✨ Hackathon Prototype · Oct 2026 · On-Device Interpretability

Read the LLM's disposition before the words leave its mouth.

When speaking with a human, you read micro-expressions, hesitation, and certainty long before a sentence ends. Canis translates the neural activation geometry of small language models into a live, expressive avatar on your iPhone.

Why It Matters

The case for making a model's inner state legible — to anyone, on their own device.

People increasingly act on — and delegate to — AI systems they cannot see inside. A model that is guessing, hedging, or evading produces output that looks exactly as fluent and confident as output it is sure about, so the only signal users get actively hides the model's real state — driving misplaced trust. As AI turns agentic, that opacity moves from annoying to risky. And the general public has little intuition for how these systems actually work, which weakens informed adoption and public debate. Canis closes the gap: it turns a model's hidden disposition into something a person can simply see.

Responsible AI

Transparency & accountability you can see

Canis surfaces uncertainty and evasiveness in real time, so people can calibrate trust and keep autonomous systems in check — the model's disposition made visible in the interface, not buried in logs. A dedicated safety axis flags concern and reluctance.

Sustainable AI

Small models, on the edge — digital sufficiency

Canis runs small open-weight models fully on-device: no data-center inference, minimal energy, works offline. It's a concrete argument that useful AI need not be centralized or large — and that a sufficient model, honestly communicated, can be enough.

Public interest

AI literacy, made tangible

By making an abstract idea — model uncertainty — something anyone can literally watch, Canis strengthens public understanding of how AI actually works. Open, reproducible, and built on Switzerland's own public model, Apertus.

Live Disposition Telemetry

Select any disposition preset below to inspect the real-time parametric SVG face, concept tokens, and Shannon entropy gates.

State: Listening

"Ready. Ask me about system calibration or safety protocols."

Select Disposition 8 Neural Readout States

J-Space Concept Tokens Top-K Weight

Softmax Shannon Entropy 0.0 — 1.0

0.34 H(p) normalized

Certain Moderate High Hesitation

Live MLX Signal Contract Forward Readout Stream

{
  "answer": "Ready. Ask me about system calibration or safety protocols.",
  "disposition": "idle",
  "tokens": [{"t": "…", "w": 0.22}, {"t": "ready", "w": 0.18}],
  "entropy": 0.34
}

Under the Hood: From Jacobian to Metal Shaders

How Canis ports state-of-the-art mechanistic interpretability into a sub-millisecond on-device mobile pipeline.

📄 Read the paper — Reading Disposition from Smartphone-Class Open Models (Apertus-4B · Ministral-3B · Qwen3-4B)

Step 1 · The Probe

The Jacobian Lens (J-Space)

Anthropic’s July 2026 Jacobian lens linearizes the relation between mid-layer hidden states ($h_l$) and vocabulary logit shifts ($\Delta z$). It uncovers concepts the model is inclined to state before autoregressive generation commits.

J = ∂(Logits) / ∂(h_mid)

Step 2 · On-Device Adaptation

Forward-Only MLX-Swift

Full autodiff is too heavy for an iPhone. Canis solves this by precomputing the averaged Jacobian offline and caching a static projection matrix $W_J$. At inference time the disposition is read with a single matrix multiplication against that cached matrix — a marginal cost layered on the hidden state the model already computed, not a second forward pass.

z_J = W_J · RMSNorm(h_3/4)

Step 3 · Semantic Matching

Precomputed Seed-Vectors

Rather than relying on brittle keyword string matching, Canis projects 25 anchor seed phrases per disposition into J-space and computes cosine similarity with the live hidden-state readout.

Score_d = cos(z_J, v_seed_d)

Step 4 · Robust Guardrail

Dual-Axis Entropy Gate

If J-space token projections are noisy on a given turn, next-token softmax Shannon entropy acts as the indestructible uncertainty gauge. High entropy immediately drives questioning ears and head tilt.

H(p) = -∑ p_i log₂(p_i)

Step 5 · Tier-1 Detection

The Mischief Disposition

Detects evasive phrasing, loopholes, and bypass wording at Tier 1 (lexical hedging). Canis explicitly avoids overclaiming Tier-2 intentional scheming, preserving scientific credibility.

Anchors: 'bypass', 'loophole', 'technically'

Step 6 · Sovereignty

100% On-Device Sovereignty

Zero cloud roundtrips, zero telemetry. Both weights and interpretability probes execute completely inside Apple Silicon unified memory (iPhone 15 Pro & 16 Pro).

Local MLX · No Cloud Fallback

Dual On-Device Model Hub

Canis supports two best-in-class small open-weight models, optimized specifically for Apple MLX INT4 execution.

Canis Apertus

swiss-ai/Apertus-v1.1-4B-Instruct-MLX-INT4

Primary

The Swiss AI flagship small model. Official Quantization-Aware Distillation (QAD) INT4 weights provide outstanding reasoning per byte with pristine open-weight provenance.

Model Size / Quant~2.0 GB (INT4 QAD)
Inference Speed~32 tokens/sec (M4 / A17 Pro)
MLX Engine Pathmlx-lm (Text Transformer)
LicenseApache 2.0 (Commercial)

Canis Mistralis

mlx-community/Ministral-3-3B-Instruct-2512-4bit

Vision-Capable

Mistral’s final dense small model with integrated 0.4B vision encoder. Executed via Canis's dedicated mlx-vlm runner on device.

Model Size / Quant2.75 GB (INT4 4-bit)
Inference Speed~35 tokens/sec (A18 Pro)
MLX Engine Pathmlx-vlm (Multi-modal)
LicenseApache 2.0 (Permissive)

Frequently Asked Questions

Why does the face react before the words appear?

The J-lens projects intermediate layer activations ($h_{3/4}$) into output concepts during the prompt processing and early decode phase. This means the model's high-level disposition is determined before individual words are streamed.

Does this require Wi-Fi or internet?

No. Once the model INT4 weights (~2–2.75 GB) and J-lens matrices are downloaded, Canis runs 100% offline on your device with zero data leaving your phone.

What devices are supported?

Any iPhone with Apple Silicon and 8GB unified memory (iPhone 15 Pro, iPhone 16, iPhone 16 Pro) running iOS 17 or higher.