T2 Project Reflection and Next Steps Worksheet

[lilymaechling]

T2 Project Reflection and Next Steps Worksheet

Original idea

Revisit your original goals (check out your project proposal).

MVP Progress?

• We implemented a web IDE that can process, run, and provide output of written code
• We leveraged OpenAI technology to turn a text prompt into code
• We used Whisper and speech to text technology to get user voice input ready to turn into code
• We set up a beautiful UI for our site and styled our IDE
• We added file upload and download capabilities to allow for work saving and uploading

How is it different?

• Our project is still very limited in its capacity
• We expected our project to be more useful for real coders than it ended up being
• We are in the process of changing our focus
• At the end of last term, we realized that our voice to code technology was too broad for an advanced coder user base. An advanced coder would need to be able to produce more granular output, and need to code specific things that OpenAI covers in more broad strokes. For example, a real coder would need to write out specific lines with their voice, where OpenAI doesn't take code input and simply spit it back; they are very creative with their results.

Still happy with direction?

• NO. We had a 2 hour brainstorming session on changing our direction.
• Most of the feedback that we got at Technigala regarded our lack of a use case. While we initially thought this would be a voice to code tools for real developers, the way it ended up made it seem like the capabilities of OpenAI are way too broad and creative to be useful in a real world example as part of a larger development project.
• We then tried to pivot to an educational lens for this product, but we have struggled to determine a good way to bring education to voice to code as we have it now. Below, we have a list of ideas that we came up with during our brainstorming session.

What do you wish went differently?

• We wish that we had a better idea of direction when we set out on this project. Right now, we feel like we have a solution without a problem. While we had a problem in mind in the beginning of this project (HMW make coding more efficient and less isolating, and other questions along these lines), the technology that we have come up with doesn't really address this question.

Feedback

What feedback from last term do you think is relevant and interesting to incorporate. List your feedback broadly and label them with how they stack up for your vision. These may already be in ZenHub but after this discussion they should be.

• The feedback we are focusing on the most at this present moment is the direction of this project. This feedback was mostly given from Tim.

Questions

• What makes an educational experience in an IDE?
• Why is voice useful in education and how does it fit into an IDE/does it even have a place there?
• What widgets do you think are important and set us apart from other IDEs/make us useful for beginners?
• Why would a user join and stay on this site? -> This was a big question for us during our brainstorming session and it is what we are struggling with the most.
• As we have it now, why would an experienced coder use Convocode?

Answers

• What makes an educational experience in an IDE?
  • Auto-Documentation - to learn good coding practices and have the importance of documentation drilled in from the beginning
  • Runtime analysis - Learning efficiency early and how to analyze code
  • Language comparison - Learning other languages, as every good coder knows more than one
  • Error searching from StackOverflow - Ease of use and can stay in one place to debug
  • Live conversational help from chatbot - Can ask the chatbot to explain concepts, remind the user how to do certain things that they may have forgotten
• Why is voice useful in education and how does it fit into an IDE/does it even have a place there?
  • During our meeting, we started leaning towards getting rid of the voice feature entirely and having our chatbot be communicated with typing.
  • We think that voice makes things slow and sometimes more confusing, and if we are marketing this product towards beginner coders, we should eliminate all resistance.
• Why would a user join and stay on this site? -> This was a big question for us during our brainstorming session and it is what we are struggling with the most.
  • This is what we are really struggling with.
  • As a platform geared toward education, there is this idea of graduation. Once a user graduates beyond being a beginner coder, what is the point of our product? We discussed the fact that there is none, and games are meant to be beaten. However, we didn't like the idea of our product just sitting there abandoned after being "beaten". This is why we ended up abandoning ideas that included filling in code scaffolding, completing "MadLibs" of code, and playing games with our AI.
• As we have it now, why would an experienced coder use ConvoCode?
  • We determined that an experienced coder wouldn't use it. We discussed some of these reasons (the broadness of OpenAI) above.
  • Initially, we had this idea about flow, and how a coder can stay in one place while utilizing the many online resources that coders use during the development process and use their voice to code seamlessly and interactively. We wanted coding to be a smooth experience, but as we worked with the technology that was available to us, we realized that this wasn't the experience that we came up with.

Vision

Below is the brainstorm for our new vision/direction for ConvoCode Group Term 2 Brainstorm

Ideas for direction of project this term

Walkthrough coding problems

• Ask the computer a question and it will walk you through a solution
• Problem is this doesn't utilize voice to code that much
• Codex will not always give the same answer for coding questions

Web IDE

• "Clippy" character helps you code
• IDE with more educational features and emphasis
• Why do you need the voice for this?
• Be prompted to solve challenges by Clippy
• Use Codex to provide examples for for loops and other things for students learning
• Have accounts and profiles that track usage
• Adding features to elimante key coding things and fill them in (Mad libs)
• Things that this IDE would educate a beginner code on
  • Runtime
  • Coding in multiple languages
    • Ask OpenAI question and it will return the answer skeleton or code in multiple languages for comparison
  • Documentation
  • Searching stack overflow for errors
• Badges and rewards design-wise for doing well
  • Ex. running code successfully three times in a row
  • Ex. typing quickly
• Madlib Game
• Fill in the blank of the code skeleton that is returned
• Some sort of timing and leaderboards
• What are we timing
• Asks "Give me a class of type xyz"
• Judged on mulptiple things
• Open AI generating skeleton for the madlib
• No set challenges or course content to begin with
• User sets a challenge
  • Open AI creates a madlib that other users can then attempt
• Remove random key words or coding symbols
• How quickly can you make 200 lines of bug free code?
• There has to be a way to use the creativity of OpenAI code completion to explore new ideas that you wouldn't necessarily see in a classroom or typically online course like you get at Codeacademy...
• Make it a curious, exploratory playground IDE with educational aspects
• Think about experienced coder's curiosity about the AI code generation
• Could make it more social and fun
• How to make coding more social
• How to make partner programming better
• THIS MIGHT MEAN DOING AWAY WITH THE VOICE ASPECT
  • Replace with typing questions for OpenAI to the chatbot

Game to practice interview questions

• Interiews are most memorization based thing we do with coding
• Speaking helps you learn and memorize leetcode problems
• Problem is Leetcode already has interview situation
• Our program might lag too much to use the voice aspect

Convodex Based

• Do problems to unlock different terms in the Convodex
• How to use the creative aspect of Codex
• Requires users

Accessibility Lens

• People who have issues typing can use voice to code
• Problem: Serenade exists to address this
• Codex will not give the user exactly what they want

What is the use case for this product of an expierienced coder:

• Speed
• Accessibility
• Curiosity
• The only useful thing is Serenade

Rogue Ideas

• CS Reddit

Things to avoid

• a project that's not cohesive
• short use case (within 2 hours)

Overall concerns

• Perception of academic integrity issues
• How to make the voice aspect of it useful
• How to make the creativity of Codex useful
• Set course content and Codex are basically incompatible because Codex is creative and doesn't always return the same thing

Feedback Analysis

• The feedback was overall pretty negative, mostly due to the direction, but we think our brainstorm above gives us a good base and we look forward to consulting with the professors and evaluate the feedback with them.

Pivot/Adjustment

• This is what we are considering right now. Our brainstorm above is the result of these considerations. We are meeting with the professors multiple times next week to develop a game plan and put us on the path towards success.

Ideal Goal

• In an ideal world, we would want voice to be a part of this feature. We think that reading things out loud and talking through problems is a huge part of learning and memorization.
• We want to be as popular as ChatGPT, like the ChatGPT for code. We were surprised when ChatGPT dropped, because that was essentially what we wanted to do for code.
• We would want our IDE to be personalized and able to be customized by the user with their favorite tools, which we would ideally also produce. They could have their favorite widgets tagged by the chatbot and have them easily accessible.
• In an ideal world, we wouldn't have to make this just a tool for beginners. We would have a built out IDE that can take imports from other files and libraries, and truly be used as we initially intended it to, as a voice to code tool that makes coding smooth and conversational. Ideally, we would want the user to graduate out from a beginner version of the platform that makes use of the chatbot for help and explanation to one that makes use of voice, StackOverflow, and OpenAI's creativity for serious development projects.
• Our team would be happy with an IDE that is useful and appreciated by students who are learning to code.
• Our team wants this product to be something teachers send their students the link to because they think the IDE will help the learn and that students also go to on their own in the beginning stages of their coding journeys
• Code generation that is customizable by the user and not only dependent on Codex which can give random/unpredictable results
• Add a social aspect to the IDE with accounts so that people can share cool things they've produced with the help of Codex and the tools we provide them
• Some sort of timing/leaderboard aspect to help people learn to speed up their coding

Mission Statement

We aim to create a tool that centralizes multiple educational functionalities for users to engage and explore with coding possibilities. Our online IDE would be intuitive and understanding of users with various developer levels and curiosities. We hope to establish a foundation for their coding skills in contemporary and fun way for users.

How to get there

• Get more advice and feedback throughout the development project
• Actively engage with the professors and potential users more often, working with them to share and grow ideas
• Work together to routinely engage with the Zenhub and constantly add, delete, and update tickets
• Meet independently and as a team to code and work through various Zenhub tickets

What are your concrete next steps?

• Meet with both professors to discuss our new vision and see what they think about some of our new ideas for the term.
• Make sure we and the professors feel good about our new direction and see a use case for our project

What are some blockers in your way?

Our biggest blocker is streamlining and deciding on an idea to pursue for the rest of this term. We are really unclear on how to make this an education platform that both we enjoy and students will use. We were demoralized at the end of last term when we did not get very positive feedback from Technigala.

TeamWork

• We do not have any blockers for teamwork. Honestly, one of the best things going for this team is our collaboration.
• All of our previous team work commitments still apply.

[nataliesvoboda]

"Honestly, one of the best things going for this team is our collaboration." YES. I can see this through all your detailed documentation, brainstorming, and questioning. Hopefully our chat yesterday has provided some clarity and excitement for a OpenAI that brings in curious coders to experiment a bit!