M

I

L

A

N

Š

U

L

C

M

I

L

A

N

Š

U

L

C

(O)LLAMA

ANTHROPIC

COPILOT

OPENAI

OPENCODE

CURSOR

CLAUDE

PERPLEXITY

CODEX

github.com/nette
github.com/contributte
github.com/dockette
github.com/ynfra
...

A

🪄

🦄

(SELF-HOSTED)
AGENTIC
DEVELOPMENT

(SELF-HOSTED) AGENTIC DEVELOPMENT

  • Same better code (than human)
  • Knows my codebase, not just code
  • Many agents, one reviewer (me)

  • Not on my machine (disposable & sandboxed)
  • Always on (previewable 👀)
  • No vendor lock (if possible)
  • Self-correcting (runs, fails ❌, fixes)
  • Full Autonomous (night-shifts)
  • Gauč Mobile friendly 🆒
MY VISION
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM
BRIEF
AGENT WORKS
REVIEW
APPROVE
REJECT
NEXT

🧙‍♂️

🧙‍♂️

🧙‍♂️

REPHRASE
REWRITE
REWORK
PROJECT CONTEXT
.
├── app
├── bin
├── config
├── docs
├── tests
├── vendor
└── www

🤖

BRIEF
AGENT WORKS
REVIEW
APPROVE
REJECT
NEXT

🧙‍♂️

🧙‍♂️

🧙‍♂️

REPHRASE
REWRITE
REWORK
.
├── app
├── bin
├── config
├── docs
├── tests
├── vendor
└── www

🤖

AGENTS.md
PROJECT CONTEXT
PROJECT INSTRUCTIONS
BRIEF
AGENT WORKS
REVIEW
APPROVE
REJECT
NEXT

🧙‍♂️

🧙‍♂️

🧙‍♂️

REPHRASE
REWRITE
REWORK
.
├── app
├── bin
├── config
├── docs
├── tests
├── vendor
└── www

🤖

API
spec
DB
spec
Code
​style
Rules
PROJECT CONTEXT
PROJECT RULES
AUTO/MANUAL
APPLY
AGENTS.md
PROJECT INSTRUCTIONS
Tests
classes

BRIEF
AGENT WORKS
REVIEW
APPROVE
REJECT
NEXT

🧙‍♂️

🧙‍♂️

🧙‍♂️

REPHRASE
REWRITE
REWORK
.
├── app
├── bin
├── config
├── docs
├── tests
├── vendor
└── www

🤖

API
spec
DB
spec
Code
​style
Tests
classes

Rules
PROJECT CONTEXT
PROJECT RULES
AUTO/MANUAL
APPLY
AGENTS.md
PROJECT INSTRUCTIONS
Skills
E2E
tests

Migrations
RFC
Writing
Docs
PROJECT SKILLS
SMART
APPLY
BRIEF
AGENT WORKS
REVIEW
APPROVE
REJECT
NEXT

🧙‍♂️

🧙‍♂️

🧙‍♂️

REPHRASE
REWRITE
REWORK
.
├── app
├── bin
├── config
├── docs
├── tests
├── vendor
└── www

🤖

API
spec
DB
spec
Code
​style
Tests
classes

Rules
PROJECT CONTEXT
PROJECT RULES
AUTO/MANUAL
APPLY
AGENTS.md
PROJECT INSTRUCTIONS
Skills
E2E
tests

Migrations
RFC
Writing
Docs
PROJECT SKILLS
SMART
APPLY
IMPROVE
OPTIMIZE
EVOLVE
SKILLS, DOCS, SPECS
BRIEF
AGENT WORKS
REVIEW
APPROVE
REJECT
NEXT

🧙‍♂️

🧙‍♂️

🧙‍♂️

REPHRASE
REWRITE
REWORK
.
├── app
├── bin
├── config
├── docs
├── tests
├── vendor
└── www

🤖

API
spec
DB
spec
Code
​style
Tests
classes

Rules
PROJECT CONTEXT
PROJECT RULES
AUTO/MANUAL
APPLY
AGENTS.md
PROJECT INSTRUCTIONS
Skills
E2E
tests

Migrations
RFC
Writing
Docs
PROJECT SKILLS
SMART
APPLY
IMPROVE
OPTIMIZE
EVOLVE
SKILLS, DOCS, SPECS
WEB/CHAT
REMOTE
Analyse current session and update skills/specs
SELF-IMPROVING LOOP
Analyse all sessions and fix what I keep repeating
Analyse last 10 PR reviews and update code-style rules
Analyse rejected attempts and add a rule to prevent them
Analyse my post-merge edits and codify my taste
Analyse recent Sentry errors and write the missing tests
Analyse your own friction and update AGENTS.md.
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM

✅

AI READINESS
  • Codestyle
  • Linters
  • Testing strategy
  • AGENTS.md
  • Migrations
  • Docs
  • DoD / DoR
  • Techstack
  • Archetypes
  • ....
AI READINESS
  • Codestyle
  • Linters
  • Testing strategy
  • AGENTS.md
  • Migrations
  • Docs
  • DoD / DoR
  • Techstack
  • Archetypes
  • ....
  1. How do we write a <class>?
  2. What does a good <class> look like?
  3. How do we write tests?
  4. How do we run linters?
  5. How do we run tests?
  6. How do we execute migrations?
  7. How do we set up the project?
  8. How do we work with <area>?
  9. ...

Makefile

github.com/vibemod/vibeapp
Factory AI

Style & Validation
Pre-commit Hooks · Cyclomatic Complexity · Large File Detection · Duplicate Code Detection · Technical Debt Tracking · N+1 Query Detection · Linter Configuration · Type Checker · Code Formatter · Strict Typing · Naming Consistency · Dead Code Detection · Code Modularization Enforcement

 

 

 

 

 

 

 

 

Task Discovery
Issue Templates · Issue Labeling System · PR Templates · Backlog Health

Testing
Test Performance Tracking · Flaky Test Detection · Test Coverage Thresholds · Unit Tests Exist · Integration Tests Exist · Unit Tests Runnable · Test File Naming Conventions · Test Isolation

Development Environment
Dev Container · Environment Template · Local Services Setup · Database Schema · Devcontainer Runnable

Product & Experimentation
Product Analytics Instrumentation · Error to Insight Pipeline

Build System
Automated PR Review Generation · Deployment Frequency · Feature Flag Infrastructure · Release Notes Automation · Unused Dependencies Detection · Release Automation · Build Command Documentation · Dependencies Pinned · VCS CLI Tools · Agentic Development · Fast CI Feedback · Build Performance Tracking · Single Command Setup · Progressive Rollout · Rollback Automation · Monorepo Tooling · Heavy Dependency Detection · Version Drift Detection · Dead Feature Flag Detection

Documentation
AGENTS.md Freshness Validation · AGENTS.md File · README File · Automated Documentation Generation · Skills Configuration · Documentation Freshness · API Schema Docs · Service Architecture Documented

Debugging & Observability
Distributed Tracing · Metrics Collection · Code Quality Metrics Dashboard · Alerting Configured · Runbooks Documented · Deployment Observability · Circuit Breakers · Profiling Instrumentation · Structured Logging · Error Tracking Contextualized · Health Checks

Security
Branch Protection · Secret Scanning · CODEOWNERS File · Automated Security Review Generation · Gitignore Comprehensive · DAST Scanning · Privacy Compliance · Minimum Dependency Release Age · Dependency Update Automation · PII Handling · Secrets Management · Sensitive Data Log Scrubbing

±80 REPORT CRITERIA
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM

✅

✅

github.com/contributte/contributte
github.com/contributte/barekit
$ opencode serve
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM

✅

✅

✅

👩‍💻

VPS

CODER SERVER

CODER

TEMPLATE

CODER WORKSPACE

CODER WORKSPACE DIND

SOURCE CODE

VS CODE

OPEN

CODE

YOUR
APP

YOUR
DB

YOUR
SVC

CODER

LB

opencode--<workspace>--<user>.coder.felix.vpn
🔓/🔒
  • Preinstalled Tools (claude, opencode, ...)
  • Preconfigured ENV (gh, glab, git, ...)
  • Preconfigured Git (email, name, co-author)
  • Workspace Apps [cool url] (opencode server)
  • Management Scripts (opencode-ctl)
  • Startup Scripts (repo cloning)
  • Persistent Volumes (per user)
  • Port Sharing (opencode server)
  • Docker-in-Docker (DIND)
  • Open-in-Coder Feature (deeplink)
github.com/ynfra
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM

✅

✅

✅

✅

AI LOOP
  1. Analyse phpstan errors
  2. Find top 30 easy errors
  3. Fix one by one (skip blockers, add comment)
  4. Run phpstan
  5. Commit
  6. Move to next error

Text

LOOP
HARNESS
  1. Coder workspace
  2. a) Systemd timer trigger opencode API
b) Claude /loop trigger opencode API
c) Claude /loop with LLM router
d) <anytool> /loop

  3. AGENTS.md - workspace instructions
  4. GOAL.md - (loop) what we are doing
  5. MEMORY.md - current progress/decisions

 

NIGHT SHIFT
PHPSTAN
NIGHT SHIFT
TEST COVERAGE
NIGHT SHIFT
REFACTORING

<FOOBAR>

OPECODE
OPENCODE
OPENCODE
OPENCODE
NIGHT SHIFT
ORCHESTRATOR
OPENCODE
Check remote night shifts
and get me current state
/loop 30m Check remote night shifts. Wake idle agents. Fix stuck work. Make them continue.
INNER/OUTER LOOP
  1. Analyse & research
  2. Prepare impl plan
  3. Grill my plan
  4. Implement it (...)
  5. Cover by tests
  6. Verify feature
  7. Document it
  8. Commit & push

Text

INNER LOOP
HARNESS
  1. Coder workspace
  2. Systemd => OpenCode
Claude /loop
  3. AGENTS.md
  4. PLAYBOOK.md
  5. GOAL.md
  6. MEMORY.md

 

1. Browse sentry errors and pick 10 most errors
2. Browse issue backlog and find related tasks
3. Browse user feedback and find related feedbacks
4. Browser error logs and find related ones
5. Sort by priority, implement each task in own subagent.
OUTER LOOP
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM

✅

✅

✅

✅

✅

AI EXTRAS
github.com/decolua/9router
9router

github.com/diegosouzapw/OmniRoute
omniroute
{
    "$schema": "https://opencode.ai/config.json",
    "model": "9router/cx/gpt-5.6",
    "provider": {
        "9router": {
            "npm": "@ai-sdk/openai-compatible",
            "options": {
                "baseURL": "http://9router.yourserver.tld/v1",
                "apiKey": "sk-your-key"
            },
            "models": {
                "cx/gpt-5.6": {
                    "name": "GPT 5.6 Sol",
                    "limit": {
                        "context": 1050000,
                        "output": 128000
                    }
                }
            }
        }
    }
}
~/.config/opencode/opencode.json
export ANTHROPIC_BASE_URL=http://9router.yourserver.tld
export ANTHROPIC_AUTH_TOKEN=sk-your-token
export ANTHROPIC_MODEL=cx/gpt-5.6

claude

GITLAB REVIEWS
Powered-By
OpenCode
+-1000 lines
OhMyFelix

OhMyFelix

12.2.2026 was born
15.2.2026 was banned
16.2.2026 was born (again)
github.com/samanhappy/mcphub​
MCPHub
mcphub.app
  1. Plan Before Code
  2. Fresh Context
  3. Done = Tested
  4. Rules in Repo
  5. Ask Before Destroy
  6. Human in the Loop
  7. Right Tool, Right Job
  8. No Secrets in Prompts
BASIC RULES
Recap
Recap
AI SETUP
AI TOOLS
AI READINESS
AI LOOP
AI PLATFORM
AI EXTRAS

AGENTS.md, SPECS, DOCS, SKILLS

Good class, how write tests,

codestyle, linters, run migrations

Claude, Claude Desktop, Codex, OpenCode Server

Coder, Jetbrains, custom templates (ynfra)

9router/omnirouter, gitlab reviewbot, OhMyFelix, Basic Rules, MCPHub, Matt Pocock (skills), 🍗🔥

/loop <time> <task>, GOAL, MEMORY, Claude to Opencode, Opencode to Opencode, Gitlab issues (task), comments (memory)

START LOOPING

 

FUN WILL COME

Thank you!

@xf3l1x
f3l1x.io

Thank you!

@xf3l1x
f3l1x.io