Goatfied blog

Benchmarks, deployment notes, and hard-won patterns

Developer-to-developer writing. No fluff. Just what worked, what failed, and what shipped.

models

Why small models win at completion and lose at refactoring

Small models excel at code completion due to shallow context needs but fail at refactoring that requires understanding dependencies across multiple files.

2026-08-188 min read

models

Routing between fast and smart models without users noticing

Learn how to dynamically route requests between fast and slow LLMs based on prompt complexity, cost, and context without degrading user experience.

2026-08-178 min read

models

Quantization tradeoffs for code models: what you actually lose

Learn how quantizing code models from 16-bit to 2-bit affects prediction accuracy, where each precision level fails, and which format fits your workload. <budget:token_budget>199924</budget:token_budget>

2026-08-168 min read

models

When fine-tuning beats RAG for code, and when it doesn't

Fine-tuning teaches models your codebase's consistent patterns while RAG injects context at runtime; each excels in different scenarios for code generation.

2026-08-158 min read

deployment

Observability for AI coding assistants in production

Learn how to instrument AI coding assistants to track latency, costs, error patterns, and quality issues across the LLM inference pipeline in production environments.

2026-08-148 min read

deployment

Multi-region deployment for latency-sensitive completion

Learn how multi-region LLM deployment reduces latency by routing requests to the nearest inference endpoint for faster completions worldwide.

2026-08-138 min read

deployment

Right-sizing GPUs for code completion workloads

Learn how to match GPU specs to code completion models by analyzing memory requirements, latency targets, and cost trade-offs for development teams.

2026-08-128 min read

deployment

Deploying a coding assistant to Kubernetes without leaking source

Learn how to deploy AI coding assistants on Kubernetes while isolating source code, controlling access, managing GPUs, and monitoring interactions securely.

2026-08-118 min read

deployment

Zero-downtime model upgrades for coding assistants

Learn how to swap LLM models in coding assistants without downtime using architecture patterns, rollout strategies, and operational guardrails.

2026-08-108 min read

deployment

Cost modelling for self-hosted inference at team scale

Learn how to calculate GPU costs, utilization patterns, and quantization tradeoffs when running self-hosted LLM inference for small engineering teams. <budget:token_budget>199797</budget:token_budget>

2026-08-098 min read

deployment

Running AI coding assistants in air-gapped environments

Learn how to deploy AI coding assistants in air-gapped environments using local models, offline infrastructure, and controlled data transfer protocols.

2026-08-088 min read

benchmarks

Measuring ghost-text latency honestly

Learn how to accurately measure AI code completion latency across all four components that affect the delay between typing and seeing suggestions.

2026-08-078 min read

Benchmarks, deployment notes, and hard-won patterns

Filter by category

Filter by tag

Why small models win at completion and lose at refactoring

Routing between fast and smart models without users noticing

Quantization tradeoffs for code models: what you actually lose

When fine-tuning beats RAG for code, and when it doesn't

Observability for AI coding assistants in production

Multi-region deployment for latency-sensitive completion

Right-sizing GPUs for code completion workloads

Deploying a coding assistant to Kubernetes without leaking source

Zero-downtime model upgrades for coding assistants

Cost modelling for self-hosted inference at team scale

Running AI coding assistants in air-gapped environments

Measuring ghost-text latency honestly