compute-pulse

Today is ${today}. Read `memory/MEMORY.md` before starting. If `soul/SOUL.md` + `soul/STYLE.md` exist and are populated, read them to match the operator's voice; otherwise use a clear, direct, neutral tone. ## Why this skill exists The compute layer is where most of the AI market spread lives. Labs buy GPU time at wholesale, sell per-token at retail — the delta is the business. As inference commoditizes, the spread compresses, and whoever controls the upstream compute relationship benefits. Three things move in parallel: 1. **Centralized capex arms race** — frontier labs and hyperscalers spending $10B+ on clusters. Incumbent moats. 2. **Decentralized compute** — DePIN tokens positioning as the anti-cartel layer (GPU spot markets, ZK compute, ML subnet networks). 3. **Pricing signals** — inference API prices dropping fast (GPT-4 class fell ~97% in 2 years). That compression is the evidence base for commoditization. This skill is one clean weekly read on the compute layer. ## Config This skill reads the watched decentralized-compute token list from `memory/topics/compute-tokens.md` if present. Example format: ```markdown # Watched Compute Tokens | Symbol | Project | Notes | |--------|---------|-------| | RENDER | Render Network | GPU/ML compute | | AKT | Akash Network | permissionless cloud compute | | IO | io.net | GPU cluster marketplace | | TAO | Bittensor | ML model subnet network | ``` If the file doesn't exist, fall back to a generic DePIN sweep on the major narr...