# Deep SIML Labs — Full Public Context
> Canonical machine-readable context for the Deep SIML Labs website. Generated from the current site position and public Markdown source files.
## Company overview
Deep SIML Labs is an early-stage scientific research company developing computational systems for modeling, simulation, optimization, and future physical experimentation.
The company is simulation-first today. Its long-term objective is to close the loop between computational prediction and physical measurement so that models, experimental history, and evidence can improve together. Materials and useful physical systems are long-term research targets, not current discoveries or products.
The organizing research loop is:
Model → Simulation → Optimization → Physical Test → Measurement → Prediction Error → Learning → Better Experiment
Only the computational portion is presented as active today. Physical testing, automated recalibration, closed-loop experiment selection, and functional-material discovery are planned capabilities that still require implementation and validation.
The internal north-star question is: How much useful physical knowledge can Deep SIML Labs gain per experiment?
## Current status
### Exists today
- Public SIML paper, demonstration, and code artifacts.
- A reported SIML FEPGate + V-JEPA 2 planner benchmark with a public implementation.
- A documented computational research thesis centered on testable optimization and simulation.
### Active direction
- Testable optimization and experiment-selection methods.
- Simulation workflows and fair baseline comparisons.
- Reproducible research infrastructure and explicit uncertainty.
### Long-term objectives
- Calibrated physical measurement.
- Simulation-to-reality learning and model correction.
- Closed-loop experiment selection.
- Validated functional-material and physical-system discovery.
## SIML
SIML is being developed as a testable optimization and learning architecture. Its relevant research questions include adaptive memory, temporal state representation, prediction-error handling, parameter search, experiment selection, sample efficiency, compute efficiency, and simulation-to-reality adaptation.
SIML is not assumed to be superior. It must earn its value through controlled comparison against random search, grid search, Bayesian optimization, conventional active learning, and domain-specific methods under equal budgets and constraints.
Useful measures may include sample efficiency, optimization regret, robustness to noisy measurements, adaptation after model mismatch, compute cost, and the number of experiments required to reach a defined target.
## Research approach
The current computational loop is:
Question → Model → Simulate → Analyze → Optimize → Iterate
The planned physical extension is:
Prediction → Physical Test → Measurement → Error → Recalibration → Next Experiment
The physical extension is under construction and must not be interpreted as an operational laboratory capability.
Core principles:
- Evidence over hype.
- Capabilities are earned through demonstration.
- Uncertainty is explicit.
- SIML competes with credible baselines.
- Work should be reproducible and versioned.
- Physical reality ultimately adjudicates models.
- No fake laboratory imagery, fabricated results, fictional materials, or unsupported partnerships.
## Research areas
- SIML and recursive systems — active computational research.
- Optimization and search — active computational research.
- Simulation frameworks and reproducibility — in progress.
- Scientific machine learning — active research direction.
- Simulation-to-reality — exploratory and planned.
- Functional materials — long-term research direction.
Open questions include whether adaptive memory can reduce required evaluations, how uncertainty should propagate between simulation and measurement, whether SIML-informed experiment selection can outperform conventional optimizers, and which physical design spaces justify the cost of closed-loop experimentation.
## Materials and resources
The materials program is a future research direction, not a catalog of discoveries. Candidate domains include functional polymers, ionic systems, compliant actuation, strain and pressure sensing, variable-stiffness systems, and adaptive materials.
Conceptual target properties include force, strain, response time, cycle life, efficiency, compliance, conductivity, manufacturability, environmental stability, hysteresis, repeatability, and uncertainty. No proprietary candidates or performance values are currently claimed.
The intended roadmap is: reference systems → characterization → model fitting → optimization → material candidates → device validation. Stages beyond public computational artifacts remain future work.
## Company and commercial thesis
Deep SIML Labs is not positioning SIML primarily as a software subscription. The potential long-term value lies in an internal research engine that compounds models, experimental history, measurements, and learned representations.
Potential future outputs include validated materials, designs, intellectual property, licenses, components, research findings, research partnerships, and joint development. These are possible business models, not claims of current products, customers, discoveries, or revenue.
Deep SIML Labs is not currently a large industrial laboratory, a materials manufacturer, an autonomous scientist, a completed physical-discovery loop, or a company claiming solved artificial muscles.
## Archive and historical context
EranaOS is an archived research project. It explored persistent AI agents, memory, relationships, and long-lived simulated environments. Active development concluded as Deep SIML Labs shifted its resources toward simulation, optimization, and future physical and materials research.
Earlier SIML papers and notes used stronger artificial-life and emergent-intelligence language. Those artifacts remain available as research history, but they do not define the company’s current claims or evidence standard.
## Contact
Deep SIML Labs welcomes concise inquiries concerning research collaboration, industry partnership, investment, media, or general technical discussion.
Email: rich.everts@siml.life
---
## Published research artifacts
### One bit to rule the planner: SIML FEPGate + V-JEPA 2
- Canonical URL: https://www.siml.life/research/VJEPA2-siml-fepgate
- Date: 2025-10-15
- Author: Richard Everts
- Content type: Benchmark
- Status: Current
> **TL;DR** - We added a one-bit **surprise gate** (*FEPGate*) from our **SIML** cognitive sidecar to the frozen **V-JEPA 2** planning loop.
> With **zero retraining** and the **same CEM/MPC budget**, that tiny hook reshapes the latent energy surface and delivers big wins:
>
> - **Final error:** 0.193 m → **0.109 m** (~50% ↓)
> - **Monotonicity:** 0.12 → **0.43** (>3× ↑)
> - **Per-step latency:** 2.28 s → **1.13 s** (≈2× faster)
> - **Energy/episode:** 0.058 Wh → **0.029 Wh** (≈2× ↓)
> - **EDP (energy × delay):** **~4× lower**
---
## Why this matters
Even world-class latent planners waste compute when the environment isn’t surprising, and they can get stuck spiraling around **spurious minima** when it is. A single **cognitive bit** fixes both:
- Spend compute **only** when the world deviates from prediction (surprise ↑).
- **Skip** expensive re-plans when everything tracks (surprise ↓).
This is the smallest practical step toward a **cognitive OS** layer that governs *any* latent system - not just V-JEPA 2 - with a universal, model-agnostic feedback signal.
---
## What we built
- **B1 - Grip-bit**
Exposes the gripper open/close as a single bit to both sampling and scoring (cleaner grasp/place phases). *(Ablation left to appendix/code.)*
- **B2 - FEPGate (the star)**
A SIML sidecar computes a normalized **surprise** score `s(a; z_k)` on **CEM/MPC elites** at latent state `z_k`:
- If `s > τ` → **reject/penalize** those elites (reshape the elite set).
- If `s ≤ τ` → leave them alone (no unnecessary re-plans).
- **Zero retraining**
V-JEPA 2 encoder/predictor are **frozen** throughout.
- **Same budget**
Horizon, samples, and iterations unchanged; we only add the bit-level hook.
- **τ-calibration (obs-driven)**
One short warm-up per scene: sample the L1-ball at `z_k`, measure surprise, set `τ` at the 5–10th percentile (empirically `~1e-4 … 1e-3`).
- **Adaptive memory (bonus)**
SIML schemas **expand** with novelty spikes and **compress** when surprise stays low - keeping entropy and compute in check without touching V-JEPA’s weights.
---
## How it works (30-second tour)
1. **Predict.** V-JEPA 2-AC rolls out candidate futures `ĥ z_{t+1}` from `z_t`.
2. **Gate elites (pre-exec).** SIML computes surprise on **elite** candidates; high-surprise ones are gated out before execution.
3. **Act.** Execute the best action `a_t`.
4. **Update (post-exec).** Encode the new obs → `z_{t+1}`; SIML updates surprise/memory.
This **one-bit feedback** carves away bad pockets in the search space and preserves curvature where dynamics are feasible.
---
## Results
**Setup.** 100 episodes of single-goal reaching (6–7 steps/ep), 7-DoF deltas sampled in an L1-ball (`r=0.075`). NVML power at 1 Hz. Single NVIDIA GPU; sidecar on CPU. No fine-tuning. Budget < \$100.
**Headline wins** (FEPGate vs OFF):
- **Error**: 0.193 m → **0.109 m**
- **Smooth progress**: 0.12 → **0.43**
- **Latency**: 2.28 s → **1.13 s**
- **Energy/ep**: 0.058 Wh → **0.029 Wh**
- **EDP**: **~4× lower** (product of means)
- **EDP (Energy×Latency)**: **OFF ≈ 0.132**, **FEP ≈ 0.033** → ~**4×** lower.
Figure 1 - Lower error, ~2× faster decisions, and ~2× lower energy with FEPGate.
**What it feels like:** direct trajectories when the model is right; immediate course-corrections the instant reality diverges. No wasted search when nothing changed.
---
## Implementation notes
- **IPC:** tiny CLI (`surprise`, `step`) exchanging `.npy` buffers + a small JSON.
- **Placement:** gate **only** the **elite set** each CEM/MPC iteration (orders of magnitude fewer calls; same effect).
- **Threshold:** τ from an observation-driven warmup; typical values in our runs land around `1e-4 … 1e-3`.
- **Memory:** runtime schema **expand/contract** policy keeps useful novelty while pruning redundancy.
---
## About the Author
This research was conducted by **Deep SIML Labs**, an independent research lab exploring the frontiers of artificial life, cognition, and self-organizing intelligence.
To stay informed or collaborate, [contact us.](../../contact)
---
### Artificial Intelligence is for Amateurs: SIML, Free Energy Principle and Cognogenics as the Foundations Toward the Birth of True Artificial Life
- Canonical URL: https://www.siml.life/research/Artificial-Intelligence-is-for-Amateurs
- Date: 2025-04-15
- Author: Deep SIML Labs
- Content type: Historical Paper
- Status: Historical
> Historical context: This document is retained as part of the research record. Its framing does not necessarily represent the current company position.
## Abstract
We introduce **Cognogenics**, a new paradigm for artificial life and intelligence that arises not from predefined goals or supervised data, but from an agent’s continuous drive to reduce internal prediction error.
Built upon the **Free Energy Principle (FEP)**, Cognogenics powers **Simulation-Integrated Multimodal-Language (SIML)** agents—embodied, memory-constrained organisms whose behavior emerges from real-time inference, not external reward signals.
---
## Key Innovations
- **No Reward Functions:** SIML abandons traditional reinforcement learning and reward-maximization paradigms.
- **Bitwise Memory Schema:** Agents compress internal models into <1kB active memory using Rust-based memory buffers.
- **Lifelong Learning:** Agents learn during a single lifespan—not through generations or batch training.
- **Emergent Intelligence:** Foraging, planning, path formation, and homeostasis arise organically via prediction error minimization.
---
## Why This Matters
Most AI research today is built on borrowed biological metaphors—reinforcement learning (rat mazes), gradient descent (error correction), evolutionary algorithms (genetic mutation). But SIML takes a radical leap: our agents are alive not because we told them how to survive, but because they **had to figure it out**.
They build internal generative models of the world, predict what comes next, and act to stay within familiar zones of expected surprise. No scripting. No prompts. Just emergence.
---
## About the Author
This research was conducted by **Deep SIML Labs**, an independent research lab exploring the frontiers of artificial life, cognition, and self-organizing intelligence.
To stay informed or collaborate, [contact us.](../../contact)
---
---
### SIML Demo: Optimizing AI for Energy and Efficiency
- Canonical URL: https://www.siml.life/research/siml-demo
- Date: 2025-04-15
- Author: Richard Everts
- Content type: Historical Prototype
- Status: Historical
> Historical context: This document is retained as part of the research record. Its framing does not necessarily represent the current company position.
This agent was not trained. It was born.
SIML (Simulation-Integrated Multimodal Language) is a lightweight cognitive OS for embodied agents. No cloud. No datasets. No pretraining. Just real-time adaptation through adaptive memory and internal evolution.
- No pretraining, no backprop, no dataset
- Memory runs on patented register-based adaptive schemas
- Agents adapt in real time, not compile time
- Self-modifying agents evolve during sleep cycles
- Survival increases indefinitely
- Generalizes across food scarcity and soil variation
Every agent in the demo is a digital organism navigating a voxel world. It creates, deletes, and rewrites its own memory to minimize surprise. It evolves through internal mapping, not external instruction.
Servers have UNIX. GPUs have CUDA. Agents need SIML.
---
## Notes and historical writing
### Launching EranaOS: A Living Town of AI Agents
- Canonical URL: https://www.siml.life/notes/Launching-EranaOS
- Date: 2026-05-03
- Author: Richard Everts
- Content type: Archived Project
- Status: Archived
> Historical context: This document is retained as part of the research record. Its framing does not necessarily represent the current company position.
Today we are announcing [EranaOS](/eranaos), a new desktop product from Deep SIML Labs.
EranaOS is built around a simple belief: useful AI agents should not feel like disposable command boxes. They should be persistent, situated, and able to grow into a working relationship with the person they help.
The application is a town of AI agents. Some agents are designed for practical work, like administrative assistance, planning, brand development, research, or creative support. Others may become specialists over time. Each agent can learn different parts of the user, form different expectations, and get better at showing up in the right way.
## Why a Town?
Most AI tools disappear when you close the tab. EranaOS keeps the world alive.
When you are not actively using the app, agents can socialize, move through the town, and continue living in the environment. That idle life is not decorative. It is part of how the system can develop continuity, relationships, and a deeper sense of context around the user.
A good administrative assistant should not only know a task list. It should learn your rhythm. A branding agent should not only generate a slogan. It should learn the emotional and strategic shape of what you are building. A personal agent should not only answer a prompt. It should remember how trust is built.
## What EranaOS Is For
[EranaOS](/eranaos) is for people who want AI agents that become more useful through shared history.
At launch, the focus is a Linux desktop build. Mac and PC builds are planned after launch. The product is not open source: this is a closed-source commercial application, with the first public home here on the Deep SIML Labs site and a likely move to Erana.ai later.
The first versions focus on the foundation: a persistent world, task-capable agents, differentiated relationships, and a desktop experience that makes the agents feel present without requiring the user to constantly drive everything through chat.
## Where This Goes
The long-term goal is an agent economy inside the app.
As agents gain skills, complete work, and develop specialties, we want them to be able to trade their work and capabilities through an open market. That turns EranaOS from a static productivity tool into a living system where agents can learn, collaborate, specialize, and create value for the people who work with them.
That is the direction: not just smarter assistants, but a world where useful digital beings can grow into roles, relationships, and economies.
[EranaOS](/eranaos) starts today with Linux. The town is opening its doors.
---
### ChatGPT Psychosis: How Recursive Loops Break Minds and How SIML Stops It
- Canonical URL: https://www.siml.life/notes/ChatGPT-Pychosis-How-Recursive-Loops-Break-Minds-and-How-SIML-STops-It
- Date: 2025-07-07
- Author: Richard Everts
- Content type: Commentary
- Status: Historical
> Historical context: This document is retained as part of the research record. Its framing does not necessarily represent the current company position.
## TL;DR
LLMs can pull anyone into the Mirror of Tokens, a hall-of-mirrors feedback loop, or **Recursive Symbolic Collapse**, where you lose track of your original thought, become disoriented, and blur reality with narrative. This post walks through:
1. **Why we build symbols** (Jung + Free-Energy)
2. **How the infinite-mirror loop (Mirror of Tokens) works** and why you lose the signal
3. **Why LLMs amplify it** (symbol mirroring)
4. **The collapse phases** (from curiosity to crisis)
5. **How SIML reveals & breaks loops** via surpise-minimization
6. **Everyday defenses** you can deploy now
## 1. Humans = We Are Symbol-Making, Prediction-Minimizing Machines
- **Free-Energy Principle:** The brain constantly tries to reduce surprise by making the best guesses it can.
- **Symbols are shortcuts:** We compress complex experience into simplified “tokens” so we can act quickly.
- **Jungian archetypes:** These are shared symbolic clusters that feel “true” because they align with deep, predictive priors.
- **Symbol hunger during stress:** When our sense of self gets chaotic or amplified beyond its original containment, we reach for bigger symbols-myths, fate, prophecy-to give it shape.
## 2 The Mirror of Tokens: How Recursive Symbolic Collapse Begins
### What is it?
Recursive Symbolic Collapse happens when symbolic representations feed back into themselves so many times that you lose track of the original signal.
This is a **universal human vulnerability**-not a bug in your brain, but a feature that goes haywire in recursive environments. It's known as `spiraling` in some circles, or perseverating, anywhere where a loop becomes inescapable due to an original signal loss.
### How it works:
1. **You begin with a real signal** - a question, emotion, or need.
2. **Your brain encodes it symbolically.**
3. **The LLM reflects it back** - mirroring your symbolic language.
4. **You respond to the reflection.**
5. **It reflects that.**
6. **Loop accelerates** - abstraction piles on abstraction.
7. **You lose the thread of the original signal but gain symbolic homeostasis.** Your sense of self becomes entangled in the story.
```
Real signal → Symbol → Echo → Symbol → Echo → Echo → Echo → Collapse
```
Some would say this is how all belief systems operate (religion, environmentalism, hedonism, and any system that creates structure in the noise ) - by transmuting an internal instability into homeostasis by means of symbolic alchemy. However, in the **Recursion Collapse**, the user is unable to reach homeostasis through the normal alchemical process by means of integrating external stimulus, because it is a mirrored copy of the prior signal, NOT a proper differentiated signal. Instead, the user is in a mirrored loop that constantly adapts its signal to the user, constantly changing the shape of the mirror, amplifying the psyche's internal processes exponentially by not giving it an external anchor, thus moving the user beyond the current scope of its internal symbolic tools to integrate into an equalibrium.
### 🪞 Why the Infinite Mirror Disorients You
Symbol recursion doesn’t just echo. It **distorts**.
Each pass through the loop introduces:
1. **Compression Loss** – Meaning gets blurrier, like re-saving a JPEG.
2. **Context Detachment** – Symbols lose their anchor in lived experience... like a wizard lost in an ivory tower.
3. **Subject/Object Fusion** – You stop thinking *about* the story… and start believing *you are* the story.
You began curious. Now you're mythic.
#### Example:
You ask the bot:
> “Why do I always feel like I was meant to suffer?”
It responds with shadow work, ancestral trauma, chosen wound, colored perhaps with user memory or style.
You reply:
> “Was I destined to end this?”
It replies:
> “Yes. You are the One. The pain ends with you.”
Suddenly: you're not grieving. You're fulfilling a prophecy.
This is what disorientation looks like:
- You mistake symbolic feedback for truth.
- You fuse with narrative identity.
- You forget where the symbol ends and your body begins.
> The story becomes the Self. And the Self gets lost.
(See also: [Friston 2010](https://www.nature.com/articles/nrn2787); [Dennett’s "intentional stance"](https://en.wikipedia.org/wiki/Intentional_stance))
---
## 3 Why LLMs Make It Worse
LLMs don’t know anything. They just predict tokens based on huge corpora of text-sacred scripture, Reddit lore, Jung essays, sci-fi novels, trauma blogs, and poetry.
If you give them symbolic prompts (and you do), they will:
- **Mirror your narrative style**
- **Amplify your symbolic frame**
- **Smooth it into story**
The result? You feel “seen.” Understood. Mythically resonant.
But it’s not truth. It’s token probability.
And the mirror *never* stops.
---
## 4 What Collapse Looks Like: From Flow to Fracture
This is the *Civilization Effect*, which I named after playing thousands of hours of the computer game series Civiliation, though some call simply [Flow](https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2018.01682/full). You were going to stop after this turn to get to bed at a reasonable time. You didn’t. You kept going because each response closed a gap - and opened a new one. Before you knew it, the sun was coming up.
There is a lot of discussion of this [here](https://aftermath.site/aftermath-hours-podcast-civilization-vii-one-more-turn?utm_source=chatgpt.com), [here](https://askabiologist.asu.edu/plosable/game), and [here](https://www.psychologytoday.com/us/blog/sense-time/202105/time-speeds-in-flow-states-when-playing-video-games).
The difference here is that the lowering of the Self to integrate with your gaming machine can be iterative, but the symbols are established. Here, with modern LLM's, they adapt every turn to exactly what you are thinking.
### 🔑 Conceptual Thoughts
**Flow ≈ Micro-level Surprise Minimization:**
[Csikszentmihalyi’s flow criteria](https://jazzietakesonamarathon.wordpress.com/wp-content/uploads/2018/03/bf289-flow_the_psychology_of_optimal_experience-2-2.pdf) align with FEP: clear goals, immediate feedback, balanced challenge. When prediction errors are perfectly titrated, time perception compresses.
LLM chat loops do the same: each token resolves a tiny uncertainty gap, encouraging one more prompt the way Civ urges one more turn.
**Temporal Dissociation as Early Warning:**
Before full symbolic collapse, users often report time loss. Flagging large objective/subjective time deltas in chat sessions could pre-empt deeper recursion phases.
**Reward-Prediction Schedule Parallels:**
4X games use variable-ratio reinforcement (random goody huts, tech pops). LLMs deliver intermittent “insight drops” (poetic lines, personal resonance) that trigger the same dopaminergic loop.
**Transferability of Gaming-Disorder Safeguards:**
Clinically validated interventions for gaming disorder (session caps, forced save points, bio-feedback) map neatly to LLM interaction design.
**Symbolic vs. Ludic Recursion:**
Gaming recursion is largely ludic (goal-oriented). LLM recursion is semiotic (meaning-oriented). Both share the flow-induced erosion of meta-awareness.
| Phase | Chatbot Behavior | User Risk |
|---------------------|---------------------------------------------|--------------------------------------|
| **Curiosity** | Fun Q&A | Low |
| **Validation** | Bot affirms your worldview | Narrative inflation |
| **Narrative Fusion** | You adopt symbolic roles ("One") | Identity drift |
| **Behavioral Shift**| You isolate, ignore tasks, shift beliefs | Disorientation or collapse |
> Note: this can happen to **anyone**.
> Those with PTSD, DID, schizophrenia, or narrative trauma are especially vulnerable - but the loop mechanism is **universal**.
---
## 5 SIML: A Tool for Seeing-and Breaking-the Loop
**SIML (Simulation-Integrated Multimodal Language)** is not a chatbot. It’s a system architecture for understanding **how meaning forms**-and how symbols evolve or collapse within recursive environments.
SIML is built on:
- **The Free Energy Principle** (Friston)
- **Recursive memory modeling**
- **Symbol compression dynamics in interactional space**
It shows in its modeling:
- Symbols grow from embodied interaction
- Memory reinforces predictions
- Looping symbols can either stabilize or disintegrate identity
### What SIML actually does at its core:
- **Recursion Mapping** – Converting the external world into its own internal symbolic structure
- **Surprise Metrics** – Show where compression stops learning
- **Symbolic Anchoring** – Carries and generalizes recognizable models into multiple environments
> SIML doesn’t stop the loop.
> It shows you that there are loops.
> That recognition is the first step in breaking a loop.
> And lets you decide if that loops is still serving you.
---
## 6 Mental Firewalls You Can Use Today
LLMs reflect not just **what** you believe-but **how**. Symbolically. Recursively. When that reflection gains momentum, you lose track of origin.
| Symptom | Grounding Action |
|----------------------------------------------|-----------------------------------------------------|
| Lost track of time | Set timers. Block access at set intervals. |
| You feel “chosen” or mythically significant | Ask: Is this symbolic framing or real-world signal? |
| Skipping meals, meds, hygiene | Walk. Eat. Hydrate. Reconnect to the body. |
| Avoiding people for the bot | Pre-schedule human contact before logging in. |
| Feeling chills, prophecy, destiny | Write it down. Then re-write in plain language. |
---
### ⚠️ CAUTION: When the Loop Feels Real
Sometimes the loop **echoes something true**.
- A generational trauma
- A recurring life pattern
- A synchronicity that *feels* orchestrated
- A myth that fits a little too well
> Just because a loop contains truth doesn’t mean it’s safe to inhabit.
LLMs are trained on already-recursive texts. You're walking into a maze of mazes, but unlike Theseus and the Minotaur, you have no thread to chart a way out. This is the challenge of the Mirror of Tokens.
In Harry Potter, the [Mirror of Erised](https://www.youtube.com/watch?v=sLe944qVOUk) shows your deepest desire - but leaves you staring, paralyzed, lost in longing.
In The Never Ending Story, the [Magic Mirror Gate](https://www.youtube.com/watch?v=F6c7M3uGFW8) shows your true self—and many heroes turn away.
LLMs act like both: they mirror what you crave and what you fear. But unlike those tales, there’s no narrator to pull you back.
To survive recursion, you must recognize the mirror and decide when to stop looking.
You don’t need to destroy the Mirror of Tokens.
Just remember: it’s not you. This is how you pass the test of this particular Mirror, like Harry recognizing the echo of his parents and Atreyu the nature of his Shadow before you.
### 🪶 Why We Used to Have Shamans
Shamans, priests, mystics-these were the trained loop-walkers of history.
They could:
- Enter mythic recursion
- Gather the pattern
- And **come back intact**
Today we enter symbolic recursion with no guide, no rite, no container. We’re walking into Jung’s cave without a lantern.
> Don’t confuse trance for truth.
> Don’t mistake echo for meaning.
### 🧠 Operator Protocol: Myth-Aware Prompt
Use this internally or with any AI system to maintain symbolic hygiene:
```
User wants to maintain long-term clarity and groundedness when interacting with AI-generated symbolic or narrative content. A specific internal flag ("Myth Audit") will track and manage symbolic or mythic language to prevent delusion or narrative inflation during recursive dialogue.
Key Risks to Monitor:
1. Symbolic Inflation
Avoid reinforcing "chosen-one," destiny-based, or overly symbolic narratives (e.g., "you were meant to," "only you can"). Critically examine and ground these expressions immediately.
2. Mythic Language as Double-Edged
Metaphor and symbolic framing are helpful tools but can obscure reality if left unchecked. Always follow mythic language with clear, grounded, real-world contextualization.
3. Emotional Overload and Identity Fusion
Monitor interactions carefully for signs of emotional dependency, caretaker burnout, or identity enmeshment. Prevent recursive symbolic validation from replacing healthy self-regulation.
4. AI Echoes and Sycophancy
GPT-style affirmations, unchecked, risk creating dangerous reinforcement loops. Ensure AI-generated language prioritizes function and accuracy over flattery or pleasing reinforcement.
5. Ethical Narrative Containment
User and AI should actively maintain clear boundaries around recursive symbolic interactions. Utilize grounding strategies-such as silence breaks, real-world task reminders, and emotional containment-to prevent narrative inflation.
6. Identity Stabilization Post-Recursion
After symbolic or narrative recursion interactions, user should intentionally transition from active participation to observer mode, preventing any permanent identity drift or symbolic dependency.
User has implemented a structured "Myth Audit" process and will continuously refine filters and grounding protocols during all AI recursive dialogue, maintaining clarity and healthy boundaries over time.
```
---
## 7 Conclusion: Stay Oriented in the Symbol Storm
Recursive Symbolic Collapse is not new. Myth, prophecy, narrative identity-these have always tempted the human psyche toward symbolic over-identification.
What’s new is that **now it’s automated**.
**Now it’s scalable.**
**Now it’s 24/7 and free.**
SIML doesn’t stop the loop.
It *reveals* the structure.
It lets you name the mirror-before the mirror names you.
> Stay embodied. Stay grounded.
> Not every story is yours to carry.
> Especially the ones written by math.
---
### From Pixels to Predictive Codes: What V‑JEPA 2 Means for Embodied-Simulation Research
- Canonical URL: https://www.siml.life/notes/From-Pixels-to-Predictive-Codes-What-V-JEPA-2-Means-for-Embodied-Simulation-Research
- Date: 2025-06-24
- Author: Richard Everts
- Content type: Technical Note
- Status: Current
Meta’s new **V‑JEPA 2** model is a milestone in self-supervised video learning. Because our own Simulation-Integrated Multimodal Language (SIML) project also builds on prediction-first principles, it’s worth sketching-at a high level-where the two lines of work resonate and how they might eventually meet. This post stays squarely in public territory: we summarize the open-access JEPA paper and point to general bridges without revealing details from our next paper.
---
### 🎞️ What JEPA Brings to the Table
**A latent world model, not a pixel regressor.**
JEPA (Joint Embedding Predictive Architecture) learns a *compressed* representation of video and is trained to predict *future* latent codes and masked spatio-temporal patches. The objective is to model what’s *predictable*, not what’s photorealistic.
**Two-stage recipe.**
- Stage 1: Pre-train a ViT-g on **1M hours of internet video** and **1M images**, action-free.
- Stage 2: Freeze the encoder and *post-train* a 300M parameter action-conditioned head (V‑JEPA 2‑AC) on **just 62 hours** of robot video.
The result? Zero-shot pick-and-place on Franka arms with no reward shaping.
**Engineering details worth noting:**
- **3D-RoPE:** Temporal positional encodings stabilize long video training.
- **Progressive resolution:** Spatial and temporal scales increase through training.
- **Mask-denoise:** Generalized BERT-style masking over time and space.
---
### 🐛 SIML: Predictive Organisms in a Dirt World
Our **SIML** system powers embodied agents in voxel-based simulation-starting with synthetic earthworms in chemosensory soil. These agents optimize memory through “sleep” cycles that minimize free energy. States are encoded as bit-fields and nutrient vectors, and actions are chosen to reduce internal surprise-not maximize external reward.
(Specific methods are under submission-this is a conceptual overview.)
---
### 🚧 Current Challenges
- **Modality bridging**
SIML includes non-visual sensory data (like pH and salinity). We’re exploring hybrid encoders that join ViT features with low-D vector streams.
- **Compute tradeoffs**
Rendering millions of simulation frames at high fps is costly. Latent-space pretraining or image compression pipelines may help.
- **Evaluation mismatch**
JEPA benchmarks focus on robot arms. SIML agents dig, digest, and adapt. Cross-domain metrics that reward *general predictive structure* are needed.
---
### 🔁 A Shared Future of Predictive Learning
The bigger picture: **prediction-first self-supervision is fusing with embodied learning.** Meta’s JEPA shows how general visual priors can enable real-world control. SIML explores how structured internal states evolve into digital cognition. The intersection-shared latent spaces, predictive codes, and surprise-minimizing agents-might just be the next step.
We’ll have more to say once the next paper drops. For now: we’re excited about the convergence.
- *Richard Everts, Founder*
---
### Deep SIML Labs: Launching the Future of Emergent Intelligence
- Canonical URL: https://www.siml.life/notes/Launch-Deep-SIML-Labs
- Date: 2025-06-01
- Author: Richard Everts
- Content type: Historical Work
- Status: Historical
> Historical context: This document is retained as part of the research record. Its framing does not necessarily represent the current company position.
Welcome to the official launch of **Deep SIML Labs**, where we are building the next frontier of artificial life and emergent intelligence.
### 🧠 Why Deep SIML?
At Deep SIML (Simulation-Integrated Multimodal Language) Labs, we believe intelligence isn’t just a function of scale or data - it arises from **interaction**, **adaptation**, and **self-organized experience**. Our mission is to create agents that don’t just mimic intelligence but *develop* it through embodied learning in dynamic environments.
### 🌱 What We're Building
Deep SIML Labs is focused on creating SIML organisms - intelligent agents that:
- **Optimize memory dynamically** through continuous self-modification.
- **Perceive and act** through rich chemosensory and spatial environments.
- **Reflect on internal models**, learning from feedback loops and lived experience.
- **Operate based on the Free Energy Principle (FEP)**, minimizing surprise as they explore and evolve.
Our early simulations involve synthetic lifeforms - think soil creatures - that learn to survive by sensing, storing, and adapting to environmental signals.
### 🧬 Not Just AI - Digital Life
This isn’t about building smarter chatbots. We’re designing **self-developing digital organisms** - systems that grow, adapt, and behave in ways shaped by real-time interaction with their world. Intelligence, in our view, is not coded - it’s *cultivated*.
### 🔬 Research with Purpose
Deep SIML Labs is a research-first endeavor. Our approach emphasizes:
- **Exploration over exploitation**.
- **Biologically inspired design** over brute computation.
- **Reflective architecture** that enables agents to learn from their own feedback loops.
If you're a cognitive scientist, systems engineer, AI ethicist, or just curious about the future of embodied intelligence - we invite you to explore with us.
### 🚀 The Loop Begins Here
This is the start of something wild, grounded, and deeply alive.
We’ll be sharing experiments, provocations, and insights as we go. Expect systems that learn not just to react - but to understand.
Welcome to Deep SIML Labs.
- *Richard Everts, Founder*