15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-631c4730b83c8223.js"],"Posts"] 16:T541,I’ve been exploring how recommendation systems are actually implemented in production, beyond just training models.

A common pattern I kept seeing is to split the problem into two stages:

1. Retrieve a small set of relevant candidates

2. Re-rank them using a model

Instead of doing brute-force inference across all items, I built a small prototype around this idea.

The flow looks like this:

- Store embeddings in a vector database (ChromaDB)

- Retrieve the Top-K most similar items/users based on vector similarity

- Run a TensorFlow.js model to re-rank the candidates

The goal is to reduce the search space before applying inference, which seems necessary when latency and scale matter.

What I found interesting is that once you move to this approach, a lot of the complexity shifts from the model itself to the retrieval layer:

- choosing K

- filtering candidates

- embedding quality

- latency vs recall trade-offs

Curious how others approach this in real systems:

- How do you decide on K?

- Do you rely purely on vector similarity or add heuristics?

- How do you handle re-ranking at scale?

Project: https://github.com/ftonato/recommendation-system-chromadb-tf...17:T4ea,I built an MCP server that gives AI agents access to compliance documentation — starting with the Colorado AI Act (SB 24-205), effective June 30, 2026.

The problem: organizations deploying AI in hiring, lending, insurance, or healthcare decisions need specific documentation — risk management policies, impact assessments, consumer notifications, bias testing docs, and appeal mechanisms. Most teams either pay $50K+ for a GRC platform, hire a law firm at $500/hr, or wing it.

What I built: complia

Write Modern Go Code with Junie and Claude Code

Paramount strikes deal to buy Warner Bros Discovery for $110B

Human brain cells on a chip learned to play Doom in a week

Rank5.io – a party game about the people you play with

Show HN: A production-style recommender using vector retrieval and re-ranking

Wall Street hit by UK mortgage lender collapse, raising fears

New Credit Blowup in London Has Wall Street Chasing Billions

Show HN: MCP server for AI compliance documentation (Colorado AI Act)

The fix to the iPhone Antennagate was 20 bytes (2025)

The Tools of Angel Investing

Murder is coming to AI, but not to Claude

Extracting 100K concepts from an 8B LLM

Agents with "Hands"

When Video Games Were Brown [Video]

Software as a Theater

Southern California air board rejected pollution rule after flood of AI comments

Barbara? Bush Crowley?

Amazon Wishlist Doxing Threat

Show HN: Host a Real time collaborative spreadsheet right from your pocket

Markdown Viewer – Extension for Firefox

Department of War Designates Anthropic Supply Chain Risk

Show HN: Prompt template for solo and small law firms

The First ZK Exploits Happened, and They Weren't What We Expected

President Trump orders federal agencies to stop using Anthropic

Cepstrum, Quefrency, and Pitch

The Outline Foundation: An Independent Home for Outline

Greek Politician Is Prosecuted for Admitting Ecstasy Use 36 Years Ago

Show HN: Forge-GPU – 55 C lessons for SDL's GPU API, built with Claude Code

MaxClaw – OpenClaw Powered by MiniMax

This week, the stock market lost $200B over a short story