Blog

Updates, insights, and wisdom from the Specularis team.

March 23, 2026
The need for a Coordination Layer in Agentic Software Development
TL;DR
As AI makes code generation cheap and abundant, the bottleneck in software development shifts from implementation to coordination. In an agentic environment, where multiple humans and AI systems work on the same codebase, teams need a shared, persistent source of truth for intent, constraints, and architecture. That is why a coordination layer becomes essential.
Benedikt Bingler
March 10, 2026
Human governance for living specs
TL;DR
As AI agents begin editing specifications alongside humans, software teams need a new governance layer for living specs: full history, provenance, conflict resolution, and easy restore. The core issue is not whether specs can evolve faster, but whether teams can still track who changed what, resolve overlaps safely, and keep humans in final control. Research suggests AI can help with merge and conflict resolution, but not reliably enough to remove human approval from the loop. The winning model is AI-assisted, human-governed collaboration, where specs remain dynamic without becoming chaotic. Teams that build this governance early will scale agentic development with far less risk, confusion, and rework.
Benedikt Bingler
March 1, 2026
Canonical specs that agents trust
TL;DR
Software engineering’s “source of truth” has shifted repeatedly over the past decades - from centralized version control to Git, API contracts, and GitOps. The rise of AI coding agents may trigger the next shift: prompts, specs, schemas, and tests increasingly become the primary inputs, while code becomes a generated artifact. Today these sources of intent are fragmented across docs, repos, and tools, leading to drift that AI agents can amplify. A better model is a canonical, machine-readable spec workspace where both humans and agents operate from the same authoritative context. Teams that establish a single source of truth for specifications will be far better positioned for AI-driven development.
Benedikt Bingler

The need for a Coordination Layer in Agentic Software Development

Human governance for living specs

Canonical specs that agents trust