March 21 - Smarter, Faster, Stronger Agents
Agents continue to get smarter, performance gets a bump, and free tier users have more time to see what makes Mito great.
Last updated
Was this helpful?
Agents continue to get smarter, performance gets a bump, and free tier users have more time to see what makes Mito great.
Smarter Editing: The agent can now edit existing notebook cells, not just create new ones.
On-the-Fly Problem Solving: Rather than planning everything upfront, the agent now figures things out as it goes—making it more adaptive and efficient.
Full Notebook Awareness: The agent receives the entire notebook as context, improving accuracy and coherence.
Citations: The agent can now generate citations when appropriate.
You can now open previous chat histories to review or continue past conversations.
Model upgrade: We've moved from o3-mini to GPT-4o — about 10x faster on agent prompts. There's a small quality tradeoff, but the speed boost is worth it.
Websocket Resilience: If you're idle for a while, websocket connections will now automatically reconnect.
Behind the scenes server tuning to ensure less timeouts for server key users.
Open Source users now have 50 completions and 5000 autocompletes per month.
These limits reset at the start of each month.
Markdown backticks (```) now render in purple font, as expected.
The variable dropdown scrolls properly again—no more cut-off options.
Last updated
Was this helpful?
Was this helpful?