One of the most common problems with AI assistants: everyone wants something different. One person just wants the answer, no preamble. Another needs context explained. A third wants to see the technical details. Most AI products solve this by not solving it at all — one size fits all.
Alex on CoreSynth now has a communication-style toggle. Four detail levels the user sets once and keeps across all conversations and devices.
Four styles
CONCISE → Just the result and action. Minimal text.
BALANCED → Result with a brief reason. Default.
DETAILED → Explains steps and context in plain language.
EXPERT → Dense and technical. Assumes prior knowledge.The user opens settings, clicks a row, done. The style saves to localStorage for instant effect and to the database simultaneously — so when they switch devices, it follows them.
How it works under the hood
The style is injected into the system prompt via placeholders. The prompt builder takes the saved value and adjusts instructions accordingly — from „respond briefly” to „explain technical context”. The pipeline is identical to reasoning effort: localStorage → API request → validation → service → prompt builder.
The goal wasn’t to create „four different AIs”. The goal was to give the user a dial they set once and forget — and yet it fits them perfectly.
Why it matters
Communication style isn’t just cosmetics. It affects:
- How many tokens the AI uses per response
- How quickly the user finds what they need
- How the AI feels to a new vs. experienced user
Concise for someone who just needs to know if a part is in stock. Expert for someone who wants to know why and how it works under the hood. Same agent, same data, different level of detail.
Tech details
STORAGE → localStorage (instant) + DB (persistence)
SYNC → On chat load, DB value writes to localStorage
when localStorage has none (new device)
VALIDATION → Zod schema, whitelist of 4 values
PROMPT → {{personality}} placeholder in system prompt
UI → Modal with selection, active row highlighted with ring
TESTS → Unit tests for pipeline and UI logicWhy I’m writing about this
Most AI products give you a chat and call it done. But communication with AI isn’t just about answers — it’s about how those answers look. A style toggle is a small thing with a big impact on how an AI assistant feels. It’s a classic case where the technical implementation is simple, but the decision to „just do it” is what makes the difference.