Engineering Blog
This is where we write when something actually matters. No filler. No SEO sludge.
Eagle 3.1 and the Boring Economics of Sovereign AI
A technical look at vLLM’s Eagle 3.1 speculative decoding and the hardware economics behind sovereign AI serving.
May 27, 2026
Read More
Cloudflare Agent Memory is About State, Not Memory
Cloudflare’s agent memory changes how LLM systems manage state. This article examines memory drift, debugging, and operational risks.
April 28, 2026
Read More
From manual execution to controlled automation — without losing grip
A practical progression from manual execution to assisted workflows and controlled automation, with explicit inputs, outputs, traces and human review.
April 14, 2026
Read More
MiniMax M2.7: “Open Source” (With Strings Attached)
MiniMax M2.7’s open weights come with licensing and deployment tradeoffs. A look at what teams should check before adopting the model.
April 13, 2026
Read More
The Problem With Chasing Leaderboard Glory
Why benchmark leadership can mislead production model choice, and how quality, latency, reliability and inference cost shape deployment decisions.
April 10, 2026
Read More