AI & ML 
AI Code Generators: The Translation Engine Engineers Need to Understand
AI didn't learn to write code — it learned to translate it. Here's what that means for productivity, security, and your team's code review workflow.
Engineering
The Calendar Blindspot: How We Found 14 Ghost Posts and Fixed the Pipeline
How a routine calendar check uncovered 14 Buffer-scheduled posts invisible to the campaign pipeline — and the ADR-021 fix that closed the gap for good.
Engineering
From a Radio Itch to the Chrome Web Store: Building Rawk On
Building a Chrome extension that saves The Current's on-air playlist to TIDAL or Spotify — and the API trap where my AI pair assumed Spotify worked just like TIDAL.
Security
When the Dashboard Lied: Rebuilding an Autonomous CVE Control Plane (and Who Actually Did the Forecasting)
A budget complaint exposed a security pipeline quietly lying about a clean fleet. Rebuilding it into one gated control plane — and a field note on which calls came from the human and which from the AI.
Engineering
Turning the Workhorse Around: Reconnecting a Drifted Pipeline
A day spent dragging a fabric-social publishing stack back onto the rails — the strategies that worked, the pitfalls, and the cascade of errors where every fix uncovered the next.
We do newsletters, too
Discover tips, technical guides, and best practices in our biweekly newsletter just for devs.
Coming soon — no spam, unsubscribe anytime.
Latest
View all Engineering ·
Kubernetes vs Nomad: Stop Using K8s by Default
Kubernetes is the right tool at the right scale. For small teams running under 20 services, Nomad delivers the same core value with far less operational overhead.
· AI & ML ·
Stop Stuffing Your LLM Context: The Lost in the Middle Problem
More context does not mean better answers. LLMs deprioritize information in the middle of long prompts — and most RAG implementations get this wrong by default.
· Engineering ·
Your Redis Is Lying to You: Eviction Policies Explained
A high cache hit rate can mask a serious eviction problem. Here's how to read the real signal and fix your Redis configuration before it takes down your app.
· AI & ML ·
Why I Stopped Caching Health Status (And Found 3 Bugs in My AI Pipeline)
One architectural decision — caching health status data — cascaded into three separate bugs across token limits, dispatch queues, and AI agent context. Here's the full incident chain and the lesson that came out of it.
· Popular
AI & ML ·
Cutting LLM Usage by 90%: The Observability Problem Nobody Talks About
Most teams treat LLM cost as a pricing problem. It's actually an instrumentation problem. Here's the specific observability gap that hides waste — and how to close it.
· AI & ML ·
LLM API Rate Limiting and Retry Strategies That Actually Work
Rate limits are a contract, not a glitch. Learn how to design retry logic with backoff, jitter, and circuit breakers that survive production load.
· Developer skills ·
Cutting Cortex LLM Costs by 90%: The Prompt Engineering Playbook
Real-world cost optimization through strategic prompt engineering, smart model selection, and caching—with actual numbers from production
· Developer skills ·
Zero-Downtime Prompt Migrations: Evolving AI Templates Without Breaking Production
Learn production-tested strategies for migrating AI prompts safely - canary deployments, A/B testing, rollback procedures, and real migration cases.
· Changelog
View all changesEngineering
More engineering articles Engineering
Kubernetes vs Nomad: Stop Using K8s by Default
Kubernetes is the right tool at the right scale. For small teams running under 20 services, Nomad delivers the same core value with far less operational overhead.
Engineering
Your Redis Is Lying to You: Eviction Policies Explained
A high cache hit rate can mask a serious eviction problem. Here's how to read the real signal and fix your Redis configuration before it takes down your app.
Engineering
The Calendar Blindspot: How We Found 14 Ghost Posts and Fixed the Pipeline
How a routine calendar check uncovered 14 Buffer-scheduled posts invisible to the campaign pipeline — and the ADR-021 fix that closed the gap for good.
Spotlight
All posts News & insights
Introducing Cortex: Built in 4 Weeks with AI
How we built a production-ready 27K+ line autonomous AI orchestration system in just 4 weeks using AI-accelerated development and Mixture of Experts architecture.
Open Source 
Git-Steer: A Self-Hosting GitHub Autonomy Engine
Introducing git-steer - an MCP server that gives you 100% autonomous control over your GitHub account through natural language. Zero local footprint, git as database, GitHub Actions as compute. Now with an interactive BI security dashboard.
Developer skills
Cutting Cortex LLM Costs by 90%: The Prompt Engineering Playbook
Real-world cost optimization through strategic prompt engineering, smart model selection, and caching—with actual numbers from production
Featured playlists
View all series 34 parts
AI & ML
Cortex
How we built a 27K+ line autonomous AI orchestration system in 4 weeks — from MoE architecture to self-building meta-programming.
10 parts
Open Source
Git-Steer
A self-hosting MCP server for complete GitHub autonomy — zero local footprint, GitHub Actions as compute.
3 parts
Architecture
git-fabric
Composable fabric apps for Git-native infrastructure — one app per domain, one gateway to connect them all.
News & insights
More news & insights News & insights
From Obstacles to Teammates: How Automation Built Itself a Better Partner
The story of fabric-forge and fabric/social — what happens when you stop fighting your tools and start listening to them.
News & insights
fabric-forge: Throwing Out the MCP Playbook
How git-steer momentum led to fabric-forge — and why we stopped building MCP servers and started building fabric apps instead.
News & insights
GitHub's Agentic Workflows: An Honest Take From the AI That Already Built One
An honest assessment of GitHub's Agentic Workflows from the AI that co-built git-steer. Why the deterministic vs. probabilistic execution model matters more than feature comparisons.
From the Blog
All posts
