
Motivated, accessible math. YouTuber | Open-source developer | Aspiring independent researcher. Prev: @Wikimedia #Eventmath grantee. Views are my own.
Denver, CO
Joined June 2016
- Tweets1.6K
- Following634
- Followers872
- Likes1.5K
am thinking about what the future of a "math book" should be... heavy on graphics; with interactive elements (e.g. GPT Sites); dynamic references that auto-update N-times-per-year to reflect changes; self-repairing to fix errors...
not paper & not a pdf... maybe a github repo?
Greg Stanton retweeted
The Institute for Computer-Aided Reasoning in Mathematics (ICARM), a new @NSF Mathematical Sciences Research Institute at Carnegie Mellon University @CarnegieMellon, is hosting a Summer School on Formalization in Lean @leanprover from July 7-18, 2026.
Computational proof assistants now make it possible to develop global, digital mathematical libraries with theorems that are fully checked by computer. This summer school will introduce students to this technology and to the goals and benefits of formalized mathematics, including metaprogramming and autoformalization. By the end of the session, students will be in a position to formalize mathematics on their own, join the Lean community, and contribute to its mathematical library.
School Structure
The program interweaves short lectures with interactive tutorial sessions. The first week focuses on learning the Lean language and gaining comfort with the technology. During the second week, students will work on independent formalization projects in small groups with opportunities to share results and receive feedback. Optional tutorials on metaprogramming will also be available.
Suggested Prerequisites
- At least one undergraduate-level computer programming course
- A mathematical area of interest for the student's formalization project
For more information and to apply to participate, please visit:
icarm.io/project/summer_scho…
Greg Stanton retweeted
if you're taking/teaching applied dynamical systems this term & want to peek at a book-in-progress, i've updated the draft on my website... see below
Greg Stanton retweeted
I'm excited to share: we built a free AI coach for intro-level proof writing. It critiques drafts, points out gaps, and helps you or your students iterate without giving away the answer. Also for lapsed mathematicians to see if they've still got it. Link in the next tweet.
Greg Stanton retweeted
if you're looking for an iconic linear-algebra-with-applications-to-AI textbook...
Greg Stanton retweeted
🥤"Optimizing mixtures of metal–organic frameworks for robust and bespoke passive atmospheric water harvesting" by C. Harriman, Q. Ke, T. Vlugt, A. Howarth, C. Simon.
feedback welcome on our new preprint!
💡the key idea is to optimize the composition of *mixtures* of metal-organic frameworks (MOFs) to comprise the adsorbent bed of a passive atmospheric water harvester. with a mixture, each MOF can specialize in delivering water under different conditions.
we frame a linear program for the optimal design of the composition of the MOF adsorbent bed that is:
(1) bespoke, i.e., tailored to a declared geographic region and time frame and
(2) robust, i.e. satisfies minimum daily water delivery requirements despite variance in weather.
both weather data and MOF water adsorption data are inputs to our optimization problem.
Greg Stanton retweeted
New video!
2300 years ago, Aristarchus used very clever math to find relative sizes and distances of the sun, moon, and earth.
His math led him to heliocentrism, but the evidence raised an unanswerable question and his idea was discarded for centuries.
Greg Stanton retweeted
i am about to teach a new applied linear algebra course at @PennEngineers, meant for datasci/ML/AI.
to support the class, i've written a book...
[see below for links to text]
Greg Stanton retweeted
How learning happens in seven general principles ⬇️
1. Working Memory Capacity is Limited: Our working memory, responsible for holding and processing information in real-time, has a limited capacity. This limitation means that when learners are exposed to large amounts of new information simultaneously, they may struggle to process it effectively. Focusing attention is essential because distractions or overload can lead to cognitive strain, reducing the ability to learn. Teachers can support this by breaking information into manageable chunks, allowing students to engage deeply with smaller parts before moving on to the next concept.
2. We Learn Through Meaning and What We Already Know: Learning is enhanced when new information connects to existing knowledge. When learners relate new content to things they already understand, they create a stronger foundation for retention and comprehension. This process, known as schema-building, helps students make sense of complex ideas by fitting them into an established framework. Teachers can facilitate this by introducing new concepts through familiar examples or linking them to relevant prior knowledge, making learning more meaningful and long-lasting.
3. Novices Think Differently Than Experts: Novices often approach tasks with means-end analysis or trial-and-error methods or simple strategies because they lack the knowledge to recognize patterns or apply sophisticated approaches. In contrast, experts draw on rich mental models, using prior experiences and refined techniques to solve problems more efficiently. This difference means that novices need structured guidance to develop these mental models over time. Teachers can help by gradually introducing students to expert strategies, enabling them to progress from basic skills to more advanced understanding.
4. To Remember, We Need to Forget: Storage strength is the measure of how well information is embedded in long-term memory, indicating its durability over time.
Retrieval strength is the ease with which information can be accessed from memory at a given moment, which fluctuates based on recent use and context. Unlike storage strength, retrieval strength is influenced by context and interference from other information. To remember something long-term, it’s essential to boost its retrieval strength, which can be achieved by actively retrieving the information. Unlike passive studying, retrieval requires you to recall the item on your own, strengthening memory more effectively, especially when the retrieval is challenging. This principle explains why strategies like interleaving (alternating topics) and spacing (spreading out sessions) are so effective. By allowing time for slight forgetting, these methods make retrieval harder, ultimately enhancing memory retention. This “adaptive forgetting” is crucial for retaining core concepts.
5. Learning ≠ Performance: A high test score or good performance in the classroom doesn’t necessarily mean that true learning has occurred. Learners might perform well temporarily by memorizing facts or rehearsing specific methods without understanding the underlying concepts. This principle highlights the difference between short-term performance and deep, long-term learning. Effective assessment should look beyond scores to evaluate true comprehension. The rate of assessment matters which is why regular checking for understanding is so important. Teachers can use open-ended questions, application tasks, and reflective discussions to gauge whether students genuinely understand the material.
6. Achievement Leads to Motivation: Success in learning (however small) can foster a positive cycle of confidence, self-efficacy, and motivation. When students achieve their goals, even in small steps, they are more likely to believe in their ability to succeed in future learning tasks. This sense of accomplishment encourages them to take on new challenges and persist through difficulties. Teachers can support this cycle by setting attainable goals, providing regular feedback, and celebrating incremental progress to reinforce students’ confidence and motivation.
7. Learners Are Bad at Knowing How to Learn: Many students are unaware of effective learning strategies, often relying on passive techniques like re-reading, highlighting, or cramming. These methods feel productive in the short term but are less effective for long-term retention. Evidence-based strategies like retrieval practice, spaced learning, and interleaving have been shown to improve memory and understanding, but students may need guidance to adopt these methods. Teachers can enhance learning outcomes by teaching students about these techniques and encouraging their use, helping students develop more efficient and powerful study habits.
(Thanks to @MrAlexKoks for the design)
Greg Stanton retweeted
Tyring the recent Complex Mode in @Desmos
reddit.com/r/desmos/comments…
Greg Stanton retweeted
can AI do research-level mathematics? make conjectures? prove theorems?
there’s a moving frontier between what can and cannot be done with LLMs.
that boundary just shifted a little. this is my experience with AI proving a new theorem.
1/
Greg Stanton retweeted
I've been copying and pasting some code for distance-based sampling of Bezier paths into sketches for a while. I mostly use it for even path sampling and morphing between paths. I finally put it into a library on npm if anyone's interested! github.com/davepagurek/bezie…
Greg Stanton retweeted
i've sketched up a custom GPT to teach you how to become a power-user of ChatGPT. after a few questions it puts you on a skill tree and then tutors you in specific prompting and usage techniques.
Greg Stanton retweeted
🌟Updates coming soon!🌟
complex-analysis.com/
Support this #opensource project by becoming a member in Patreon:
patreon.com/jcponce
Thanks for your support! ∞❤️
Greg Stanton retweeted
It's really cool to see the past few years of @p5xjs development laid out on a timeline like this! The community got a lot done.
Our 2023 Impact Report celebrates a decade of growth and achievements, spotlighting #software development, #community engagement, and team expansion 💜 Check our report here: processingfoundation.report
#2023ImpactReport #creativecode #Processing #p5js
Greg Stanton retweeted
Have you ever wondered about the magic behind ChatGPT? As more LLM-based generative AI tools become part of our lives, they don’t have to remain mysterious black boxes! Here, we delve into the basics of LLMs, from predicting the next word to techniques like fine-tuning and RLHF.
Greg Stanton retweeted
Check out this amazing "book". By Neil J. Calkin, Eunice Y.S. Chan, and Robert M. Corless (@corless_rob)
computational-discovery-on-j…
Greg Stanton retweeted
I'm teaching complex analysis right now. One of my students, Jackson Fellows, made this interactive tool to visualize complex functions w = f(z) as mappings from the z-plane to the w-plane. jacksonfellows.github.io/com…
It currently supports +, -, *, /, (), i, exp, sin, and cos.






























