MaplePrimes Announcement

For those who are familiar with the long history of the Maple Conference, you’ll know that we vary the format from time to time, and for the next conference, we'll be trying something new!

The next event will be a user conference, strongly focused on enabling Maple users to get the most out of Maple, increasing their knowledge and abilities through a series of technical talks by Maple experts. We’ll release more details later, but our main goal is to engage with and support the many customers who love Maple and want to learn how to do even more with it. This event will be held in the spring of 2027, which we hope will be a convenient time for those who find it difficult to attend a fall event.

Another aspect of past conferences has been the opportunity for users to share their work with the Maple community. We'd like to remind you that the Maple Transactions, an open-access online journal, is an ideal place to publish your work, and you can see an overview of the journal in this short video. If you prefer a less formal venue, we encourage you to post right here on MaplePrimes to share your work and interact with the community.

We will share more details as we get closer to the event. For now, we just wanted to give you an idea of the plans, as we know that many people look forward to the Maple Conference and expect to hear from us around this time.

Featured Post

I am delighted to share that the ExaktAI Workspace now also provides a full-featured mathematical document interface for Lean, the theorem prover and proof assistant, integrated directly with computer algebra system (CAS) computation.

The Workspace, including its Lean integration, is in free beta testing and open to everybody.

Lean files (.lean) open as Lean documents, proofs are checked step by step with goals shown under each step, CAS computations can be inserted inside proofs, and a class of CAS results can be turned into Lean theorem candidates that Lean then independently checks.



The Workspace also includes a Lean tutorial, intended to make this accessible to users who are not yet familiar with Lean syntax and workflow.

CAS computations can be inserted between Lean proof steps. A class of CAS results can also be turned into Lean theorem candidates that Lean independently checks.



Edgardo S. Cheb-Terrab
ExaktAI, Canada
Research Fellow Emeritus at Maplesoft.

Featured Post

Mathematics and physics are evolving disciplines. AI is increasingly becoming an active participant in that evolution, producing results at a speed and scale that are difficult to follow without appropriate computational tools.

In the ExaktAI project we are addressing two related issues: the reliability problem of AI mathematics through computational validation, and maximizing human participation in checking, understanding and extending the results.

For the latter, we developed the ExaktAI Workspace.

The Workspace is a full computational-mathematics document environment in which one can follow, inspect, modify and extend AI-generated mathematics, or simply do the mathematics directly. Of particular relevance here, the Workspace can run Maple directly when Maple is installed.

The public beta is now open. I would be particularly interested in feedback and suggestions, both about the Maple integration and about the broader ideas for keeping us, the humans, actively involved and increasingly participating in AI-mathematics developments.


Update, September 27: the Workspace now also provides a document interface for Lean, the theorem prover, with CAS computations inside proofs; see "The Lean theorem prover" below.

Edgardo S. Cheb-Terrab
ExaktAI, Canada
Research Fellow Emeritus at Maplesoft.