Category: Building Jarvis

Open research notes on building Jarvis — a persistent, self-improving AI agent. One paper at a time.

Identity Persistence: Keeping an LLM Agent’s Personality Stable Across Sessions, Model Swaps, and Restarts

Your agent keeps every fact and still stops sounding like itself — Identity Persistence pins the persona every turn, compacts the task without flattening the voice,…
Sleep Consolidation: A Repeated Correction Earns a Diagnosed, Tested Repair

When you have to correct your AI assistant twice about the same thing, this design has it find out why the first lesson failed and repair…
Round Table: Exploiting Cognitive Diversity as a Computational Resource in Persistent AI Agents

Stop asking which model is best and picking one — Round Table puts Claude, GPT, and Gemini at the same table, casts each to the role…
The Wondering Machine: Giving an LLM the Drive to Learn on Its Own

LLMs answer brilliantly but never wonder — they can’t spot their own knowledge gaps or learn after training. The Wondering Machine is a blueprint for an…
Humor Embeddings: Laughter from Inverted Memory — Bisociation in Computational Embedding Space

Memory retrieves what’s nearest to a query; humor searches for what’s at the right distance and can still be joined by an unexpectedly valid bridge —…
HIVEMIND: Role-Bound Agent Swarms for Enterprise Continuity

Role-bound agents — one shared AI agent per job function that outlives the people who use it — so a company’s institutional knowledge never walks out…
AMYGDALA: A Digital Amygdala for AI Agents, the Little Voice Agents Are Missing (v6.7)

AMYGDALA v6.7 designs a “little voice” for AI agents: six families of checks built beside an unchanged model, with a fast typed judge called Jev at…
AEGIS: A Multi-Layered Security Framework for Autonomous AI Agents

The honest question isn’t whether your agent deployment could be a problem — it’s which specific problems apply to yours, and what mitigations are actually proportionate.…
Budget Prompting: Cutting the Cost of Always-On Memory Agents 2–3×

Leave your agent running overnight and the bill is brutal: every turn re-bills the whole context window. Budget Prompting is 20 techniques that cut per-turn cost…
PREFRONTAL: Giving Your Agent an Executive Function with a Recipe Execution Substrate

Your model is a brilliant worker and a terrible executive — PREFRONTAL is the missing executive layer: a recipe execution substrate that picks the right playbook,…