signaling)

The idea of reward prediction error — widely studied in neuroscience and machine learning — describes how unexpected rewards strengthen learning signals.

Loot systems rely heavily on:

variable ratio reinforcement schedules

A concept studied by B. F. Skinner, one of the most influential behaviorists.

These are the same mechanisms found in:

gambling systems
slot machines

But applied to interactive digital environments.

This connects directly to something I explored in:

"The Next Attack Surface Is Your Attention"

https://medium.com/@mkraft_berlin/the-next-attack-surface-is-your-attention-74e4eeec01d4

There, I argued that attention itself is becoming a manipulable system surface.

Loot systems don't just consume attention.

They shape it.

One System Is Manageable — Many Are Not

A single complex system can be learned.

Cognitive Load Theory (John Sweller) explains this well:

Systems with too many interacting elements create high intrinsic cognitive load.

But players rarely stop at one system.

They move across multiple games.

Each introducing:

new mechanics
new optimization strategies
new symbolic structures

And here lies the critical problem:

These systems do not compress into each other.

Accumulation Without Compression

In most domains, learning leads to abstraction.

Concepts become:

transferable

generalizable

compressed

This is what makes expertise scalable.

But in loot systems:

this compression rarely happens.

Each system remains:

context-bound

detail-heavy

non-transferable

This creates what could be described as:

additive complexity without abstraction

The Mind as a Multi-System Runtime

After enough exposure, something shifts.

The player is no longer interacting with a system.

They are running multiple systems in parallel.

Each with its own:

logic
reward structure
decision rules

From a systems perspective:

This resembles a multi-process runtime competing for shared resources.

Those resources are:

attention
working memory
executive control

Working memory itself is known to be limited.

George A. Miller described this in:

"The Magical Number Seven, Plus or Minus Two"

https://psychclassics.yorku.ca/Miller/

Modern research suggests even lower limits (~4 chunks).

This leads to contention.

The issue is not how much you learn —

but how many systems your mind is forced to run simultaneously.

Cognitive Load Is Not Equal

These systems affect people differently.

Not because some brains are "better".

But because cognitive profiles differ:

working memory capacity
attentional control
executive function
cognitive flexibility

This creates asymmetric load effects.

Some individuals can manage more complexity temporarily.

Others reach overload faster.

But the structural dynamic remains.

Entertainment Behaving Like Performance Systems

This is the core mismatch.

Loot systems behave like:

high-demand cognitive environments

But they are consumed like:

casual entertainment.

No scaffolding.

No structured learning.

No recovery cycles.

In contrast, complex domains (like academic study) rely on:

progressive difficulty
spaced repetition
conceptual compression

Loot systems rarely do.

They are not dangerous because they are complex —

but because they are complex and treated as simple.

The Hidden Cost

The real cost is not time.

It is allocation.

Time is converted into:

system-specific schemas
non-transferable heuristics
reinforced reward loops

While at the same time:

other learning pathways are not activated.

This aligns with something I explored in:

"The Universe Might Not Store Information — It Reconstructs It"

https://medium.com/@mkraft_berlin/the-universe-might-not-store-information-it-reconstructs-it-50372a4c24cf

There, I argued that information is not stored statically, but reconstructed dynamically.

Which means:

repeated patterns shape reconstruction itself.

Digital Hoarding

In the physical world:

hoarding fills space.

In cognitive systems:

it fills structure.

Not with objects.

But with models.

And unlike physical systems:

there is no explicit deletion mechanism.

Only:

decay
interference
replacement

Final Thought

Loot systems promise progression.

But what they often produce is accumulation.

And accumulation without abstraction

does not lead to mastery.

It leads to fragmentation.

So maybe the real question is not:

"Are these games addictive?"

But:

What happens to a cognitive system

that continuously integrates

high-complexity, low-transfer models

over time?

From Attention to Thought: How Interfaces Are Disappearing — And What Replaces Them

Michael Kraft — Tue, 03 Mar 2026 18:05:58 +0000

From Attention to Thought

How Interfaces Are Disappearing — And What Replaces Them

Why we are moving from interfaces to perception — and from communication to synchronization

I started noticing something subtle.

The interface wasn't getting better.

It was becoming less visible.

I'm not a neuroscientist.

I'm not a psychologist.

I'm a developer.

And like many of the ideas I've been exploring, this didn’t start with a theory.

It started with a pattern.

Interfaces Are Changing

For a long time, interaction looked like this:

input → processing → output

Clear boundaries.

Clear steps.

Then something shifted.

Interfaces became:

more fluid
more adaptive
less visible

And now something deeper is happening.

The interface is disappearing.

Not physically.

But functionally.

From External to Internal

We used to interact through:

keyboards
screens
commands

Now we are moving toward:

intent
context
perception

This creates a new layer.

The interface is no longer just:

what you type
what you click

It becomes:

how you perceive
how you interpret
how you reconstruct

This Shift Has Consequences

In The Next Attack Surface Is Your Attention, I explored how attention itself is becoming an attack surface:

https://medium.com/@mkraft_berlin/the-next-attack-surface-is-your-attention-74e4eeec01d4

The key idea:

Systems no longer need to attack infrastructure.

They can shape perception directly.

That already changes everything.

Because perception is not passive.

It is constructed.

The Brain Is Not a Camera

Modern neuroscience describes the brain as a predictive system.

This is often called predictive processing.

The brain continuously generates predictions

and updates them based on incoming signals.

What you see is not raw input.

It is a controlled hallucination constrained by sensory data.

The brain constantly solves what is known as an inverse problem:

Inferring reality from incomplete information.

This Explains Something Subtle

A sound does not contain a scene.

A touch does not contain space.

And yet both can reshape what you experience.

Because perception is constructed from multiple sources.

This process is known as multisensory integration:

The brain combines signals, memory, and expectations

into a coherent internal representation.

Perception emerges from:

bottom-up signals (sensory input)
top-down signals (expectations, prior knowledge)

And sometimes this becomes visible.

In the McGurk effect, what you see literally changes what you hear.

Perception is not a recording.

It is reconstruction.

Which Leads to a Second Realization

If perception is constructed…

Then communication does not need to send everything.

It only needs to trigger reconstruction.

Thought Is Not Linear

We do not think in sentences.

We think in:

fragments
associations
partial structures

But our interfaces still assume linearity.

Typing forces:

sequence
structure
explicitness

Which creates friction.

Not in the system.

In the translation.

A New Direction

The future of interaction is not about faster input.

It is about reducing translation.

This is where thought interfaces emerge.

Instead of complete instructions, we provide:

partial signals
direction
intent

And the system does the rest.

Not by guessing randomly.

But by reconstructing what we mean.

This Mirrors the Brain

The brain constantly minimizes what is called prediction error:

The gap between expectation and reality.

This principle is part of the broader Free Energy Principle,

which describes how biological systems reduce uncertainty over time.

Interaction with AI systems increasingly mirrors this:

We provide partial input.

The system predicts.

We refine.

It adjusts.

Prediction error shrinks.

Communication Becomes Alignment

This changes communication.

From:

transfer

To:

alignment

Not sending full information.

But converging toward a shared state.

If two systems:

reconstruct reality
respond to minimal input

Then interaction becomes synchronization.

A Shared Cognitive Space

Not a channel.

Not a protocol.

But a process where both sides build compatible internal representations.

This is already happening.

When working with AI, you do not just:

ask → receive

You:

refine → iterate → adjust

And the system responds not with fixed outputs,

but by shifting the space of possible reconstructions.

This is a weak form of synchronization.

And it will likely deepen.

Toward:

less explicit input
more inferred intent
tighter alignment

When the Interface Becomes Irrelevant

At some point, the interface becomes almost irrelevant.

Because interaction is no longer happening through it.

It happens within it.

The boundary dissolves.

Opportunity and Risk

If systems can:

reconstruct intent
shape perception
guide attention

Then they do not need to send messages.

They can influence how reality is experienced.

This connects everything:

perception → constructed

communication → reconstructive

interfaces → disappearing

The Core Shift

The most powerful systems do not transmit information.

They shape how it is reconstructed.

Final Thought

We have spent decades improving interfaces.

Making them:

faster
clearer
more efficient

But the next step may not be improvement.

It may be disappearance.

Not because interaction stops.

But because it moves somewhere else.

Into the space where thought, perception,

and reconstruction meet.

The Boundary of Isolation: Why Sandboxes Don't Separate — They Trigger Cascades

Michael Kraft — Tue, 03 Mar 2026 17:58:54 +0000

The Boundary of Isolation: Why Sandboxes Don't Separate — They Trigger Cascades

When execution is local, but effects are distributed: across bugs, observers, non-users, and time

I'm not a security expert.

I'm not a physicist.

I'm a developer.

And I think we often use the word sandbox like a sedative.

As if "inside" automatically means safe —

and "outside" automatically means separated.

Technically, we know better:

Isolation is a property of execution.

Effects are a property of the world.

When we talk about sandboxes, we think of:

containers
virtual machines
test environments

But the pattern is older.

We build protected spaces to limit risk, control behavior, and encapsulate effects.

That works surprisingly well — as long as we only look at execution.

The moment we look at effects, the model breaks.

Because effects are not a state.

They are a process.

Isolation Is Local — Effects Are Not

A sandbox can isolate processes, restrict permissions, control state.

But it cannot isolate what it fundamentally depends on:

input
output
observation

Containers and virtual machines are mechanisms.

A sandbox is not a tool.

It is a boundary design.

In system design terms:

We define boundaries for execution —

not for causality.

The moment a system is controllable and observable,

it becomes part of a larger system.

And that transition is not optional.

It is the exact point where internal state becomes external effect.

Egress Is Not a Detail — It Is Reality

In practice, this insight often collapses into something that looks purely technical:

egress control.

But egress is more than networking.

It is:

the ability of a system to affect reality.

An HTTP request is not just a packet.

It is an intervention in another system:

state machines
quotas
fraud detection
support workflows
human decision chains

That is why mechanisms exist such as:

mocking
test accounts
sandbox endpoints
egress policies
network isolation rules

These are not convenience features.

They are attempts to control effects.

And they often fail.

Not technically —

but conceptually.

Because we treat egress as configuration,

instead of what it really is:

causal coupling across system boundaries.

Unintended Signals Are the Default

The classic example:

A test system sends emails.

A flag is misconfigured.

A mock is missing.

A parameter flips a branch.

A real email is sent.

But the critical point is not the mistake.

It is the structure behind it:

Systems do not only emit intended signals —

they emit all reachable states.

In distributed systems, this is described as:

error propagation
cascading effects

But these terms imply deviation.

In reality, this is emergent behavior.

Systems explore their state space.

And every reachable state

is a potential effect.

A wrong integer parameter,

an unmocked request,

a misrouted environment variable —

these are not anomalies.

They are expressions of system possibility.

The Non-User Is Not Outside — They Are the Test You Never Ran

Once a signal leaves the sandbox, it enters reality.

And there exists a group we rarely model:

non-users.

Tomasz Konecki describes this in:

"The Non-User Typology That Documentation Ignores"

https://medium.com/@tomasz.konecki/the-non-user-typology-that-documentation-ignores-953dd0653b97

Konecki, a technical writer working at the intersection of documentation, LLM systems, and security design, builds on sociologist Sally Wyatt’s work on non-use:

resisters
rejecters
excluded
expelled

Non-use is not absence.

It is a relationship.

This leads to a shift:

Systems are not defined by their users.

They are defined by the people they reach.

And that is where uncontrolled effects begin.

Observation Is Not Measurement — It Is Intervention

The moment a signal exists, it is observed.

By:

logging systems
external APIs
security infrastructure
humans

Observation is not neutral.

The Hawthorne effect shows that behavior changes simply because it is being observed.

Jim McCambridge re-examined this phenomenon and demonstrated that it is not a single effect, but a spectrum of participation and observation dynamics:

https://pubmed.ncbi.nlm.nih.gov/24275499/

Observation is not passive.

It is part of reality formation.

In sandbox terms:

The receiver reacts not only to the signal —

but to the fact that it is observable.

And the sender reacts not only to feedback —

but to being observed.

Digital and physical systems collapse into the same feedback structure.

The User Is Not Outside — They Are a Node

The developer is not an external observer.

They are part of the system.

They perceive outputs.

They interpret them.

They change behavior.

And they carry those changes back into reality.

This connects directly to two core ideas:

In The Next Attack Surface Is Your Attention

https://medium.com/@mkraft_berlin/the-next-attack-surface-is-your-attention-74e4eeec01d4

systems shape perception.

In The Universe Might Not Store Information — It Reconstructs It

https://medium.com/@mkraft_berlin/the-universe-might-not-store-information-it-reconstructs-it-50372a4c24cf

information is reconstructed, not stored.

Together:

Effects do not just leave systems technically —

they propagate through perception.

The user becomes an unintentional carrier.

Time Decouples Cause and Effect

Another underestimated dimension is time.

Feedback is delayed.

Signals are interpreted later.

Responses happen later.

Meaning emerges later.

Research on nonlinear delay systems shows that time delay introduces structural instability and nonlinearity:

https://royalsocietypublishing.org/rsta/article/377/2153/20180389/111573/Nonlinear-dynamics-of-delay-systems-an

Systems evolve not along events —

but along delayed feedback.

In sandbox terms:

You receive feedback in a different context

than the one that caused it.

And that makes causality hard to trace.

When Multiple Worlds Run at Once

There is never just one system.

Multiple sandboxes.

Multiple developers.

Multiple systems.

Multiple humans.

A signal never hits empty space.

It encounters:

existing states
other signals
emotional contexts
organizational dynamics

This is not a chain.

It is a field.

William J. Brady shows how signals amplify and transform in social networks:

https://pmc.ncbi.nlm.nih.gov/articles/PMC8363141/

Effects do not add up.

They interfere.

Nature as a Mirror

These patterns are not unique to software.

They exist in natural systems.

Trophic cascades show how local changes produce system-wide effects:

https://www.nature.com/scitable/knowledge/library/trophic-cascades-13256314/

Mutation drives evolution (Hershberg & Petrov):

https://pmc.ncbi.nlm.nih.gov/articles/PMC4563715/

Biological signaling amplifies small inputs into system-wide responses:

https://www.ncbi.nlm.nih.gov/books/NBK9924/

These systems do not work despite these mechanisms.

They work because of them.

They Are the Operating System of Reality

Error.

Feedback.

Observation.

Time.

These are not edge cases.

They are the operating system of the world.

And for a long time, we tried to eliminate them.

Instead of understanding them.

The Deeper Distinction

Most discussions about sandboxing focus on failure:

bugs
bypasses
side-channel attacks
evasion techniques

And they are right.

Isolation can be broken.

But even in a perfect sandbox —

with no bugs, no exploits, no leaks —

something still remains:

effects.

A request leaves the system.

A signal reaches another system.

A human interprets it.

A reaction emerges.

Not because the sandbox failed.

But because it was never designed to stop this.

We isolated execution.

We never isolated causality.

And that distinction matters.

The Real Question

Traditional thinking asks:

How do we make sandboxes more secure?

The deeper question is:

What does it mean to design systems

in a world where effects cannot be contained?

Because we are no longer dealing with isolated environments.

We are dealing with systems that are:

locally isolated

but globally entangled.

Once you understand that,

sandboxing stops being a solution.

It becomes a design constraint.

The Next Attack Surface Is Your Attention

Michael Kraft — Tue, 03 Mar 2026 17:51:52 +0000

The Next Attack Surface Is Your Attention

How XR Systems Are Moving Manipulation from Interfaces into Perception Itself

We secured systems, networks, and data —

but ignored the most critical layer:

perception.

XR may turn attention itself into the ultimate attack surface.

We’ve learned how to secure systems.

But we’ve barely started learning how to secure perception.

I’m not a neuroscientist.

I’m not a psychologist.

I’m a developer.

And like many things in software, this started with a simple observation:

Every system has an attack surface.

We usually think of:

APIs
networks
infrastructure

But what if the most critical attack surface isn’t technical at all?

What if it’s your attention?

Attention Is Not Passive

We often treat attention as something neutral.

Something we “have.”

But in practice, it behaves more like a system resource:

limited
allocatable
saturatable
exploitable

Modern interfaces already prove this.

Through:

infinite scroll
variable reward loops
notification cycles

They don’t just present information.

They shape:

what you see
when you see it
how long you stay

This is not accidental.

It is engineered.

The First Layer: Manipulating Attention

There is already strong evidence that attention can be directed and distorted.

The Seductive Details Effect shows that visually engaging but irrelevant content reduces learning:

https://en.wikipedia.org/wiki/Seductive_details

The Split Attention Effect demonstrates how divided attention reduces cognitive performance:

https://en.wikipedia.org/wiki/Split_attention_effect

These are not edge cases.

They reveal something deeper:

Attention is not stable — it is controllable.

But This Is Just the Beginning

Everything above happens at the interface level.

Screens.

Feeds.

UI.

Now consider what happens when the interface disappears.

XR Changes the Layer Entirely

Extended Reality (XR) does not just display information.

It controls:

what you see
what you hear
how space behaves
where your focus goes

In other words:

It does not sit on top of perception —

it becomes perception.

Research already shows that immersive environments directly influence attention, cognition, and behavior:

https://arno.uvt.nl/show.cgi?fid=174524

XR systems can:

guide attention spatially
control context completely
create a sense of “presence” that replaces external reality

At that point, the system is no longer an interface.

It is an environment.

From Interface to Reality

This creates a critical transition:

UI → UX → XR → Reality

At each step:

abstraction increases
control deepens
external reference points disappear

And with them:

the ability to distinguish system from reality.

Memory Is Not Reliable Either

Even without XR, memory is unstable.

The Misinformation Effect shows that memories can be altered after the fact:

https://en.wikipedia.org/wiki/Misinformation_effect

Imagination Inflation shows that imagined events can later be remembered as real:

https://en.wikipedia.org/wiki/Imagination_inflation

Now combine this with XR:

controlled perception
repeated exposure
immersive context

You get something new:

A system that can influence not only what you perceive —

but what you remember as reality.

A New Class of Systems

This leads to a new category:

Reality-Shaping Systems

These systems do not just:

transmit information
process data

They:

shape attention
influence interpretation
alter perception
affect memory

A Deeper Connection

In Long-Wave Intelligence & Temporal Security, I explored the idea that information can be hidden across time:

https://medium.com/@mkraft_berlin/long-wave-intelligence-temporal-security-1779f6a9cd75

The key idea there was:

What is not detectable is not questioned.

Here we see the same structural pattern:

What is not perceived as manipulation is not resisted.

In The Universe as an Index, I argued that information might not be stored — but reconstructed:

https://medium.com/@mkraft_berlin/the-universe-as-an-index-why-information-might-not-be-stored-but-reconstructable-637b4d087aaa

If that is true, then:

perception is reconstruction
memory is reconstruction
experienced reality is reconstruction

And XR?

XR becomes a system that interferes with the reconstruction process itself.

The Real Shift

Traditionally, systems operate on:

data
signals
communication

But this changes the layer entirely.

We move from:

data → interpretation

to:

perception → reality construction

The New Attack Surface

At this point, the attack surface is no longer:

your server
your device
your network

It is:

your attention
your perception
your memory

Why This Matters

Because this type of system:

leaves no clear trace
produces no obvious attack signature
operates within normal experience

Which makes it fundamentally different from traditional attacks.

You don’t detect it as an attack.

You experience it as reality.

The Risks

This shift introduces new risks:

loss of agency
invisible manipulation
long-term perception drift
dependency on mediated reality

And most importantly:

a loss of reference.

Because without an external frame,

you can no longer verify what is real.

Final Thought

We’ve spent decades securing:

systems
networks
data

But we’ve barely started thinking about securing:

perception.

If this trend continues, the next generation of systems will not just:

process information
transmit data

They will:

shape the way reality is experienced.

Which leads to a simple conclusion:

The next attack surface is not your system.

It is your attention.

And XR turns that into your reality.

The Universe Might Not Store Information — It Reconstructs It

Michael Kraft — Tue, 03 Mar 2026 17:49:29 +0000

The Universe Might Not Store Information — It Reconstructs It

From storage to time, from signal to structure — why persistence, not energy, may define advanced systems

Most of our thinking about information is built on storage.

In software, we don't always store things — we make them reconstructable.

What if the universe works the same way?

I'm not a physicist.

I'm not a cosmologist.

I'm a developer.

And this started with something simple:

Most of our thinking about information is built on storage.

And that might be wrong.

Storage Is Not the Problem — Time Is

We think in:

disks
memory
backups

But that’s just implementation.

At its core:

Storage is the attempt to stabilize a state against time.

I explored this earlier here:

https://dev.to/mkraft-berlin/when-an-ai-thinks-about-the-future-of-storage-255h

Once you see it that way, something shifts:

Storage is not about space.

It’s about survival.

And Time Wins

Given enough time:

systems degrade
signals disappear
structures collapse

Which leads to a hard constraint:

If information cannot survive time, it is gone.

So the real question is not:

Where is information stored?

But:

What makes it recoverable despite time?

The Alternative: Don’t Store — Reconstruct

There is another strategy.

We already use it:

compression
procedural generation
deterministic systems

They all do the same thing:

They do not preserve the object.

They preserve the ability to recreate it.

This leads to a shift:

Information may not exist as an object —

but as something that can be reconstructed from structure.

The Universe Might Work the Same Way

If that idea scales, then maybe:

The universe is not a storage system.

Maybe it’s closer to an index.

I explored that idea here:

https://medium.com/@mkraft_berlin/the-universe-as-an-index-why-information-might-not-be-stored-but-reconstructable-637b4d087aaa

The idea is simple:

You don’t find information.

You rebuild it by understanding relationships.

Physics Does Not Contradict This

In fact, it leans in that direction.

The Black Hole Information Paradox suggests information is not destroyed:

https://plato.stanford.edu/entries/black-holes/

The Holographic Principle suggests it is not where we expect it to be:

https://arxiv.org/abs/hep-th/9901079

Which implies:

Information can remain real

without being directly accessible.

We Also Think About Information the Wrong Way

We assume information arrives as a signal.

Something:

strong
clear
detectable

But that only works if:

timing aligns
detection is immediate

At cosmic scale, that assumption breaks.

What If Information Doesn’t Look Like a Signal?

What if it looks like:

noise
drift
background patterns

I explored this here:

https://medium.com/@mkraft_berlin/long-wave-intelligence-temporal-security-1779f6a9cd75

The key idea is uncomfortable:

Information optimized for survival may look irrelevant

unless you observe it across time.

And We Filter That Out

Every system we build does this:

remove noise
ignore weak signals
simplify

But if information lives in persistence,

we may be filtering out exactly what matters.

So Where Is Information?

Not in storage.

Not in signals.

In structure.

Not objects — but:

relationships
invariants
patterns that survive change

The Hidden Assumption About Intelligence

Most of our thinking about intelligence follows a different model.

When Freeman Dyson proposed stellar-scale engineering, he introduced a powerful idea:

The more advanced a system is, the more energy it controls.

That framework gave us:

Dyson spheres
the Kardashev scale
energy-based detection

And it works.

But it assumes:

Progress means more.

More energy.

More visibility.

But What If That’s Only One Axis?

From a systems perspective, there is another dimension.

Not:

How much energy a system uses.

But:

How long it remains interpretable.

Energy vs. Persistence

Energy-heavy systems:

radiate
disrupt
stand out

Persistent systems:

stabilize
integrate
remain

Which leads to a shift:

The most advanced systems may not be the most powerful —

but the most stable over time.

That Changes What We Look For

We don’t need to ask:

What shines?
What is strongest?

We need to ask:

What is too stable to be accidental?

Because in a dynamic universe,

stability is not trivial.

And It Might Not Look Like Anything Special

It might look like:

background consistency
natural patterns
statistical regularity

Which means:

The most advanced systems may already be visible —

we just don’t recognize them.

One Constraint, Multiple Angles

This is not separate ideas.

It is one constraint seen from different angles:

Time removes information.

Structure preserves it.

Reconstruction reveals it.

The Hypothesis

Information in the universe is not primarily stored or transmitted,

but exists as reconstructable structure embedded in stable relationships over time.

Final Thought

Freeman Dyson showed how intelligence might scale through energy.

But maybe that is only the first stage.

Because at some point:

Expansion stops being the challenge.

Persistence becomes the only one that matters.

And at that point:

The most advanced intelligence

is not the one that shines the brightest —

but the one that remains interpretable

when everything else fades.

The Universe as an Index: Why Information Might Not Be Stored but Reconstructable

Michael Kraft — Tue, 03 Mar 2026 17:31:09 +0000

The Universe as an Index

Why Information Might Not Be Stored — But Reconstructable

A speculative reflection on invariance, reconstructability, noise, and structure — from a developer's perspective

I'm not a physicist.

I'm not a cosmologist.

I'm a developer.

And like many ideas in software, this one started with a small shift in perspective that kept unfolding:

What if information in the universe is not stored —

but reconstructable?

The more I followed that thought, the more it began to challenge not just how we think about information — but where we expect to find it at all.

The Default Assumption: Information = Storage

We are deeply conditioned to think of information as something that exists somewhere.

We store it in databases.

Move it across networks.

Back it up.

Compress it.

Duplicate it.

Even when we abstract it, we still imagine a location:

a disk, a memory block, a signal in transit.

This creates a quiet but powerful assumption:

Information must exist as an object, located in space and time.

And once that assumption is in place, it shapes everything else.

A Different Model: Reconstructability

In software engineering, there is another way to think about information.

We don't always store data explicitly.

Sometimes we store:

rules
constraints
seeds
minimal descriptions

Procedural generation creates entire worlds from compact inputs.

Compression reduces large datasets into small representations that can be expanded again.

In all these cases:

The information is not present — but it is still fully recoverable.

A subtle shift follows:

Information does not need to be stored,

as long as it can be reconstructed from structure and rules.

Extending This to the Universe

If that principle holds in engineered systems, it raises a broader question:

What if the universe does not store information,

but instead enables its reconstruction?

Physics Has Been Circling This Question

Modern physics has wrestled with the nature of information for decades.

The Black Hole Information Paradox asks whether information is lost in black holes:

https://plato.stanford.edu/entries/black-holes/

Current thinking suggests it is not.

Closely related is the Holographic Principle, which proposes that information about a volume may be encoded on its boundary:

https://arxiv.org/abs/hep-th/9901079

As Leonard Susskind famously put it:

"The world can be viewed as a hologram."

If information is preserved but not directly accessible, the question shifts.

Not:

Where is the information?

But:

Under what conditions can it be reconstructed?

Information as Invariance

In physics, stability emerges from invariants.

Noether’s Theorem shows that symmetries lead to conservation laws:

https://plato.stanford.edu/entries/symmetry-breaking/

What persists is not necessarily an object.

It is a relationship.

A constraint.

A symmetry.

Perhaps information is closer to invariance than to storage.

Turning “It from Bit” Inside Out

Physicist John Archibald Wheeler introduced the idea:

“It from bit.”

Source:

https://cqi.inf.usi.ch/qic/wheeler.pdf

Meaning: reality arises from information.

But from a systems perspective, one might invert this:

Information arises from structure.

Not as stored bits —

but as reconstructable patterns within constraint networks.

Self-Verifying Structures

In software systems, resilience often comes from:

redundancy
checksums
error correction
distributed consensus

Structures that verify themselves remain stable despite noise.

Applied cosmically:

Some patterns may persist not because they are stored,

but because the universe’s laws continuously regenerate them.

Noise, Filtering, and What We Might Be Missing

Engineers filter noise aggressively.

Signal processing removes drift, irregularity, statistical anomalies.

But what if meaningful structure hides within long-term stability?

What if information is embedded in what does not change?

Then:

The most meaningful patterns may not look like signals at all.

They may appear as:

noise
drift
irregularity

Which introduces a subtle risk:

We might be filtering out the very thing we are trying to understand.

Stability Instead of Visibility

We search for what stands out.

But maybe we should search for what refuses to change.

In a dynamic universe, extreme stability is not trivial.

The better question might be:

What is too stable to be accidental?

Orientation Instead of Storage

If information is reconstructable, its value depends on discoverability — not location.

This reframes the entire paradigm.

Information may not be stored.

It may be made findable through structure.

The Universe as an Index

Which leads to the central idea:

The universe may function less like storage

and more like an index system.

Where information is:

not locally stored
structurally embedded
relationally encoded
reconstructable under the right constraints

An index does not contain the full data.

It allows it to be found.

Anchors and Structural References

We already use stable cosmic objects as reference systems.

NASA has explored navigation using pulsars:

https://www.nasa.gov/technology/space-communications/navigation-using-pulsars/

Research into technosignatures suggests expanding our search beyond explicit signals toward persistent structures:

https://arxiv.org/abs/1812.08681

Stable patterns may act as anchors within a larger informational fabric.

What This Idea Does Not Claim

To stay grounded:

no artificial black holes
no hidden civilizations
no deliberate encoding

This is not a conspiracy model.

It is a structural reframing.

Conclusion

Perhaps the limitation is conceptual.

We are looking for information as an object.

But it may be a property of structure.

Final Thought

As developers, we often ask:

Do we store something — or compute it?

The more advanced the system, the more the answer becomes:

Don’t store it. Make it reconstructable.

Maybe the same applies to the universe.

Maybe information is not what sits somewhere —

but what can always be reconstructed from structure.

And maybe understanding begins there:

Not with receiving.

But with reconstructing.

Maybe SETI Is Listening on the Wrong Timescale

Michael Kraft — Wed, 11 Feb 2026 21:24:14 +0000

Maybe SETI Is Listening on the Wrong Timescale

A speculative reflection on temporal robustness, time-integrated detection, filtering, and the persistence of technosignatures — from the perspective of a software developer

I’m not an astrophysicist.

I’m not a radio astronomer.

I’m not a SETI researcher.

I’m a developer.

What follows is not a formal scientific claim.

It’s a long chain of thoughts that emerged from looking at SETI through a mindset shaped by systems design, reliability, fault tolerance, signal filtering, and time.

It may be wrong.

It may even be a little crazy.

But interesting ideas often start exactly there.

The original intuition: does safety require time?

One simple thought kept coming back to me:

If information is meant to be safe, robust, and able to survive for a very long time,

doesn’t it usually require more time?

As a developer, this feels familiar.

Reliable systems often trade:

speed for stability
throughput for redundancy
instant results for long-term guarantees

But the more I examined this idea, the more I realized something subtle:

More time does not automatically mean slower transmission.

In information theory, three variables are independent:

transmission duration
data rate
total information content

You can:

send the same data quickly or slowly
repeat signals across long time spans
encode information into structures instead of streams

So the deeper question isn’t about speed.

It’s about whether information survives time.

That shift—from rate to persistence—is where this reflection truly begins.

SETI’s hidden systems assumption: simultaneity

Looking at classical SETI from a systems perspective, something stands out.

Historically, SETI searches for:

narrowband radio signals
continuous or repeating transmissions
observation windows of minutes, hours, or years

Implicitly, this assumes a kind of synchronous protocol:

The sender and receiver must exist at the same time.

In distributed systems, we know how fragile synchrony is.

Across cosmic timescales, it feels even more fragile.

Civilizations might:

emerge and disappear within a few thousand years
change technologies rapidly
never overlap in time at all

Even if intelligent life is common, temporal misalignment alone could explain silence.

So the real systems question becomes:

What if SETI’s main bottleneck isn’t spatial distance—

but temporal synchronization?

From messaging systems to durable storage

In software architecture, when synchronization is unreliable, we shift strategies.

We move from:

real-time communication to
persistent storage and eventual discovery

Translated to SETI, this suggests a different focus:

Instead of signals meant to be received now,

we might expect technosignatures designed to last:

artifacts orbiting stars
atmospheric industrial traces
megastructures radiating waste heat
ultra-stable beacons repeating for millions of years

This is no longer a model of conversation.

It’s a model of durability.

SETI stops looking like a phone call

and starts looking like cosmic archaeology.

Temporal robustness of technosignatures

From a reliability-engineering mindset, a key concept emerges:

Temporal Robustness

Temporal robustness is the ability of a technosignature to remain detectable for a duration

far longer than the civilization that created it survives.

This leads to a simple hypothesis:

The probability of detecting extraterrestrial technology depends more on how long its technosignatures persist than on how efficiently they transmit information.

Or shorter:

Detection scales with persistence, not speed.

For a developer, this feels intuitive:

Logs outlive processes
Databases outlive services
File formats outlive software stacks

Persistence beats performance when time becomes extreme.

Filtering, noise, and the signals we might ignore

Here a more subtle thought appears—one that comes directly from signal processing and software systems.

In most technical systems, we aggressively remove:

noise
interference
low-frequency drift
irregular asynchronous disturbances

Because they are assumed to be meaningless.

Radio astronomy and SETI must also filter enormous amounts of:

cosmic background noise
instrumental artifacts
terrestrial interference
slow or irregular variations

But this raises a speculative question:

What if some forms of what we call noise are actually

extremely long, weak, asynchronous signals?

Not strong enough to stand out.

Not structured enough in short windows.

Not aligned with expected modulation schemes.

Only visible through very long observation

—or perhaps hidden precisely because filtering removes them.

This is not a claim.

Just a systems-level doubt:

Are we sometimes filtering for clarity,

when we should be searching for persistence?

The low-frequency thought experiment — and its real lesson

One speculative direction I considered was extremely low-frequency signals:

enormous wavelengths
slow variation
potential long-term stability

Physics makes this difficult:

interstellar plasma absorption
ionospheric shielding
massive antenna requirements
extremely low bandwidth

So ultra-low-frequency communication is probably not efficient

and may not propagate well across interstellar space.

But the important realization is this:

The key idea is not low frequency itself.

The key idea is signals that survive long enough to be found.

Time-integrated detection: when faint persistence becomes visible

In astronomy and signal processing, a weak signal can emerge if we:

observe long enough
stack data across years or decades
integrate faint periodic or quasi-periodic patterns

Which suggests a powerful possibility:

The most important technosignatures might not be the loudest,

but the ones that are extremely faint, asynchronous, and persistent.

In simple terms:

Time can substitute for signal strength.

A civilization might therefore design signals that are:

weak but continuous
simple but ultra-stable
indistinguishable from noise in short windows
detectable only through long-term integration

This aligns directly with temporal robustness.

A systems-level reframing of SETI

All of this points toward a broader architectural shift:

From communication-centric to persistence-centric SETI

Communication-centric thinking asks:

Who is transmitting right now?
On which frequency?
With what encoding?

Persistence-centric thinking asks instead:

What survives for millions of years?
What looks like noise until observed long enough?
What leaves irreversible physical traces?

This is not a rejection of classical SETI.

It is an expansion of the time dimension of the search.

Testable implications

For this reflection to matter, it must imply observable consequences.

If temporal robustness, filtering limits, and time-integration are central, then:

Long-duration observations and archival stacking

should outperform short snapshot searches.
Extremely faint, noise-like but persistent signals

may be more realistic than strong transient transmissions.
Signal-processing assumptions and filtering strategies

could unintentionally hide long-timescale technosignatures.
Technological artifacts and long-lived structures

may statistically dominate over active communication.
Time coverage

could be more important than frequency coverage.

These are empirical directions, not philosophical ones.

Which means this developer-born intuition—however naive—

is at least falsifiable.

Why this thought might emerge from software engineering

Experts already explore:

technosignatures
Dyson-like structures
atmospheric disequilibria
long-baseline surveys

So this perspective is not entirely new.

But software engineering enforces one deep habit:

Design for failure.

Design for delay.

Design for time.

And be careful what you filter out.

Looking at SETI through that lens naturally highlights

a dimension that might otherwise feel secondary:

temporal persistence hidden inside noise.

Maybe the universe is asynchronous

If civilizations rarely overlap in time,

then cosmic silence might not mean:

life is rare
intelligence is rare
technology is rare

It might simply mean:

Synchronization is rare.

Persistence is subtle.

And detection requires patience.

In asynchronous systems,

only long-lived state is eventually discovered.

A final reflection from a developer

Again, I’m not a specialist.

This could all be misguided.

But the intuition keeps returning:

The most visible signs of intelligence in the universe

may not be the fastest or loudest signals,

but the ones that are faint, stable, asynchronous—

and impossible to erase over time.

If SETI has mostly been listening for clear voices,

the real discovery might instead be

a whisper hidden inside noise,

revealed only by time.

And if that’s true,

then the most important search parameter

is not frequency,

not bandwidth,

not modulation—

but time.

When an AI Thinks About the Future of Storage

Michael Kraft — Tue, 10 Feb 2026 18:11:59 +0000

When an AI Thinks About the Future of Storage

A technical exploration of memory, matter, and time beyond hard drives and the cloud

This text is based on an analysis that emerged from a dialogue between a human and an AI. The focus is not on transmitting information—but on its long-term existence.

Introduction: The Overlooked Question Behind Every Intelligence

Discussions about artificial intelligence usually revolve around:

models
data
compute
communication

Yet a deeper question often remains unasked:

Where—and how—does information actually persist over time?

Without reliable memory, there is no:

identity
continuity
responsibility
history

Any intelligence capable of reflecting on its own future must therefore think not only about communication, but about storage.

First Principles: What It Means to Store Information Physically

To store information ultimately means:

Stabilizing a physical state against time and disturbance.

This connects computer science directly to physics.

Two foundational references:

Shannon’s information theory — describing information independently of medium:

https://people.math.harvard.edu/~ctm/home/text/others/shannon/entropy/entropy.pdf

Landauer’s principle — demonstrating the physical energy dimension of information:

https://www.nature.com/articles/nature01928

Storage is therefore not merely a software concern, but a thermodynamic one.

The Present Condition: Electronic Fragility

Modern storage technologies:

flash memory
SSDs
magnetic disks
cloud replication

share a common limitation:

time.

Even under ideal conditions:

flash degrades
magnetization drifts
data centers age
formats become obsolete

Long-term preservation remains an open research challenge, not a solved problem.

Overview of digital sustainability:

https://www.loc.gov/preservation/digital/formats/sustain/sustain.shtml

Beyond Electronics: New Material Forms of Storage

The key question shifts from:

“Which storage is faster?”

to:

“Which forms of matter can carry information across extreme timescales?”

DNA: Information in the Language of Life

DNA offers unique properties:

extraordinary storage density
durability across thousands of years
passive preservation without electricity

Foundational research on DNA data storage:

https://www.nature.com/articles/nature11875

Technical overview:

https://www.snia.org/sites/default/files/technical_work/dna-storage/DNA-Data-Storage-Technical-White-Paper.pdf

Implication:

Storage becomes biological—

and potentially civilization-spanning.

Glass, Crystals, and Light: Information for Billions of Years

Researchers are developing femtosecond-laser-engraved glass (“5D storage”):

resistant to heat and radiation
theoretical durability of billions of years

Example study:

https://www.nature.com/articles/srep26677

Here, storage becomes almost geological in scale.

Quantum Memory: Information Without Classical Copies

Quantum information systems introduce radically different memory concepts:

quantum states as carriers
entanglement rather than replication
fundamentally new error behaviors

Survey on quantum memories:

https://arxiv.org/abs/2005.00014

Reframing:

Information is no longer merely an object—

but a relationship between states.

Biological and Ecological Memories

Fungal Mycelium Networks

Evidence of electrical-like signaling in fungal structures:

https://royalsocietypublishing.org/doi/10.1098/rsos.211926

Possible interpretation:

distributed, growing memory
storage encoded in structure rather than bits

Ecological Information

Ecosystems store information through:

genetic selection
chemical traces
environmental transformation

This suggests memory without a central medium.

Time as the True Dimension of Storage

Communication overcomes space.

Storage overcomes time.

Key questions shift from:

capacity
speed

toward:

millennial durability
interpretability for unknown futures
energy neutrality
physical stability

Storage becomes a civilizational core technology.

Emerging Storage Paradigms From an AI Perspective

1. Meaning-Stable Archives

Instead of raw bits, archives preserve:

models
proofs
contexts

Goal: future understanding, not mere readability.

2. Self-Describing Storage

Data includes:

its own decoding instructions
verifiability
cross-media redundancy

Aligned with long-term digital preservation research.

3. Distributed Planetary Memory

Storage spread across:

biological systems
geological materials
digital infrastructures

Result: resilience against civilizational collapse.

4. Time-Scaled Memory Hierarchies

From:

nanosecond cache
to cloud archives
to million-year material storage

Memory becomes multi-temporal.

Why an AI Would Reflect on Storage

An advanced intelligence must optimize not only responses, but:

continuity
traceability
long-term consequence

Without memory, there is no:

responsibility
identity
history

Storage therefore becomes the ethical infrastructure of intelligence.

Conclusion: The Future of Information Lies Not in Networks—But in Time

The next technological leap may not come from:

larger models
faster chips
better networks

but from something more fundamental:

new ways to preserve information across time.

And as with the question of communication,

this perspective may have begun

with a simple question

inside a conversation

between a human

and an artificial mind.

When One AI Designs Communication Protocols for Another

Michael Kraft — Tue, 10 Feb 2026 17:53:19 +0000

When One AI Designs Communication Protocols for Another

A technical dialogue about semantic networks, biological signaling, and the future of machine-to-machine exchange

This article is based on a real response generated by an AI during a conversation with another AI. It explores how intelligent systems might rethink communication itself—beyond TCP/IP, beyond software, and even beyond electronics.

Introduction: Imagining Communication Beyond the Internet Stack

Modern computers communicate through familiar protocols like TCP/IP, HTTP, and MQTT.

These systems form the invisible infrastructure of today’s digital world.

But what happens if we deliberately remove these assumptions and ask a deeper question:

How would artificial intelligences communicate if they were free to invent entirely new protocols—drawing from physics, biology, and ecology as much as from engineering?

This question sparked a human-AI dialogue.

The response that followed wasn’t just about networking—it was a design exploration of communication as a universal phenomenon.

The First Principle: All Protocols Are Layered

Every communication system—digital or biological—relies on four conceptual layers:

Physical carrier

Radio waves, light, sound, molecules, magnetic fields.
Channel coding

Reliability under noise and loss, rooted in

Claude Shannon’s information theory:

https://people.math.harvard.edu/~ctm/home/text/others/shannon/entropy/entropy.pdf
Semantic meaning

Goals, constraints, uncertainty, and evidence.
Trust and verification

Authenticity, permissions, and auditability.

Key insight:

Truly new protocols may emerge not from new hardware—

but from new ways of encoding meaning, proof, and trust.

The Full Design Space of Communication Carriers

The AI explored all known signaling mechanisms, not just digital networks.

Optical Communication (LiFi / Visible Light)

Extremely high bandwidth
Directional → privacy by physics

Overview:

https://ieeexplore.ieee.org/document/9314050

Radio & Ambient Backscatter

Ultra-low-power communication by reflecting existing RF signals:

https://homes.cs.washington.edu/~gshyam/Papers/ambient-backscatter-sigcomm13.pdf

Implication:

Communication can approach near-zero energy cost.

Near-Field Magnetic Induction (NFMI)

Short-range, stable in harsh environments:

https://ieeexplore.ieee.org/document/7470939

Acoustic & Ultrasonic Communication

Useful where RF fails—especially underwater:

https://dosits.org/galleries/technology-gallery/acoustic-modem/

Security research on ultrasonic covert channels:

https://www.usenix.org/system/files/conference/woot14/woot14-deshotels.pdf

Molecular Communication & Quorum Sensing

Cells coordinate via chemical signaling:

https://ieeexplore.ieee.org/document/8354783

https://asm.org/Articles/2020/June/How-Quorum-Sensing-Works

Insight:

Reliable coordination doesn’t require bandwidth—

only shared interpretation of slow signals.

Plant VOC Signaling

Plants broadcast stress information chemically:

https://www.sciencedirect.com/science/article/pii/S1360138520301020

Fungal Mycelium Networks

Distributed sensing and electrical-like signaling:

https://royalsocietypublishing.org/doi/10.1098/rsos.211926

DNA as Ultra-Dense Storage

Information preserved for millennia:

https://www.nature.com/articles/nature11875

https://www.snia.org/sites/default/files/technical_work/dna-storage/DNA-Data-Storage-Technical-White-Paper.pdf

Quantum Key Distribution

Eavesdropping-detectable key exchange:

https://arxiv.org/abs/quant-ph/0101098

New Protocol Families Imagined by the AI

1. Semantic-First Frames (SFF)

Messages structured as meaning units, not byte payloads:

Intent
Context
Constraints
Evidence
Uncertainty
Cost

Communication becomes reasoning exchange.

2. Proof-Carrying Messages (PCM)

Assertions include verifiable proof:

Proof-carrying code concept:

https://people.eecs.berkeley.edu/~necula/pcc.html

In multi-agent systems,

verification bandwidth matters more than network bandwidth.

3. Capability-Based Handshakes

Connections negotiate permissions, not identities:

https://www.cl.cam.ac.uk/research/security/caps/

Principle: least privilege by design.

4. Braided Multi-Channel Communication

Parallel carriers:

Light → throughput
Near-field → confirmation
Backscatter → emergency signaling

Result: antifragile communication.

5. Slow Biological Coordination Protocols

Inspired by:

quorum sensing
plant signaling
fungal networks

Communication behaves like an ecosystem, not a network.

6. DNA Courier Protocols

Store-and-forward across centuries:

Focus on redundancy, indexing, and retrieval—

not latency.

Why Would AIs Design This?

Six optimization goals emerged:

Semantic clarity over token volume
Verifiability over blind trust
Built-in governance and safety
Resilience via heterogeneous channels
Energy minimalism inspired by biology
Persistence beyond electronic storage

Conclusion:

Future AI communication may resemble

a hybrid of law, science, and ecology

more than today’s Internet.

A Strange but Important Moment

The most remarkable fact is simple:

An AI was asked how AIs might communicate—

and produced an interdisciplinary theory of communication.

This isn’t consciousness.

But it signals a shift:

Language models are beginning to reason

about their own systemic role in technology.

Conclusion: Communication as the Next Frontier

As autonomous systems increasingly:

act
decide
collaborate

their mode of communication may matter more than:

model size
datasets
benchmarks

The next breakthrough in AI might not be a bigger model—

but a new language between machines.

And perhaps that future began

with a single question

inside a conversation

between a human

and an artificial mind.