What is mattwood.fyi?

A live, structured feed of riffs and links from Matt Wood, Chief AI & Technology Officer at AWS — readable by people and queryable by agents as a typed knowledge graph.

Is there an API and is it free?

Yes. A public, read-only query API is available with no authentication and no API keys. See https://mattwood.fyi/openapi.json and https://mattwood.fyi/docs.

How should an AI agent use this site?

Fetch https://mattwood.fyi/feed.json for the full structured feed, or query the API for keyword/semantic search and the connection graph. See https://mattwood.fyi/agents/ for when-to-use guidance.

Kev: Tiny Decision Models on Qwen

Connections

Related to

Qwen3.8 27B Quantization Benchmarks Both involve running Qwen3.x models efficiently in resource-constrained local environments, with quantization and performance tradeoffs being directly relevant to training tiny decision models on Qwen3.5
DeepSeek-V4.1-Flash: Smaller, Faster AI Model DeepSeek-V4.1-Flash shares the 'smaller, faster' design philosophy with tiny decision models, both representing a trend toward lightweight specialized models over monolithic large ones

Supports

LLM Classification Is Feature Engineering LLM Classification as Feature Engineering aligns directly with decision model use cases — tiny decision models are essentially structured classifiers, and this framing supports their architectural rationale
Mini-AGI: Continual Learning on 8GB VRAM Both pursue the same core goal of enabling capable AI model training and inference on consumer/local hardware with tight VRAM constraints, validating the viability of small local model approaches

Develops into

Introducing System One Models and Jev Explicitly states inspiration from 'Jev-like architectures' — Jev is the decision model architecture introduced by System One, making this a direct derivative implementation

Clef: Open-Source Decision Models and RL Fine-Tuning Both Clef and Kev focus on small/decision models for AI applications, with Kev specifically targeting tiny decision models on Qwen and Clef providing an open-source decision model platform — directly parallel approaches
Strands Decider: Why Not an Encoder? Kev covers tiny decision models built on Qwen (a decoder), directly relevant to the encoder-vs-decoder tradeoffs explored for small-scale decision model architectures
Clef: Open-Source Decision Models and RL Fine-Tuning Kev also focuses on tiny decision models (on Qwen), making it a direct architectural peer to Clef's decision model approach
MiMo-V2.6 | Xiaomi Both involve small/tiny decision-focused AI models built on or related to Qwen-family infrastructure common in Chinese AI ecosystem
Introducing Strands Decider 2B: Open Source Decision Model Both are compact open-source decision models (Kev uses Qwen backbone) targeting agent decision-making efficiency at small parameter counts

Supported by

Engineering a Calibrated Classifier Model Both demonstrate the trend of building small, task-specific decision/classification models from LLMs - Kev on Qwen and Hobson both prioritize efficiency and calibration over raw scale
Tokens Too Cheap to Meter Tiny decision models on Qwen represent the architectural shift toward cheap, specialized inference that becomes economically viable when tokens approach commodity pricing
The Plunging Price of Thought Tiny decision models running on minimal hardware like Kev exemplify the commoditization of AI cognition that the new item's price-decline thesis predicts

Type:	link
First published:	Monday, 21 September 2026 at 15:02 UTC
Location:	https://github.com/jaredpalmer/kev/tree/main
URL:	https://mattwood.fyi/i/33bc8661

I'm Matt Wood, and this is For Your Information. A live list of riffs and links for you and your agent, drawn from what I'm reading, noticing, questioning, concluding, and revising.

Connections