Run Local LLMs on Mac to Cut Claude Costs
A practical hybrid workflow that uses costly LLM APIs for planning and local models (via Ollama + OpenCode) for execution, guarded by deterministic evals.
Speedscale Blog
Latest insights on Agentic AI workflows, cloud native architectures, and performance optimization best practices from the Speedscale team.
Featured Post
Recent Posts
What the Warp Terminal Sends Home
We recorded Warp traffic to see what gets sent back to the home base. Spoiler: It's everything.
Replace API Synthetics with Traffic Replay
UI synthetics only tell you something is broken. Traffic replay per microservice isolates failures before any human walks up. Zero scripts required.
Beyond AI Vibes: Deterministic Foundations for Agentic Coding
AI coding adoption is high and trust is dropping. A testing pyramid for agents, plus reproducible production context that grounds AI in real behavior.
Dark Code: The AI-Generated Software Nobody Understands
Dark code is software no human has written, read, or reviewed. As AI tools accelerate, the gap between shipped code and understood code is widening fast.
OpenTelemetry Trace Testing for CI Release Gates
Trace-based testing uses OpenTelemetry traces as replayable test input so CI catches production regressions before deploy, not after incident review.
When Your Observability Literally Stops Traffic
Observability tells you what failed—but not how to recreate it. Why reproducibility is the missing fourth pillar, and what that means for incident response.
Why Autonomous AI Agents Can't Run on SaaS Infrastructure
SaaS AI fails when agents need continuous access to your codebase and internal APIs. Here's why BYOC is the only deployment model that works at scale.
From Datadog to CI Tests: Catch Regressions Before Deploy
Learn a practical workflow to convert Datadog metrics, traces, and incidents into CI tests that catch regressions before deploy.
Why You Should Stop Buying SaaS and Start Building It
LLMs have collapsed the cost of custom internal tools. Here's the startup distribution problem I've watched kill companies — and how I vibe-coded my way out.