How it works

Simple by design.

01

Ask anything.

Type naturally into one conversation. No need to think about which AI to open first.

What are the risks of launching a marketplace in a new market?
02

Mention an AI.

Use @claude, @gpt, @gemini, or any supported model. They pick it up instantly.

@claude What do you think?
Claude
The biggest risk is usually supply-side quality…
03

Let them collaborate.

Ask models to critique, compare, research, or build on each other's answers — all in the same thread.

@gpt Challenge that.
ChatGPT
I'd push back on that — the demand side is the harder problem here…

The core interaction

Talk to the AI you want.

Type @, pick a model, and it joins the conversation. It's that simple.

New conversation
@claude@gpt@gemini@grok@deepseek@qwen@r1@llama@glm@mistral@gemma@kimi

Type @ to mention a model — try it above

Multi-model

Why choose one?

Ask the same question to multiple models at once. Get different perspectives, then ask one to synthesize the rest.

✓Mention multiple models in a single message
✓Each responds with their own perspective
✓Ask one to compare or synthesize
✓The conversation stays in one place
Ask all of them. Then ask one to compare.
@claude @gpt @gemini
What are the biggest risks here?
Claude

I'd focus on regulatory risk first. In emerging markets, policy shifts can invalidate your core model overnight. Build for adaptability from day one.

ChatGPT

Distribution is the hardest part. Most startups underestimate channel costs. I'd stress-test your CAC before scaling any supply-side investment.

Gemini

The macro timing is interesting — infrastructure investment in the region is at a decade high. That creates tail winds, but also inflates competition expectations.

You
@claude Propose a solution to the cold-start problem.
Claude
Start with a curated seed set of high-quality content. Use collaborative filtering with implicit signals — dwell time, scroll depth — before explicit ratings accumulate. The key is bootstrapping trust without asking users to do extra work.
You
@gpt Challenge that approach.
ChatGPT
The seed content approach risks baking in the curator's biases early. I'd argue for a hybrid: collect enough behavioral signal in the first session to personalize, rather than starting from editorial assumptions.
You
@claude Respond to GPT's point.
Claude
That's fair. The bias risk is real. A middle path: use the seed set as a safety floor, but let behavioral signal override it quickly. Decay the editorial weight over the first 10 interactions.

AI collaboration

Models see the
whole conversation.

When you bring a new model into the thread, it reads everything that came before. It can respond to other models, build on their ideas, or push back.

Critique
Ask GPT to challenge Claude's answer
Build on
Have models extend each other's work
Synthesize
Ask one to summarize the rest
Debate
Let models argue a position

Your next conversation
can include every AI.

Free to start. No credit card required.