Practical DevOps, Cloud, AI &amp; Linux engineering guides

A 1T-parameter MoE model with open weights coming doesn't make self-hosting the right call for most teams.

••2 days ago

Mistral Large 4 'Le Chonk': Self-Host or Use the API?

A 1T-parameter MoE model with open weights coming doesn't make self-hosting the right call for most teams.

A compromised npm release went after cloud credentials as expected, and local AI coding-assistant config files, which most teams never treated as secrets.

••3 days ago

The Tensorlake npm Compromise Targeted Your AI Tool Configs

A compromised npm release went after cloud credentials as expected, and local AI coding-assistant config files, which most teams never treated as secrets.

Haiku 5.5 is cheap enough to stop rationing subagent calls, but the 75% savings claim only holds if your prompts are short.

••4 days ago

Claude Haiku 5.5: When to Route Down From the Frontier Tier

Haiku 5.5 is cheap enough to stop rationing subagent calls, but the 75% savings claim only holds if your prompts are short.

Nano Banana 2.1 is a routing update, not a new model to chase, and it mostly matters for batch and multi-reference image work.

••5 days ago

Gemini Nano Banana 2.1: What Actually Changed, and for Whom

Nano Banana 2.1 is a routing update, not a new model to chase, and it mostly matters for batch and multi-reference image work.

A single Availability Zone in AWS's Spain region degraded for roughly 2.5 hours, and most "multi-AZ" architectures would have failed anyway.

••6 days ago

AWS EU-SOUTH-2 Packet Loss: The Single-AZ Lesson Nobody Learned

A single Availability Zone in AWS's Spain region degraded for roughly 2.5 hours, and most "multi-AZ" architectures would have failed anyway.

GitLab is capping unauthenticated API calls at 60 an hour starting October 19, and the preview windows land before most teams will have noticed.

Kiril Urbonas·2 min read·51

GitLab's New Rate Limits: What to Fix Before Oct 19

GitLab is capping unauthenticated API calls at 60 an hour starting October 19, and the preview windows land before most teams will have noticed.

Most moderation, routing, and triage calls aren't generation problems. Clef proves it by answering in 38.8 milliseconds instead of seconds.

Kiril Urbonas·2 min read·19

Cloudflare's Clef: When a Decision Model Beats an LLM Call

Most moderation, routing, and triage calls aren't generation problems. Clef proves it by answering in 38.8 milliseconds instead of seconds.

Mods make Claude Code a real runtime you can program. They also run with your full native permissions, no sandbox, so a bad one is a bad CI step.

Kiril Urbonas·2 min read·11

Claude Code Mods: Programmable, Unsandboxed, and Your Problem in CI

Mods make Claude Code a real runtime you can program. They also run with your full native permissions, no sandbox, so a bad one is a bad CI step.

Google shipped a 1M-token output ceiling and strong cyber-defense benchmarks, then handed the keys to almost nobody. Here is what that gating pattern tells you.

Kiril Urbonas·2 min read·8

Gemini 4 Argon: A Frontier Model You Mostly Can't Use Yet

Google shipped a 1M-token output ceiling and strong cyber-defense benchmarks, then handed the keys to almost nobody. Here is what that gating pattern tells you.

OpenAI priced GPT-6.1 Sol at a fifth of GPT-6 Astra and called it nearly as good at agentic coding. That claim is now a routing decision, not a headline.

Kiril Urbonas·2 min read·16

GPT-6.1 Sol's 80% Price Cut Rewrites Agentic Coding Math

OpenAI priced GPT-6.1 Sol at a fifth of GPT-6 Astra and called it nearly as good at agentic coding. That claim is now a routing decision, not a headline.

A slow database made a backend service look dead, and the health check that was supposed to protect it killed it faster. The lesson has nothing to do with AI.

Kiril Urbonas·2 min read·19

Azure OpenAI's Sweden Central Outage: A Health Check Postmortem

A slow database made a backend service look dead, and the health check that was supposed to protect it killed it faster. The lesson has nothing to do with AI.