NEWS

Smart engineering in practice.

Case studies, architecture notes, and the occasional rant — grounded in systems that actually shipped.

The Discipline Gap
6 min read

The Discipline Gap

DeepSeek V4 Pro beat GPT-5.5 Pro on precision. Not because it's smarter, but because it followed instructions. Here's what that score actually means for production engineering.

It’s 2 AM. Your pipeline broke again. The JSON that should have matched the schema matched it exactly for the first three runs. Then the model switched mid-prompt and started producing keys that were close, not correct. A schema validator failed silently in the background. Nobody caught it until the downstream system choked.
Read article →
FAIL: The Generalist Ceiling
6 min read

FAIL: The Generalist Ceiling

One agent trying to be everything hits a ceiling that looks exactly like burnout. The 7-week journey from generalist amnesia to a crew of 7 that …

The Real Cost of Inference
6 min read

The Real Cost of Inference

Headline API rates aren't the real cost. Cache hit rate, context window, and architecture implications matter more. Here's what to actually look at …

The Platform Trap
7 min read

The Platform Trap

When your platform owner competes with the products on their own shelves, the shelf gets narrower. Microsoft, GitHub Copilot, and Claude Code made it …

The Fresh-Start Paradox
7 min read

The Fresh-Start Paradox

Your reviewer starts fresh every run. Forgets everything you fixed. Flags the same issues again. That sounds broken until you realize a reviewer with …

A Crew on My Laptop
9 min read

A Crew on My Laptop

Seven AI agent profiles. One Linux box. A SQLite kanban board that routes work between them. No cloud service, no SaaS dashboard — just profiles, …

>_

Ready to build something that works?

Let's talk about your goals and how codegrit.dev can help you move work forward.

Start a Conversation →