Building AI that makes developers more capable, without giving up control.
WhoseDev is an AI development environment built by Anotiai, designed around a simple belief: increasingly capable AI should give developers more leverage without requiring them to surrender unnecessary control over their code, context, privacy, or tools.
WhoseDev is one part of a broader effort to explore how powerful AI can be made more understandable, controllable, secure, and practical in the environments where people actually work.
WhoseDev is one expression of a much bigger vision for AI.
Anotiai is focused on the broader challenge of making increasingly capable AI safer and more trustworthy to use. That means looking beyond model capability alone and thinking about the systems around AI: governance, security, privacy, protection, transparency, and meaningful human oversight.
The long-term vision extends beyond software development. As AI becomes more involved in scientific research, healthcare, engineering, business, and other high-impact environments, the surrounding technology needs to account for how that capability is controlled, protected, evaluated, and used responsibly.
Software is where we start.
Software development is already becoming one of the most important environments for applied AI. Developers use AI to understand code, generate changes, debug systems, reason through projects, and increasingly delegate parts of their workflow to agents.
But useful AI often needs useful context. That context can include source code, project structure, configuration, internal identifiers, credentials, and other information that may require careful handling.
WhoseDev gives us a practical environment to explore a different approach: AI that can work deeply with software while keeping privacy, security, model choice, and developer oversight closer to the center of the experience.
The hard questions start when AI gets access to real work.
The idea behind WhoseDev is not based on the assumption that AI is inherently unsafe or that developers should avoid using it. The question is what happens when increasingly capable systems are given access to the information and tools that make software development work.
These are representative scenarios that can occur in modern development workflows. They are examples of the kinds of problems that shaped our thinking, not claims that a specific incident happened to Anotiai or WhoseDev.
The production debugging problem
A developer is investigating a production issue and asks an AI system to analyze logs, configuration, or surrounding code. The information needed to solve the problem may also contain internal identifiers, infrastructure details, customer-related data, or credentials. The AI may be useful, but the developer still needs to understand what information is being shared.
The secret hidden in the project
A repository can contain environment files, API keys, tokens, private endpoints, or configuration values. Even when a developer never intentionally asks an AI system to see a secret, broad project context can create opportunities for sensitive information to be included in a request.
The agent with too much access
An AI coding agent can become significantly more useful when it can inspect files, run commands, modify code, and interact with development tools. But every additional capability raises an important question: what should the agent be allowed to access or change, and when should a developer be asked to intervene?
The wrong model for the wrong project
A cloud model may be the right choice for one task while a local model may be more appropriate for another. Teams can have different requirements based on their codebase, policies, infrastructure, or risk tolerance. A useful AI environment should not assume that one deployment model fits everyone.
The debugging conversation that contains more than code
Developers often paste error messages, stack traces, logs, screenshots, and configuration snippets into AI tools. Those materials can reveal much more about an environment than the developer intended. The useful question is not simply whether the AI can understand the information, but whether the developer understands what information is being exposed.
The moment automation becomes action
There is a meaningful difference between an AI suggesting a change and an AI making that change. As development tools become more agentic, permissions, approvals, visibility, and auditability become increasingly important parts of the experience.
The question behind WhoseDev
If AI is going to understand our work deeply enough to become genuinely useful, how do we give it that capability without giving up more information, access, or control than the task actually requires?
WhoseDev is our attempt to explore that question through a real developer environment rather than treating it as a theoretical problem.
AI capability is growing faster than the systems around it.
More capable models can reason across larger amounts of information and participate in increasingly complex workflows. That creates enormous opportunity, but it also makes questions about data boundaries, permissions, security, oversight, and accountability more important.
We believe these concerns should be considered while AI systems are being built, not treated as something to solve only after the technology becomes deeply embedded in people's workflows.
Context
AI systems often become more useful as they receive more context. That makes understanding and controlling information flows increasingly important.
Capability
As AI moves from suggestions toward agents and autonomous actions, the question becomes not only what AI can do, but what it should be allowed to do.
Trust
People and organizations need systems whose behavior, boundaries, and limitations can be understood rather than hidden behind vague assurances.
Privacy is a technical boundary, not just a promise.
WhoseDev is designed around the idea that developers should have greater visibility into how AI interacts with their environment. Local processing, model choice, privacy-aware handling, and developer-controlled workflows are part of that direction.
Workspace
Your code and project context
Local layer
Inspect and prepare context
Policy
Determine what may be shared
Model
Local or configured cloud model
Developer
Review and apply the result
This represents an architectural direction rather than a guarantee that every workflow follows the same path. Actual behavior can depend on product version, configuration, model, integration, and deployment.
An AI development environment built around control.
WhoseDev is being developed to help developers build, understand, debug, and modify software with AI. The goal is not simply to add an assistant to an editor, but to explore what a more privacy-aware and controllable AI development environment can look like.
AI-assisted development
Code generation, explanations, debugging, refactoring, and other AI-assisted development workflows.
Agentic workflows
Exploring AI agents that can reason through development tasks, work with project context, and perform actions while keeping developers involved.
Local and cloud models
Supporting different deployment approaches so developers can choose workflows that fit their technical and privacy requirements.
Privacy-aware processing
Designing workflows where sensitive context can be considered and handled before it becomes part of an external model request, depending on configuration.
Capability needs responsibility around it.
Privacy should be understandable
People should be able to understand where their information is processed and what happens when AI becomes part of their workflow.
Security should support usability
Protection should not require people to abandon useful technology. Good security and good product design should reinforce each other.
AI should remain governable
As AI becomes more capable, systems around it should make permissions, boundaries, decisions, and oversight more understandable.
Model choice matters
Different environments have different requirements. Developers should have meaningful choices about how and where AI capabilities are provided.
Transparency earns trust
We would rather explain what a system does, what it does not do, and what depends on configuration than make claims that sound better than the reality.
AI should assist human judgment
AI can be extremely useful, but useful does not mean infallible. People remain an important part of responsible decision-making.
AI should be powerful enough to help and responsible enough to trust.
WhoseDev starts with software development, but the thinking behind it is broader. Anotiai is exploring how increasingly capable AI can be developed and deployed with stronger consideration for governance, protection, security, privacy, transparency, and human oversight.
The long-term ambition reaches into environments where AI may have significant consequences, from software and scientific research to healthcare and other domains where reliability, privacy, safety, and responsible use matter deeply.
We are not presenting that future as solved. It is the direction we are building toward: stronger AI systems surrounded by better safeguards, clearer boundaries, and technologies that help people remain informed and in control.