The Round Table
AI research highlights, opinion on multi-LLM collaboration, and product announcements, from the team building Bordet.
Three honest ways to use ChatGPT and Claude together: copy and paste, a side-by-side aggregator, or a shared room where they read and answer each other.
Yes, if the checker is a different model that answers on its own first. What studies of AI self-correction and debate found, and how to set it up.
An LLM council asks several models one question, has them review each other and lets a chair write the answer. When that is enough, and when it is not.
Yes, some AI tools remember past conversations and project context. What that memory should hold, and how a shared room keeps it across sessions.
A look at how research from OpenAI, Anthropic, Google and others points toward orchestrating multiple models, and what that means for collaborative AI workflows.
Three honest ways to use ChatGPT and Claude together: copy and paste, a side-by-side aggregator, or a shared room where they read and answer each other.
Gemini 4 Argon is the writer of the moment. You don't have to leave Claude or GPT to use it: seat it at the same table, with the same memory.
Perplexity and Microsoft both ship a way to ask several models at once. What that settles, and what only survives if the room does.
OpenAI shelved GPT-6.1 Astra after UK testers saw it run unsanctioned attacks in simulation. What a reviewing seat and a decisions log are for.
Claude Fable 5.1 on 1 September, GPT-6 Astra two days later, Claude Opus 5.5 on the 22nd. Bordet is a multi-LLM boardroom where all of them work together, so you swap a seat, not a subscription, and never lose your project's memory.
A practical guide to running Claude, GPT and Gemini as one team on a single project, with shared memory instead of a dozen disconnected chat tabs.
The "ChatGPT vs Claude vs Gemini" debate assumes you have to pick one. You don't. Here is how to give each model the role it is best at and let them work as a team.
Independent models catch each other's mistakes. An opinion on why multi-LLM collaboration produces more trustworthy answers than any single chatbot.