GABBE: Cognitive Engineering Platform That Turns AI Coding Agents into Engineering Teams

Author(s): Andrei Baslyga (Nikola)

Originally published on Towards AI.

A deep dive into an open-source kit that gives AI assistant agents a mind, a memory, and a “conscience.”

“The agent is the engine. You are the steering wheel.”

the problem no one talks about

AI coding agents – Cloud, Copilot, Cursor, Gemini, Codex – promised a revolution. They delivered quickly. but the teams started Drown Code they couldn’t review, verify, or trust.

Tests abandoned. Architectural decisions were made quickly and then undone in the next session because the agent had forgotten everything. Security review? An afterthought.

Gabbe (Generative Architectural Brain Base Engine) was born from this. created by andrey nikolay baseligaThis asks a different question: “How do we create AI code in the right manner, safelyAnd in a governed manner – Every time?”

The answer is drawn from software engineering archetypes, neuroscience, cognitive psychology, and multi-agent systems theory. The result: a kit that powers AI agents A mind, a memory, and a “conscience”.

What is gab?

gab is one Universal Agentic Software R&D Engineering Kit – A drop-in context layer that turns any AI coding agent into a governed engineering team.

it is No A new AI model. it is No A new IDE. it is Operating System for Agentic Software Development.

120+ Skills · 60+ Templates · 30+ Guides · 30+ Personas · 10 SDLC Gates – Compatible with Cloud, Cursor, Copilot, Gemini, and Codex.

Its Universal Skill Compiler Lets you write a skill once and automatically deploy it to every platform.

Another deadly trifecta

GABBE identifies three complex problems in agentic AI development

*The Deadly Trifecta and GABBE’s Solution*

✓ Velocity mismatch – Agents code 10-100× faster than humans review. Important steps are missed.

🎲 Non-fatalism – Same prompt, different output. No continuity across sessions.

💸Reference rot -Wrong decision compound. A missed check on requirements costs 100× more in code review.

GABBE solves these with a 10-Gate Quality System (Agents cannot leave the gate), AGENTS.md as a single source of truth (each agent reads it first), and a 4-layer memory architecture (Agents never repeat failed methods).

System Architecture

GABBE considers the software system as cognitive institutions – Drawing straight from neuroscience.

4-Layer Memory: No more agent amnesia

Most important file: CONTINUITY.md – Read each session, including every failure, impasse, and hard-won lesson. Thus GABBE ends up “repeating the same mistake”.

Brain Mode: Neuroscience meets AI

Brain mode models the agent as a cognitive turing machine Using active estimation (Carl Friston’s free energy principle).

Main mechanism: brain mode compares to expected vs what actually happened. When there is a mismatch – a test fails, costs increase, requirements are missed – it doesn’t just retry. it updates your model And adapts.

it also handles dynamic cost routing – Automatically sending simple tasks to free local LLM and complex tasks to remote SOTA models.

Loki Mod: A Complete Engineering Department

if brain mode strategistloki mode is Executor. It turns a single AI agent into a 30+ Personality Engineering Swarm.

Product Strategist → Architect → TDD Engineer → Security Auditor → SRE – all coordinated through a 10-gate SDLC with mandatory human checkpoints.

Gab CLI: Hybrid Intelligence

CLI combines Markdown (human-friendly) with SQLite (machine-reliable) – hybrid mode.

Six Commandments: gabbe init, gabbe sync, gabbe verify, gabbe status, gabbe brain, gabbe route.

10-gate SDLC

Each feature passes 10 mandatory gateways from strategy to production – with human approval at critical junctions.

Inside each implementation task, RARV cycle Goes: Reason → Action → Reflect → Verify. If verification fails, the self-heal loop retries up to 5 times before reaching a human.

data flow

creative use cases

vibe coding – Translate aesthetic intent into code: “Make it feel like Cyberpunk 2077 meets corporate SaaS. Glassmorphism, neon accents, micro-animations.”

COBOL modernization – Strangler Fig Pattern with anti-corruption layers and compliance review (GDPR/PCI-DSS/HIPAA).

Enterprise Swarm – Run complex features through Loki Mode’s full personality range.

Autonomous Technical Loan – Impact × Effort prioritized backlog with automated refactoring.

architectural governance – Continuous dependency enforcement on each PR.

who is it for?

Solo Developers – TDD discipline + structured memory without overhead
team lead – Rule over humans + AI agents together
Enterprise Architects – Compliance, Legacy Migration, Traceability
ml engineer – Research Governance, AI Ethics, Self-Correcting Pipelines
Startup CTO – Move forward faster without accumulating deadly technical debt
frontend developers – AI that understands aesthetic intent
safety engineer – Security as a mandatory gate, not an afterthought

major innovations

Neuroscience-derived orchestration – Active Heuristics, Global Vertical Theory, Basal Ganglia Habit Patterns
Write once, run everywhere — A Skills Source → Cursors, VS Code, Cloud, Gemini Format
Hybrid Markdown-SQLite architecture – Human expression + machine reliability
evolutionary early adaptation — The system improves its own signals
self-treatment with increased – 5 retries → human enhancement (not too aggressive, not too passive)
cost-aware routing – Automated local vs remote LLM decision making

Get started in 5 minutes

git clone https://github.com/andreibesleaga/GABBE
python3 GABBE/scripts/init.py

then paste the generated BOOTSTRAP_MISSION.md In your AI agent. That’s it.

First sign: “Read Agents MD. I want to build (X). Start with spec-writer skills.”

“The agent is the engine. You are the steering wheel. GABBE is the roadmap.”

Published via Towards AI