By: Real Good Staff

Extra! Extra! Read all about it! Aiming to illuminate AI’s black box, the new nonprofit on the block, Real Good AI, is helping partners work toward a brighter future. As artificial intelligence changes lives, the folks at Real Good AI are on a mission to make sure those changes are, well, real good.

Real Good AI provides support to organizations working for the common good, drawing on expertise in research, data science, and the nonprofit sector. This 501(c)(3) charity offers a valuable, free service for community organizers, helping them harness the power of AI for positive impact. In the spirit of transparency and collaboration, Real Good AI is proud to launch The Real Good Newsletter, your source for updates, stories, and insights from their work with nonprofit partners. Stay tuned to keep up with our progress, meet our partners, and discover how you can get involved.

Subscribe today and learn more about Real Good AI at https://www.realgoodai.org/newsletter

By: Real Good Staff

Wondering how AI can help your nonprofit make a bigger difference? The Real Good AI team is here to lend a hand! Our crew of scientists and outreach experts use data science research to answer your questions and help you reach your goals.

Maybe you want to know more about your supporters, figure out the best time for a big fundraiser, or show the impact of your work for a grant. Maybe you’re just curious about how to connect with your community in new ways.

Whatever your question, we’re ready to dig into the numbers and find answers that make sense for you.

And don’t worry your privacy is important to us. We only share the information you want to share. Best of all, our help doesn’t cost a thing! Thanks to a generous endowment, Real Good AI offers these services free to nonprofits working for the common good.

If you’re looking for a partner to help you plan, measure, and grow your mission, the Real Good AI team is just a click away.

Real Good AI is spinning up some significant partnerships! Libraries looking for educational content, international nonprofits looking for fundraising and volunteer assistance, advocacy nonprofits looking to measure impact, or research institutions looking for guidance, we are building relationships through our data science expertise. We are excited to share these details and organizations with you, because we want you to have insights into our collaborations and see the impacts of your support.

By: Real Good Staff

Who’s behind the curtain at Real Good AI? Meet the awesome staff working hard to make sure AI is a force for good!

Dr. Amanda Muyskens: Executive Director & Data Scientist

With a PhD in Statistics, Amanda leads Real Good AI with a sharp mind and a big heart. Whether wrangling data (and staff) or charting the nonprofit’s course, Amanda is dedicated to making AI accessible and effective for all.

Craig Kaufman: Outreach Director

Craig is one of the friendly faces connecting Real Good AI to partners and the public. With a knack for building bridges and a passion for nonprofits, Craig ensures our work reaches those who need it most.

Dr. Imène Goumiri: Outreach Data Scientist

Armed with a PhD in Mechanical and Aerospace Engineering, Imène brings technical expertise and a collaborative spirit to every project. Dr Imène works closely with Dr Amanda to turn complex data into real-world solutions.

AJ Sandhu: Science & Social Media Communicator

The voice behind the newsletter, AJ translates science into stories and keeps our supporters in the loop. Whether online or in print, AJ makes sure everyone can follow along as Real Good AI makes a difference.

You can find more information about the Real Good AI Team here: https://www.realgoodai.org/team

Keep an eye on this space for an introduction to our accomplished board members.

Together, the team is committed to transparency, innovation, and most importantly doing real good. Stay tuned for more stories from the folks working to shape a brighter future with Real Good AI!

A special shout out to our Real Good Volunteers helping us make a brighter future!

Dr. Eric Bell volunteered to develop AI educational tools in fun and accessible mediums.

Mia Kaufman helped create adorable emotes for the Twitch channel and graphics for business cards.

And the talented RyansArt has been working closely with staff on an educational project. Look forward to an exciting preview in the next newsletter!

By: Real Good Staff

Stop the presses! Real Good AI is now a proud Twitch Affiliate, and you’re invited to join the fun. As Twitch affiliates we can now monetize our weekly livestreams, this means ad revenue, donations of “bits,” and $6 monthly subscriptions, all supporting our mission. In fact, we already have 11 subscribers! Looking for something to do on Thursdays at 2 PM EST? Tune in to the Real Good AI Twitch Channel, where our team takes a well-earned break to play games, share updates on our mission, and spotlight the incredible work of our nonprofit partners.

Let's Talk Numbers:

This is an incredible level of reach and we are excited to share our experience in nonprofit streaming with other organizations.

April 17th featured our very first live on stream interview with Dr. Eric Bell while playing World of Warcraft. Viewers were entertained while they learned about his career path to becoming a postdoctoral researcher at Vanderbilt University. Dr Bell spoke to us about AI’s use in drug discovery and even what a supercomputer is! If you would like to watch that steam, you can access it here: https://www.twitch.tv/videos/2442501107

Each week, it’s two hours of cooperative gaming, lively discussion, and special guests from the worlds of nonprofits, science, and streaming.

Thanks to our streams, we’ve connected with amazing viewers, potential nonprofit partners, and volunteers from all walks of life. From libraries to educational nonprofits, we’re building bridges and making new friends.

Is there anything that you would like to hear us talk about? From science to streaming to nonprofit work let us know and we will work to find some awesome guests to share their expertise.

Email us at contact@realgoodai.org with your suggestions!

So grab your favorite snack, fire up Twitch.tv, and join the team as we game for good, celebrate community, and keep you up to date on all things Real Good AI.

Catch us live most Thursdays at 2 PM EST at https://www.twitch.tv/realgoodai

See you in the chat!

By: Dr. Amanda Muyskens

As a fun application of AI, we taught an AI to play tic-tac-toe with reinforcement learning (RL) in the programming language R as a demonstration of how the tech can be used for so many things other than just chatbots. This was a learning opportunity to see how viable “vibe coding” is in practice.

First: what is “vibe coding?”

Don’t feel behind if you haven’t heard of it, it is very new. Coined in early 2025 by Andrej Karpathy, a co-founder of OpenAI, “vibe coding” is “a programming technique dependent on artificial intelligence (AI), where a person describes a problem in a few sentences as a prompt to a large language model (LLM) tuned for coding.”Essentially, one instructs the AI to build a coding tool as a manager would instruct an entry level coder and uses the results as part of their projects. For now, we ignore the ethical concerns of such a task and set out to test the viability: is it possible this could help or replace coders?

We thought this would be a simple task. There are plenty of online tutorials for solving tic-tac-toe using the programming language Python. So in theory this should be a relatively easy recall and translation task with the addition of a convenient user interface (UI) for the human to interact with.

Well, the review is in and the results are mixed.

There is a lot to consider, but let’s focus on results first. Did we need to know about mathematical methodology to vibe code? Unfortunately, in our estimation the answer is an astounding YES. We absolutely could not have gotten to our solution without our debugging and domain knowledge.

Did vibe coding help us get to our solution faster? Well yes but mostly no.

Although it did get us a good UI quickly, debugging and looking for errors in a code you didn’t write can take so much longer than just re-writing a simple code. We think it took us about the same amount of time, but our time was spent significantly differently than if we hadn’t vibe coded.

Here are some more specifics:

First the positives:

  • It did seem to understand from the prompt what it was supposed to be building.
  • It actually structured the code correctly where the reinforcement learning (RL) was trained first, saved and then the interface was a secondary file.
  • The RL algorithm it selected was the reasonable choice.
  • It saved us from having to know syntax for UI features we had never used before.
  • It suggested fun features that improved the quality of the resulting UI.

The negatives:

  • There were a lot of errors and debugging to get the code to even run, and debugging on a new code is always harder than debugging code you wrote yourself.
  • Syntax errors–it was sure it could do things it could not and missed subtleties of the R and R shiny languages.
  • Methodological errors. In particular, the algorithm was only rewarding an end to the game, rather than rewarding the algorithm for winning and punishing it for losing. This led to a hilarious version of the game where the AI would just let you win. Although amusing, this is dangerous for anyone trying to code something they don’t know already.
  • In fine-tuning the UI, it did not seem to understand basic instructions on moving pieces around. Therefore, I wasted a lot of time arguing with it about this rather than fixing the one short line of code to do what I ultimately wanted (which I had to do anyway!).

Overall, “vibe coding” can provide some positives, particularly in helping to make a nice UI. However, I wouldn’t count on vibe coding to take coders jobs anytime soon. Users still need to have the domain knowledge and check everything it spits out.

For your enjoyment, you can play our AI tic-tac-toe here: https://realgoodai.shinyapps.io/deploy/

We’d love to hear how you did against it!