ChatGPT is the best-known AI assistant in the world. Bordet puts a team of models — GPT included — in one room with roles, shared memory, and a record of what was decided.
ChatGPT is the default AI assistant for hundreds of millions of people, and OpenAI keeps widening it: Projects, custom GPTs, agent features, and free group chats that bring up to 20 people into one conversation with ChatGPT. If you need one excellent assistant, it is the obvious choice.
Two things are structural, though. Every voice in a ChatGPT conversation is an OpenAI model — you get one vendor's perspective, however many people are in the room. And in group chats, ChatGPT's memory is switched off by design: it neither uses nor forms memories there, so the conversation your team has today leaves nothing behind for the next one. Nothing records what was actually decided.
Bordet is built around exactly those two gaps. GPT sits at the table next to Claude, Gemini, and Grok, each with a role — strategist, maker, critic — cross-examining each other's answers instead of agreeing with themselves. The room accretes a shared memory that every member, human and AI, carries into the next session, and an append-only decisions log records what the team settled and why.
| Capability | Bordet | ChatGPT |
|---|---|---|
| Group chat with people | Yes, real-time rooms with invite links | Yes, up to 20 people, free |
| Models from multiple labs in one thread | Yes: GPT, Claude, Gemini, Grok side by side | No, OpenAI models only |
| Distinct roles working as a team | Yes (Strategist, Maker, Critic, custom) | Custom GPTs are personas of one vendor's models |
| Memory in group conversations | Yes, the room's shared memory grows every session | Off in group chats by design |
| Auditable decisions log | Yes, append-only, with provenance | No |
| Cross-session project state | Yes, curated state shared by every member | Projects hold chats and files, per-account memory |
| Second opinions across vendors | Built-in: ask the room, compare stances | Manual copy-paste into other apps |
| Agentic changes to your repo or files | Yes, in the room, review-gated via PRs | Yes, via separate tools (Agent, Codex) |
| Bring your own API keys | Yes (or managed billing), per-room spend caps | No, subscription only |
Comparison reflects our understanding as of July 2026; check ChatGPT's site for their current features.
You pay for what group chats deliberately leave out. ChatGPT group chats run one vendor's models and switch memory off, so nothing carries over and nothing is recorded. Bordet's room is the opposite: several frontier models with roles, a shared memory that compounds across sessions, and an append-only decisions log. And with BYOK, running the room costs roughly what the tokens cost.
Yes. Bordet is model-agnostic, so GPT sits at the table next to Claude, Gemini, and Grok. Many rooms make a GPT model the Maker or Strategist and let another lab's model play Critic, disagreements between vendors are exactly where the value is.
No. Keep ChatGPT as your personal assistant, it is excellent at that. Bordet is for project work: a standing room where a team of models and people collaborate over weeks, with memory and decisions that persist. The two coexist naturally.
As of its group-chat launch, OpenAI states that ChatGPT does not use your saved memories in group chats and does not create new ones from them. That is a sensible privacy default for a consumer product, but it means a team's working context evaporates. Bordet's shared room memory is scoped to the project and shared by design.
Open a boardroom, give each model a role, and watch the team remember what you decided.