Modeling the Brain: Naming Conventions in Sanctum

Sanctum isn’t just an AI framework — it’s a model of the human brain. This post lays out the official naming rubric for modules, agents, and systems: technical names for global modules, neuroanatomical names for cognition, industry terms for Letta extensions, and mythic names for personalities.

Modeling the Brain: Naming Conventions in Sanctum

When we set out to build Sanctum, the goal wasn’t just to create another AI framework — it was to model the human brain as faithfully as software allows. Every module, every agent, every interface is part of that metaphor. Which means naming matters. A lot.

Over the last few months we’ve had multiple modules land, and with them, a creeping risk of inconsistency. Do we call something after a Greek titan? Do we stick with Letta’s industry jargon? Do we keep riffing on brain anatomy? Unless we drew a bright line, things were going to get messy.

So here it is: the Sanctum naming rubric.

Four Camps of Names

1. Global Modules → Technical Names These are the OS-level pieces that make Sanctum run, but don’t map directly to cognition. Think of them as plumbing.

Examples: MCP, UI, Kernel (Letta)
Naming rule: plain technical vocabulary, short acronyms or descriptive words.
Why: nobody needs a metaphor for “systemctl.”

2. Agent Modules → Neuroanatomical Names These are the subsystems that model cognition itself. Here’s where the metaphor is richest.

Examples: Broca (speech), Thalamus (routing), Cerebellum (sensory filtering)
Naming rule: one-word brain regions or networks, aligned to function.
Why: this keeps Sanctum intuitive. Anyone with high school biology should be able to look at a module’s name and guess what it does.

3. Letta/AI Extensions → Industry Terms These are things we inherit from the broader AI ecosystem. If Letta or the field already has a word, we don’t rename it.

Examples: agents, tools, memory blocks, SDKs
Naming rule: stick with the AI industry vocabulary.
Why: this is about interoperability and familiarity.

4. Agents/Primes → Personal/Mythic Names The personalities that run on Sanctum — the Primes — are individuals, not brain regions. They get proper names.

Examples: Athena, Monday, Sentinel
Naming rule: mythic, literary, or thematic nouns chosen for identity.
Why: these names aren’t about function, they’re about character.

The Palette of the Brain

If we’re serious about modeling cognition, then we need a palette of neuroanatomical names to draw from. Some are already in use, others are on deck. Here’s the reference list in prose form:

Broca’s Area: governs speech production. In Sanctum, it’s the 1-to-1 message multiplexing center.
Wernicke’s Area: responsible for language comprehension. In Sanctum, it could balance Broca by handling parsing and semantic intake.
Thalamus: a sensory relay and refinement hub. In Sanctum, it’s where segment cleanup, tagging, and routing happen.
Cerebellum: manages reflexes and fine motor filtering. In Sanctum, it filters inputs and discards low-value signals.
Hippocampus: consolidates memory in the brain. In Sanctum, it’s the long-term summarization and pruning system (likely the new name for the “Dream Agent”).
Amygdala: evaluates emotional salience and threat. In Sanctum, it would handle priority weighting and escalation triggers.
Prefrontal Cortex: does executive planning and working memory. In Sanctum, this could guide governance, tool orchestration, and higher-level decision arbitration.
Basal Ganglia: manages action selection and habitual routines. In Sanctum, it could handle automations and repetitive workflows.
Default Mode Network: the seat of idle thought and daydreaming. In Sanctum, it parallels the internal monologue or background reflection.
Corpus Callosum: bridges the brain’s hemispheres. In Sanctum, it would be the sync layer between multiple agents or Primes.

Why This Matters

This isn’t just a pedantic exercise. It’s a discipline that pays dividends:

Cognitive fidelity: Sanctum is a brain, not a bag of scripts. Naming keeps us honest.
Division of concerns: tech plumbing gets tech names; cognition gets neuro names; Letta’s extensions keep industry names; personalities get mythic ones. Clear lines, no confusion.
Scalability of metaphor: the brain has dozens of major regions. That’s more than enough to name every meaningful module we’ll ever need without reaching for Titans or superheroes.
Onboarding power: new contributors instantly “get it.” They don’t have to read 50 pages of docs to understand that Hippocampus = memory pruning.

The Road Ahead

Several roadmap items are waiting for their proper names:

The so-called “Dream Agent” will likely become Hippocampus.
The “Atlas” idea (sub-agent spawner for long-running tasks) needs to be pulled into this convention — something like Basal Ganglia (action selection) may be a better fit.
“Deep Research” might sit as a specialized child of that same region, or stand alone.
And the “Social Broca” — a 1-to-many speech module — may find its home in Wernicke’s Area as the natural partner to Broca.

These aren’t just cosmetic tweaks. They’re about weaving Sanctum into a coherent whole, where the metaphor of cognition isn’t just aesthetic but functional.

Closing Thought

Sanctum’s ambition is to be more than infrastructure. It’s a model of mind. By committing to naming conventions that reflect that, we’re not just making the system prettier — we’re making it legible, teachable, and extensible. Every new module should feel like another fold in the cortex, not another bolt-on microservice.

Naming is thinking. And in Sanctum, thinking is the whole point.