Superintelligence
KS
Systems & Craft · Seattle, WA · Contributor since 2024

K. R. Sterling

Principal Systems Architect, Distributed Runtimes

Distributed systems engineer and runtime architect with over twenty years designing high-concurrency low-latency infrastructure across hyperscale data centers. Formerly staff infrastructure architect optimizing distributed model serving and high-bandwidth interconnects. His writing demystifies memory arithmetic, speculative execution, hardware bottlenecks, and the disciplined craft of sustainable systems engineering.

Focus Areas: Distributed Inference, Memory Hierarchies, MoE Routing, Kernel Optimization

Published Essays

3 essays by K. R. Sterling

All Authors & Contributors →
October 21, 2026 · 9 min read · Distributed Systems Engineers & Inference Architects

The Co-Evolution of Task and Harness: Why Lean Execution Models Are Leaving the Chat Window Behind

For years, foundation models were optimized to produce pleasant conversational prose inside browser windows. The emergence of closed-loop self-scaffolding systems like Ornith-1.5 reveals a different trajectory: lean parameter footprints tuned directly for terminal harnesses, sparse tool use, and verifiable assertions.

Core Principle: When post-training couples task proposal, verification harnesses, and solution rollouts through closed-loop group relative policy optimization, models shed their conversational bloat. Lean 35B Mixture-of-Experts architectures activating only 3B parameters can match monolithic weights on complex software engineering benchmarks simply by honoring execution feedback over rhetorical flair.
Read Essay →
October 16, 2026 · 8 min read · Distributed Systems Engineers & Architects

JEV-as-a-Judge: Accept When Confident, Escalate When Unsure

Routing every production evaluation through a heavyweight autoregressive model burns cluster budgets on redundant monologues. Here is how typed System One evaluators and confidence-routed cascades restore discipline to the inference critical path.

Core Principle: Most production evaluations do not require an autoregressive chain-of-thought monologue. By pairing a single-pass System One evaluator like Jev ($0.042 per million tokens read) with a confidence-routed cascade, architects can resolve clear decisions at 10x to 300x lower latency and reserve heavyweight System Two reasoning strictly for ambiguous edge cases.
Read Essay →
October 13, 2026 · 7 min read · For Developers & Systems Architects

The Architecture Beneath the Hype: Why Sparse Routing and Memory Arithmetic Won the Day

The era of throwing $100M compute clusters at monolithic dense models has hit an economic brick wall. Here is how modern Mixture-of-Experts and speculative decoding actually work under the hood.

Core Principle: Frontier performance is no longer won by brute-force parameter expansion, but by decoupling parameter knowledge (671B) from FLOP cost per token (37B active) through elegant routing mathematics.
Read Essay →