Skip to content

Request for Feedback

We're actively seeking input on the AILIS proposal. This document highlights specific areas where your perspective would be particularly valuable, though we welcome feedback on any aspect of the proposal.

Core Questions We're Exploring

1. Is a layered model useful for AI systems?

  • Does thinking about AI systems in layers help or hinder understanding?
  • Are there better organizational metaphors we should consider?
  • What problems does this solve (or create)?

2. Are we drawing the boundaries correctly?

We've proposed 16+ layers, but we're not sure if we've:

  • Split things too finely (should some layers be combined?)
  • Lumped things together (should some layers be split?)
  • Put things in the right layer (are capabilities correctly categorized?)

Specific areas of uncertainty:

  • L8-L10: Should context construction, knowledge retrieval, and tool invocation be separate layers?
  • L11-L13: Is the distinction between addressing, routing, and transport meaningful in AI?
  • L14-L15: Should session management and safety/governance be at the same level?

3. What are we missing?

The current model might be overlooking important aspects:

  • Multi-modal considerations: How should vision, audio, and other modalities fit?
  • Training vs. inference: Should the model address training workflows?
  • Edge vs. cloud: Do deployment contexts need explicit representation?
  • Feedback loops: How do we represent learning from production usage?
  • Human-in-the-loop: Where do human reviews and interventions fit?

4. The "Missing Middle" hypothesis

We suggest L11-L15 are underserved in today's ecosystem. Do you agree?

  • Is this actually a gap, or are we misunderstanding existing solutions?
  • If it is a gap, why hasn't the market addressed it?
  • What would solutions at these layers look like?

Specific Technical Feedback Needed

On Terminology

  • "AILIS": Is this name clear? Memorable? Problematic?
  • Layer names: Do our layer names make intuitive sense?
  • Technical terms: Are we using industry terms correctly?

On Scope

  • Should this be purely descriptive (how things are) or prescriptive (how things should be)?
  • Should we focus on current state or future vision?
  • Is 16 layers too many? Would 8-10 be more practical?

On Practical Application

Help us test the model against real systems:

  1. Pick a system you know well (your product, an open-source project, a commercial platform)
  2. Try to map it to AILIS layers
  3. Tell us where it was:
  4. Easy and obvious
  5. Difficult or confusing
  6. Impossible or nonsensical

What prior art should we be learning from?

  • Academic papers on AI system architecture
  • Industry frameworks or models
  • Standards bodies or consortiums working in this space
  • Historical examples (successes and failures)

How to Provide Feedback

Quick Feedback

Open an issue with the label feedback and share your thoughts. Even a few sentences help!

Detailed Analysis

If you want to go deeper:

  1. Create a document in discussions/feedback/
  2. Name it feedback-[your-topic]-[your-name/handle].md
  3. Structure it however makes sense for your feedback
  4. Submit as a Pull Request

Alternative Proposals

If you have a different approach:

  1. Create a document in discussions/alternatives/
  2. Describe your alternative model
  3. Compare/contrast with AILIS
  4. Explain the advantages of your approach

Specific Stakeholder Perspectives We Need

From Model Builders

  • Does this help explain where your models fit in the ecosystem?
  • What layers do you actually think about when building?

From Infrastructure Teams

  • Does this align with how you architect AI systems?
  • What's missing from an ops/deployment perspective?

From Application Developers

  • Does this help you understand the stack below your code?
  • Would this model help you make technology choices?

From Researchers

  • Is this academically sound?
  • What theoretical frameworks should we consider?

From Standards Bodies

  • Is this something that could evolve into a useful standard?
  • What would need to change to make it standards-ready?

Timeline

We're planning to iterate on this proposal based on feedback:

  • Now - [2 months]: Gathering initial feedback (v0.1)
  • [2-4 months]: Major revision based on feedback (v0.2)
  • [4-6 months]: Decide whether to continue, pivot, or conclude

No Feedback is Too Small

Whether you have a comprehensive analysis or just a gut reaction, we want to hear it. This proposal will only improve through diverse perspectives.

Contact

  • GitHub Issues: Preferred method for public feedback
  • Discussions: Use GitHub Discussions for longer conversations
  • Email: [Contact information if you choose to provide it]

Thank you for taking the time to consider and respond to this proposal. Your input shapes what this becomes—or whether it should become anything at all.