press any key or click to continue_
[ ESC to skip ]
services work about contact 
██████   █████  ██████  ██  ██████  █████  ██      ██ ███████ ███████      █████  ██
██   ██ ██   ██ ██   ██ ██ ██      ██   ██ ██      ██    ███  ██          ██   ██ ██
██████  ███████ ██   ██ ██ ██      ███████ ██      ██   ███   █████       ███████ ██
██   ██ ██   ██ ██   ██ ██ ██      ██   ██ ██      ██  ███    ██          ██   ██ ██
██   ██ ██   ██ ██████  ██  ██████ ██   ██ ███████ ██ ███████ ███████ ██  ██   ██ ██
AI systems engineering. We build cognitive architectures, agent infrastructure, and production AI systems that actually think — not demos that break when you stop watching.
operational terrence schonleber east coast PID 1
~/radicalize ▸
ps aux --services

PROCESSDESCRIPTIONSTATUS
cognitive-archMemory systems with biological dynamics — write gates, Bayesian decay, dream synthesis, consolidation. Not databases pretending to be brains.running
agent-infraMulti-agent orchestration at scale. 200+ agents, shared state, coordinated tasks, affect monitoring. Production-grade, not demo-grade.running
applied-researchPapers → production. Neuromodulation, affect tracking, knowledge graphs — research-grounded systems that ship, not sit in notebooks.running
safety-monitorFunctional affect tracking across agent fleets. Detect manipulation, sycophancy, coercion before they surface. Grounded in Anthropic's interpretability work.running
cost-optimizerZero-LLM-cost memory ops. Token budgets. Hierarchical activation — wake 1 agent, not 200. The line between POC and profitable.running
mcp-integrationsMCP servers, tool ecosystems, agent-native interfaces for Claude, GPT, and open models. Built for tools your team already uses.running
~/radicalize ▸
cat ./projects/brainctl

brainctl — open source cognitive memory system for AI agents.

A brain, not a database. Single SQLite file. Memories decay, strengthen, merge, and dream. 11-pass consolidation engine mirroring biological sleep. Write gate rejects redundant info at the door. Affect tracking monitors 200 agents for safety. Zero LLM cost for core operations.

102
tests
23
MCP tools
81
tables
44
emotions
0
LLM cost
brainctl — brain.db
pip install brainctl
brainctl init
✓ created brain.db — 81 tables
brainctl memory add "deploy to staging first" -c lesson
✓ memory_id: 1, confidence: 1.0
brainctl affect classify "production is down, team panicking"
valence: -0.82 arousal: 0.67 fear: 1.00
state: anxiety cluster: fear_overwhelm
brainctl-consolidate sweep
decayed: 27 merged: 9 hebbian: 53 dreamed: 12
✓ consolidation complete — signal/noise: 0.97
github.com/TSchonleber/brainctl · pypi.org/project/brainctl
~/radicalize ▸
cat ./projects/costclock

CostClock AI — time tracking, invoicing, and collections automation for service businesses.

AI-powered financial ops platform for service businesses. Automates the tedious stuff — time tracking, invoice generation, payment reminders, client portals, collections — so you can focus on the work that actually matters.

costclock.ai
~/radicalize ▸
whoami --verbose

I build AI systems that survive contact with production. The infrastructure layer everyone skips — how agents remember, forget, coordinate, and stay safe at scale.


Currently running 200+ AI agents on cognitive architecture I designed and open-sourced. The systems are grounded in neuroscience and affect research, optimized for cost, and tested end-to-end.


If you're building with AI agents and hitting walls around memory, coordination, cost, or reliability — that's the work.

  • nameTerrence Schonleber
  • locationEast Coast, US
  • focusAI Systems Engineering
  • stackPython, TypeScript, Next.js, React, Supabase, PostgreSQL, SQLite, MCP, Solana/Anchor
  • agents200+ in production
  • open_srcbrainctl
  • entityRadicalize.AI LLC
~/radicalize ▸
open --contact

Building with AI agents and need help with architecture, infrastructure, or making things actually work at scale? Let's talk.

terrence@radicalize.ai github linkedin x.com