10X WORKFLOW // BRIDGE // PORTAL MAP

10XDEVS 4.0 // AI-NATIVE SOFTWARE ENGINEERING

Bridge 4.0: learn to ship complete changes with AI agents - from research to review.Learn to ship complete changes with AI agents — from research to review.

10xDevs 4.0 is a practical five-week program for developers and teams who want to use AI agents in real projects, including existing codebases. You learn to prepare context, plan a change, drive implementation and prove its quality. This page maps the complete method, tools and program outcomes.

CONTEXTstored in the repo, not chat memory
LOOPresearch → plan → implementation → review
EVIDENCEtests and gates instead of agent claims
01FROM PROMPTS TO PROCESS

AI is fast. Without a system, it produces chaos faster.

One good model response does not scale across a project. The next session forgets the decisions, a new agent scans the repository again, and “done” often means only “code was generated.” 10x Workflow stores context, decisions and quality gates in the repository.

Marcin and Przemek supervise an AI agent moving one software change through a controlled process
ONE CHANGE // A CLEAR PROCESS, VISIBLE PROGRESS, HUMAN CONTROL
CHAT FROM ZEROCONTEXT IN THE REPO

The project map, rules and patterns become shared infrastructure. Every agent starts from the same point.

“BUILD THE FEATURE”SMALL ARTIFACTS

Research, plan, implementation and review leave a trail on disk. Decisions outlive a single session.

“LOOKS GOOD”A HARD GATE

Tests, lint, build and review decide whether the work is finished.

The shift:

you build an environment where the agent starts with the right context and ends with a verifiable result.

02WHAT 4.0 CHANGES

From using AI to designing the entire working system

The new generation of tools is only the context. The product is one portable process in which an agent can drive a change end to end while the human retains control through explicit decisions, artifacts and quality gates.

Main frame
Building a full-stack product with AI
A system for driving an entire change with an agent
Way of working
Separate techniques and tools
One process for most code changes
Quality
Tests and automation before deployment
Risk-based checks and tests proven with an intentional failure
Existing code
Project onboarding and modernization
Decisions based on change history, dependencies and production evidence
Team
Each developer experiments alone
Shared rules, automated review and versioned instructions
Outcome
A product and a more mature AI toolkit
A product, process and versioned artifacts in the repo
For edition 2.0 alumni (October 2025) — 2,500 graduates:

you do not need another greenfield MVP. Bring your previous project or a repository from work and turn it into an environment ready for today’s agents.

03SYSTEM ARCHITECTURE

Five layers of one working system

10x Workflow joins things that usually live apart: project knowledge, agent instructions, delivery stages, automated verification and continuous repository improvement. Each layer strengthens the next. The Hub is the documentation of this method.

01 // MAP

Record the project map

The project describes its structure, risks and good examples. The agent receives only the context needed for the current task.

02 // SKILLS

Turn knowledge into skills

Recurring tasks become skills with a clear input, output and boundaries, ready for the whole team.

03 // WORKFLOW

Drive the change in stages

Understanding the problem, planning, implementation and review are not one leap. Every stage leaves its own document or result.

04 // GATES

Stop work at quality gates

Tests, type checks, the build and project rules decide whether work can continue. An agent’s claim is not enough.

05 // LOOPS

Make one change at a time

Small iterations select a target from risk evidence and make one safe change at a time.

Marcin Czarkowski and Przemek Smyrdek operate a holographic six-step agent workflow loop in a cosmic command center
CORE SKILLS CHAIN // DECISIONS AND ARTIFACTS INSTEAD OF ONE LONG SESSION
The mechanism:

context → decision → artifact → gate → next decision. The result no longer depends on one long, fragile conversation.

04MODULES 00–05 + SCALING

The complete program: from a personal environment to team scale

Every part ends with a capability you can demonstrate on your own code. You build a new way of working, from project preparation to safely scaling agents across a team. You are not doing it alone: every week brings a live session with the authors, office hours and the course community on Circle.

  1. 00 // PREWORK

    Choose the environment

    15 lessons, 4 sections and a quiz help match the model, tool, permissions, cost and context to the task.

  2. 01 // FOUNDATION

    Prepare the project

    Turn an idea into clear requirements, compact project rules, a change roadmap and a deployment-readiness check.

  3. 02 // WORKFLOW

    Drive the change

    Move through six explicit stages, from defining scope to saving lessons, with a clear result at every stage.

  4. 03 // QUALITY

    Prove risk protection

    Select the highest risks, deliberately break behavior and verify that the test detects the problem.

  5. 04 // EXISTING CODE

    Defend an architecture decision

    Combine a project map, change history and dependencies to choose a safe modernization path.

  6. 05 // TEAM

    Build a shared standard

    Move private instructions into versioned team rules, automated review and a safe execution environment.

  7. SCALING

    Increase autonomy without losing control

    Agents can work longer and in parallel, but stop after a failed check and return control to a human.

Marcin and Przemek inspect a holographic map of a large repository with modules, dependencies, a feature path and highlighted hot spots
LEGACY // DETERMINISTIC MEASUREMENT, AGENT INTERPRETATION, HUMAN DECISION
05THE DAY-TO-DAY EFFECT

What changes once the workflow is in place

Every change leaves context, a plan, code, tests and review in the repository. You can resume work, switch models or hand the task over without reconstructing decisions from chat history.

Faster start
The agent does not begin with archaeology. It reads the map, contracts and nearest patterns, then spends the context budget on the real problem.
Less drift
The plan and artifacts hold the direction. It is easier to resume, switch models or hand work over without losing decisions.
Safer changes
Small scope and automated gates limit the blast radius. A failure becomes input for the next iteration, not something to hide.
Team scale
Knowledge written once keeps working. The same rules, skills and patterns support the next developer and the next agent.
Marcin and Przemek lean over project preparation artifacts: a PRD checklist, a roadmap, an AGENTS.md file and a deployment readiness panel
ARTIFACTS IN THE REPO // PRD, ROADMAP, AGENTS.MD AND A DEPLOYMENT READINESS PANEL
Check it in the Hub:

open the Core Skills Chain to inspect the input, artifact, mechanism and boundaries of every stage.

06THE ECOSYSTEM AFTER THE COURSE

The Hub, 10x-cli and 10xBench solve three adoption problems

Together they give you three elements needed to adopt the method: a place for guidance, workflow delivery into the repo, and evidence-based model comparison.

DOCUMENTATION

10xDevs Hub

Permanent, searchable documentation of the method: modules, skills, mechanisms, boundaries and examples.

DISTRIBUTION

10x-cli

Delivers and updates rules, commands, skills and resources without manually copying fragments across projects.

MODEL DECISIONS

10xBench

Compares models on the same production task, including the result, code and final product image.

Marcin and Przemek connect method documentation, repository resources, model comparison and a practice route
10XDEVS HUB + 10X-CLI + 10XBENCH // KNOWLEDGE, DISTRIBUTION AND EVIDENCE-BASED DECISIONS
You can move into your own repo
  • analysis grounded in code evidence
  • a plan with clear scope and measurable criteria
  • review showing plan compliance and unapproved changes
  • a risk map and tests verified with an intentional failure
  • an architecture report for an existing system
  • a versioned set of team instructions, commands and rules
  • automated checks for autonomous and parallel work
07PROVEN IN PRACTICE

The workflow was proven across the full 3.0 cohort

10x Workflow was introduced in edition 3.0. In 4.0, you use its expanded version, adapted to today’s tools, longer-running agent tasks and parallel work.

6700graduates across three editions
2915graduates of edition 3.0
44%of graduates are senior developers
1500projects submitted by July 27, 2026
Marcin and Przemek supervise three isolated parallel agent work streams with separate gates and human review points
SCALE TIER // MORE PARALLELISM, THE SAME GATES AND AN EXPLICIT RETURN OF CONTROL

LEARNER VOICES // REAL PROJECTS

A system matters only when it works beyond the course

These voices come from the public 10xDevs website and project submissions in 10xForge. They show how the workflow performed at work and in personal products. Quotes remain in their original language.

WATCH THE STORY

„Na dzień dobry w pracy rzeczywiście poczujesz różnicę.”

Kacper KustraSoftware Engineer
WATCH THE STORY

„Jestem w stanie (…) zmierzyć się z problemami dużo większymi.”

Mateusz Jarzębowski-BownikSoftware Engineer

FROM PROJECT SUBMISSIONS

Nauczyłem się konfigurować środowisko tak, żeby AI pracowało dla mnie, a nie na odwrót.

Mariusz SzustkaSenior SysOps Engineer · PSE Innowacje10xDevs.pl
MORE PROOFRead all graduate stories44% of graduates are senior developers. See recent 10xDevs 3.0 LinkedIn posts, recordings, projects and feedback.

Look inside: demonstration pages and the complete Hub map

Evidence inside · start with these pages

The complete Hub 4.0 map

08COMPARED WITH OTHER SYSTEMS

A lightweight core, specialist skills and the complete change cycle

10x Workflow is an original, maintained system. It draws on patterns known from the tools below — isolated agent environments, spec before code, prepared context — and combines them into one portable process: a simple flow for most changes and specialist skills for quality, legacy, teamwork and controlled autonomy.

HumanLayer

WHAT IT DOES WELL

Coordinates many agent tasks and isolated working environments.

10X WORKFLOW

10x Workflow adds a portable method, shared documents and quality checks that work beyond one tool.

OpenSpec

WHAT IT DOES WELL

Helps agree on a change specification before an agent writes code.

10X WORKFLOW

10x Workflow extends that scope with repository discovery, implementation, plan compliance, quality and teamwork.

OpenMercato

WHAT IT DOES WELL

Provides an application foundation with prepared context and rules.

10X WORKFLOW

10x Workflow applies that principle in a process independent of domain, framework and programming language.

Compare the scope and working model of each system. Every named project develops independently.

09WHO 4.0 IS FOR

For developers using AI in real repositories

The program is universal and stack-independent. You do not need to be a full-stack expert: commercial experience and solid knowledge of your own specialty — backend, mobile or data, for example — plus experience working on a real product are enough. It teaches the layer above: how to prepare the project and process so an agent becomes a predictable participant in engineering work.

DEVELOPER

You want control back

You use agents, but results vary, sessions drift, and fixing generated code consumes the promised productivity gain.

TECH LEAD

You want a common standard

You need shared rules, repeatable review and a way to move team knowledge into the agent environment.

TEAM

You want knowledge to scale

You do not want to pay the repository-discovery cost on every task. Context, skills and gates should work for everyone.

Requirements:

this program is for people with commercial experience, but you do not need expert frontend and backend knowledge at once. If you know your specialty well and want to build your own repeatable workflow, you are in the right place.

10THE NEXT STEP

Explore the 10xDevs 4.0 program

The program takes you through a complete change on your own code: project preparation, the Core Skills Chain, tests, review and a shared team standard.

Check the program scope, working format, requirements and current enrollment terms.