Top AI Product

Spotify’s engineering team said the quiet part out loud: most of what a coding agent does isn’t thinking, it’s I/O. Reading 2,000-line files, stamping out boilerplate — none of that needs a frontier model. Portal is Spotify’s platform for declarative agents on ephemeral runtimes (“AWS Lambda, but for agents”), and its new shunt plugin intercepts the grunt work, hands it to a cheap model like Gemini 2.5 Flash, and lets Claude keep only the reasoning.

How the shunt works

Any file read over 350 lines gets blocked by a hook and rerouted to a bulk-reader mode; a code-writer mode handles boilerplate. On Spotify’s Java monorepo, bulk reads use about 90% fewer tokens. The tradeoff is 10–30 seconds of latency per delegation, and editing stays with Claude — cheap models fumble line numbers.

Plug it in today

The whole thing is an open-source Claude Code plugin, spotify/portal-ai-plugins: a CLI plus hooks and skill files, zero code changes. That’s the interesting part — routing at the harness level, not a model router. With teams burning $200–500 per developer monthly on tokens, the 239 points on HackerNews aren’t a mystery.

You Might Also Like

• Pi Mono 29k Stars and a 200 Token System Prompt That Rivals Claude Code
• Zcode Zhipu z ai glm 5 2 Coding Agent hit hns 1 Spot Chinas Open Answer to Claude Code
• Claude Code Security Just Dropped and it Already Found 500 Zero Days Nobody Knew About
• Claude Code Remote Control Just Turned my Phone Into a Coding Terminal and im Weirdly Into it
• 27k Github Stars in Weeks Learn Claude Code by Shareai lab Breaks Down ai Coding Agents Into 12 Lessons