Tools

News

Notícias

Classificados

Cursos

Broker

IPv4:
IPv6:
UpOrDown
Ping
MTR
MTU Detect
Portscan
DNS
HTTP/SSL
My IP
IP Calc & Sum

ChatGPT Branching Enables Parallel Conversations

Image © Arstechnica
OpenAI has launched a new ChatGPT feature that lets users split a single chat into parallel threads, keeping the original conversation intact while exploring different paths.

ChatGPT now supports conversation branching, allowing users to run parallel threads from a single chat without discarding the original discussion.

Users can branch by hovering over a message, selecting ‘More actions’, and choosing ‘Branch in new chat’. A fresh thread begins that includes all messages up to that point, while the original chat remains intact.

This enables teams to explore different approaches from the same starting point—for example, testing formal versus humorous tones or alternative strategies—without overwriting the main conversation.

The update addresses a long-standing limitation of linear dialogues in large language models, where users had to alter prompts or start anew to test different directions. A 2024 study by researchers from Tsinghua University and Beijing Institute of Technology argued such linear interfaces can increase cognitive load during brainstorming and complex analysis.

Industry observers have welcomed the feature, with some comparing it to Git branches: both let you experiment with alternatives while preserving the original work. The development highlights how AI chatbots are tools that can be steered and revised, rather than fixed personas.

 

Arstechnica

Related News

Wasco Gets $10M Fiber Upgrade
Pixnapping Threat Expands to Android 2FA
Bluepeak Upgrades 25k Customers in SD MN
Signal's Post-Quantum Upgrade: Engineering Triumph
Comcast Begins Teays Valley Internet Buildout
Metronet expands in 4 Michigan towns

ISP.Tools survives thanks to ads.

Consider disabling your ad blocker.
We promise not to be intrusive.

Cookie Consent

We use cookies to improve your experience on our site.

By using our site you consent to cookies. Learn more