Software development with coding agents.
Get a defined software outcome with one developer responsible for the architecture, implementation, and handoff. Coding agents support the work within an agreed scope.
Coding agents support delivery. Marc owns the technical decisions.
Agentic development means using AI coding agents to assist with tasks such as implementation, code exploration, and test preparation. Marc defines the architecture, bounds the work, reviews changes, and checks the resulting product against agreed acceptance criteria.
AethDesign is Marc Kubernus’s independent studio in Germany, working remotely with businesses. You work directly with the developer responsible for your project, from the initial repository review to the agreed handoff.
This service concerns how your software is developed. Adding a customer-facing assistant or an automated AI workflow inside your product is a separate AI implementation scope.
Where agentic development fits
A focused first product
Build a minimum viable product around a specific user task. Agree what the first version needs to demonstrate, which integrations it needs, and what stays outside the initial release.
Internal business tools
Turn a defined operational workflow into an interface, data model, and integrations. Include the permissions, error recovery, and day-to-day maintenance the workflow needs.
Changes to existing software
Start with the current codebase, conventions, and tests. Scope a feature, integration, or technical improvement that fits the architecture and preserves existing behavior.
From an agreed scope to reviewed software
The project starts with observable acceptance criteria and access boundaries. Progress is reviewed through working increments and the checks relevant to each change.
- Define the outcome
Review the user task, existing systems, constraints, and release needs. Agree on deliverables, exclusions, acceptance criteria, and who will maintain the result.
- Set the boundaries
Agree repository access, tool use, and how project information may be processed. Keep credentials and production access restricted to what the work requires.
- Build and review
Use coding agents for bounded tasks under Marc’s direction. Review code and test the relevant behavior, including failure paths and integration points.
- Validate and hand over
Check the agreed user journeys and document the result. Deliver the source, setup instructions, known limitations, and any deployment or maintenance work included in the scope.
Typical deliverables
- Agreed scope and acceptance criteria
- Implemented features in the project repository
- Relevant tests and validation evidence
- Technical decisions and known limitations
- Setup and handoff documentation
- Deployment preparation when included in scope
Questions and answers
Is this the same as building an AI agent for my business?
No. Here, coding agents help develop the software. Your finished product does not need to contain AI. Assistants, retrieval, and agent workflows embedded in your product belong to AI implementation.
Can you work in our existing repository?
Yes, subject to an initial review of the codebase and access requirements. We agree on the permitted tools, repository permissions, review process, and handling of project information before implementation.
Does using agents guarantee a faster or cheaper project?
No. Suitability depends on the codebase, clarity of the task, integrations, and verification needs. Estimates follow a scope review and account for human architecture, code review, and testing.
Who maintains the software afterward?
The handoff and maintenance responsibilities are agreed before work begins. The scope can include documentation for your team and a separate arrangement for ongoing changes or support.