New blog post: Ralph needs a test train split


For those who don't know, Ralph Wiggum is the name given to the dumb, tail-chasing AI coding loop that often churns and churns, getting itself into trouble generating endless nonsense. With all sorts of mitigations to avoid Ralph ralphing up meaningless, crazy code

Another way to think about Ralph is the endless epochs of model training. Sometimes after enough epochs, training goes off the rails, gets overfit, and stops actually improving on the task.

That's why, coming from a search domain, I think about controlling Ralph's behavior in terms of overfitting. After all, a lot of my AI coding is for things like search rerankers and classifiers - not ust your everyday CRUD app.

Here overfitting is the big risk. And manipulating AI coding agents to have an evaluation system the agent can't peak matters a great deal.

New blog post: ​https://softwaredoug.com/blog/2026/01/17/ai-coding-needs-test-train-splits

Enjoy,

-Doug

Events · Consulting · Training (use code search-tips)

You're subscribed to Doug Turnbull's daily search tips where I share tips, blog articles, events, and more. You can always manage your profile:

Doug Turnbull

I share search tips, blog articles, and free events I'm hosting about the search+retreval industry, vector databases, information retrieval and more.

Read more from Doug Turnbull

Have you been to a conversion-crazy site? It’s nuts. Their site screams at you. They probably have the modern version of the HTML blink tag. Popups everywhere just won't go away. Buy buy buy! It’s fun to go to a physical store when you can browse the shelves, talk to customer service, and get help. People avoid stores lacking information and only high pressure salespeople in your face. If your search stinks of pressure, users will retreat. They’ll stay on Google. They win precisely because...

My buddy John Berryman and I were nicely hosted by Hugo Browne-Anderson on the Vanishing Gradients podcast. We talked about how agentic search stands poised to be more disruptive to the Information Retrieval space than RAG. Check it out! Other upcoming events Tomorrow (free) Cheat at Search Essentials, Vector Search (free) ReasoningLayer.ai - symbolic reasoning with LLMs Tuesday (free) Cheat at Search Essentials - Search Evaluation (WTF is an NDCG?) Friday (free) Doug Turnbull + Daniel...

Youtube masterminded how to turn engagement into insights. Whether search or their feed, you can learn from how they learn from you! They: Create low-friction, sticky, addictive little interactions you do subconsciously on the surface. How often have you found yourself hovering over a video on your feed? Give you many actions to take on a video, even from search results themselves (bookmark, share, etc) Treat their monetization (ads) more as a guardrail. They know engaging you with a sticky,...