Show HN: Tangents – Non-linear LLM chat with hands-on context control

[_boffin_]

LLM chat breaks when you're learning. Side questions either clutter the main chat, or you copy/paste into a new chat and lose the context.

Tangents makes the workflow explicit. You can branch off any message to chase a rabbit hole, then merge the findings back without polluting the main thread.

Key Features:

  - Branching: Select text in any message to fork a new chat.

  - The Collector: Highlight snippets across different branches to build a "shopping cart" of context.

  - Compose: Send a prompt using only what is in your Collector (explicit context control).

  - Console: Inspect exactly what context is being sent to the LLM before you hit enter.

How it's different:

It is not a node-graph canvas. It keeps the linear chat UI but allows inline branching. It is not an agent framework; it is a tool for humans who want manual control over the LLM's context window.

-----

View it here: https://tangents.chat/hn

Note: Currently supports OpenAI (BYOK supported).

[_boffin_]

Quick clarification: the `/hn` page is a no-login, no-API-key interactive demo (autoplay is just the default walkthrough; you can pause and click around).

OpenAI BYOK is only for the full app when you want real model calls.

More detail on what’s different under the hood:

  - Branches are anchored to a source message + selected span (not freeform threads).
  
  - Collector items are references back to those spans, so "Compose" can build a prompt from explicit citations rather than chat history drift.

  - "Context compiler" shows the exact prompt stack + token budget, and lets you exclude/pin items to control what survives truncation.

Feedback I'd love: does Branch + Collector + Compose feel faster than "open a second chat window + copy/paste", or does it feel like extra steps?