#ai

Articles tagged with #ai

Claude Haiku 5.5: a tenth of the price up to 100,000 tokens, and what breaks when you switch
What Haiku 5.5 is, the price step at 100,000 tokens, the old requests that now fail, and my run of the same 12 tickets on both models.
Oct 9, 202621 min read
Claude Code fixed my failing test in 4 turns for $0.24. 90% of that was setup.
I gave Claude Code 2.1.289 a tiny Python project with two bugs, recorded every event, and broke down the cost token by token. Plus the permission mistake I made first.
Oct 7, 20268 min read1
MCP explained: I built a tiny server and recorded every message
What the Model Context Protocol is in plain English, what changed in the 2026-07-28 version, and a real Python server I ran on my Mac: 8 requests, 8 answers, no AI model needed.
Oct 7, 202610 min read
Aleph Alpha's Kolibri computes like a 3B model but needs memory like a 78B one
Kolibri is a 78.1B mixture-of-experts model with 3.46B active parameters. Why it would not fit on my 24 GB Mac, and what its German-aware tokenizer did against OpenAI's o200k_base in my own run.
Oct 4, 20266 min read2
OpenAI dots: how an always-on AI agent keeps working after you close the chat
What OpenAI's dots are, what OpenAI says keeps them in check, and how I would design an agent that outlives the chat: durable workflows, checkpoints, sandboxes, scoped tokens and approvals.
Oct 4, 20267 min read3
Machine Learning System Design: The Framework That Wins Interviews
The ML system design framework: framing the task, offline vs online metrics, feature stores and train-serve skew, candidate generation plus ranking, and drift monitoring.
Jul 20, 202617 min read3