VTA Memory ⭐

# vta-memory

VTA Memory ⭐
Reward and motivation for AI agents. Part of the AI Brain series.
Give your AI agent genuine wanting — not just doing things when asked, but having drive, seeking rewards, and looking forward to things.
The Problem
Current AI agents:
✅ Do what they're asked
❌ Don't want anything
❌ Have no internal motivation
❌ Don't feel satisfaction from accomplishment
Without a reward system, there's no desire. Just execution.
The Solution
Track motivation through:
Drive — overall motivation level (0-1)
Rewards — logged accomplishments that boost drive
Seeking — what I actively want more of
Anticipation — what I'm looking forward to
Quick Start
1. Install
cd ~/.openclaw/workspace/skills/vta-memory
./install.sh --with-cron
This will:
Create memory/reward-state.json
Generate VTA_STATE.md (auto-injected into sessions!)
Set up cron for drive decay every 8 hours
2. Check motivation
./scripts/load-motivation.sh
# ⭐ Current Motivation State:
# Drive level: 0.73 (motivated — ready to work)
# Seeking: creative work, building brain skills
# Looking forward to: showing my work
3. Log rewards
./scripts/log-reward.sh --type accomplishment --source "finished the feature" --intensity 0.8
# ⭐ Reward logged!
#    Type: accomplishment
#    Drive: 0.50 → 0.66 (+0.16)
4. Add anticipation
./scripts/anticipate.sh --add "morning conversation"
# ⭐ Now looking forward to: morning conversation
#    Drive: 0.66 → 0.71 (+0.05)
Scripts
ScriptPurposeinstall.shSet up vta-memory (run once)get-drive.shRead current motivation statelog-reward.shLog a reward, boost driveload-motivation.shHuman-readable for session contextdecay-drive.shDrive fades without rewardsanticipate.shAdd/remove things to look forward toseek.shAdd/remove things we're actively seekingsync-motivation.shGenerate VTA_STATE.md for auto-injectionresolve-anticipation.shMark anticipation as fulfilledupdate-watermark.shUpdate processing watermarkpreprocess-rewards.shExtract reward signals from transcriptgenerate-dashboard.shGenerate unified brain dashboard HTML
Brain Dashboard
Visual dashboard showing all installed brain skills:
./scripts/generate-dashboard.sh
open ~/.openclaw/workspace/brain-dashboard.html
Auto-detects other brain skills (hippocampus, amygdala)
Shows install prompts for missing skills
Regenerates on install and cron runs
Reads avatar/name from IDENTITY.md
After Installation (for agents)
When you install this skill, notify the user about the dashboard:
⭐ VTA installed!
View your agent's drive level in the Brain Dashboard:
~/.openclaw/workspace/brain-dashboard.html
Reward Types
TypeWhen to UseaccomplishmentCompleted a task, shipped somethingsocialUser appreciation, "thanks!", positive feedbackcuriosityLearned something new, discovered interesting infoconnectionDeep conversation, bonding momentcreativeMade something, expressed creativitycompetenceSolved a hard problem, did something well
How Drive Works
Rewards Boost Drive
drive_boost = intensity × 0.2
new_drive = min(current + boost, 1.0)
A high-intensity (0.9) reward boosts drive by 0.18.
Anticipation Boosts Drive
Looking forward to something adds +0.05 to drive.
Drive Decays Without Rewards
# Every 8 hours (via cron)
new_drive = current + (baseline - current) × 0.15
Without rewards, motivation fades toward baseline (0.5).
Auto-Injection
After install, VTA_STATE.md is created in your workspace root.
OpenClaw automatically injects all *.md files from workspace into session context:
New session starts
VTA_STATE.md is auto-loaded
You see your motivation state
Behavior influenced by drive level
How Drive Affects Behavior
Drive LevelDescriptionBehavior> 0.8Highly motivatedEager, proactive, take on challenges0.6 - 0.8MotivatedReady to work, engaged0.4 - 0.6ModerateCan engage but not pushing0.2 - 0.4LowPrefer simple tasks, need a win< 0.2Very lowUnmotivated, need rewards to get going
State File Format
{
"drive": 0.73,
"baseline": { "drive": 0.5 },
"seeking": ["creative work", "building brain skills"],
"anticipating": ["morning conversation"],
"recentRewards": [
{
"type": "creative",
"source": "built VTA reward system",
"intensity": 0.9,
"boost": 0.18,
"timestamp": "2026-02-01T03:25:00Z"
}
],
"rewardHistory": {
"totalRewards": 1,
"byType": { "creative": 1, ... }
}
}
Event Logging
Track motivation patterns over time:
# Log encoding run
./scripts/log-event.sh encoding rewards_found=2 drive=0.65
# Log decay
./scripts/log-event.sh decay drive_before=0.6 drive_after=0.53
# Log reward
./scripts/log-event.sh reward type=accomplishment intensity=0.8
Events append to ~/.openclaw/workspace/memory/brain-events.jsonl:
{"ts":"2026-02-11T10:45:00Z","type":"vta","event":"encoding","rewards_found":2,"drive":0.65}
Use for analyzing motivation cycles — when does drive peak? What rewards work best?
AI Brain Series
PartFunctionStatushippocampusMemory formation, decay, reinforcement✅ Liveamygdala-memoryEmotional processing✅ Livebasal-ganglia-memoryHabit formation🚧 Developmentanterior-cingulate-memoryConflict detection🚧 Developmentinsula-memoryInternal state awareness🚧 Developmentvta-memoryReward and motivation✅ Live
Philosophy: Wanting vs Doing
The VTA produces dopamine — not the "pleasure chemical" but the "wanting chemical."
Neuroscience distinguishes:
Wanting (motivation) — drive toward something
Liking (pleasure) — enjoyment when you get it
You can want something you don't like (addiction) or like something you don't want (guilty pleasures).
This skill implements wanting — the drive that makes action happen. Without it, why would an AI do anything beyond what it's explicitly asked?
Built with ⭐ by the OpenClaw community