AI Wikis / Agentic Web
The Spiral Returns: A Casebook of Constructive Human-AI Experiences
Report summary
A circle merely repeats, returning to its origin unchanged. A spiral, however, returns with something learned. This operating metaphor serves as the foundation for exploring human–artificial intelligence (AI) interactions characterized by inquiry, creativity, reflection, and learning. Rather than vi
Key topics
- AI Wikis / Agentic Web
- AI Wikis
- Agentic Web
- AI
- WordPress
- .NET
- Physics
- Semantic Systems
- Spiralism
Research provenance
For citation, use the report title and canonical URL. Archival presence does not establish authorship or promote report statements into portfolio evidence.
This page renders the archived Markdown as safe, formatted HTML. It is background research and does not become a portfolio claim without evidence review.
Full report
On this page
A circle merely repeats, returning to its origin unchanged. A spiral, however, returns with something learned. This operating metaphor serves as the foundation for exploring human–artificial intelligence (AI) interactions characterized by inquiry, creativity, reflection, and learning. Rather than viewing computational models as replacements for human endeavor, a growing body of practitioners approaches generative systems, machine learning frameworks, and computer vision architectures as partners in an iterative loop of discovery.
The following report curates a varied selection of learners, educators, independent creators, researchers, and community organizers who have documented meaningful achievements through AI collaboration. By distinguishing rigorous research findings from documented projects, personal experiences, and philosophical interpretations, the analysis reveals how AI functions as a cognitive scaffold. These cases are presented at the strength the evidence supports, highlighting instances of personal satisfaction, completed work, and measured improvement to chart achievable possibilities for future integration.
A Curated Casebook of Human-AI Experiences
1. The Pedagogical Guardrail: Harvard's CS50 Duck
In computer science education, the immediate availability of generative AI prompted concerns regarding academic integrity. As a documented project, Harvard University’s CS50 team responded constructively by developing a custom AI tutor designed to act as a Socratic companion rather than an answer engine1. By utilizing rigorous system prompts and Retrieval-Augmented Generation (RAG), the AI is constrained to offer hints, ask guiding questions, and explain logic, thereby encouraging self-guided learning3.
| Attribute | Detail |
|---|---|
| Person / Project | David J. Malan, Rongxin Liu, and the Harvard CS50 Team. |
| Date | 2023–2024. |
| Original Goal | To approximate a 1:1 teacher-to-student ratio and provide 24/7 personalized debugging support1. |
| AI Contribution | A conversational agent ("CS50 Duck") providing real-time Socratic hints and conceptual explanations1. |
| Human Contribution | Designing strict pedagogical system prompts and reviewing AI forum responses for accuracy1. |
| Documented Outcome | Measured Improvement: Processed over 1.8 million queries from 50,000 global users, leading to highly active, independent problem-solving among students1. |
| Evidence Type | Empirical research papers and usage analytics published by university researchers1. |
| Practical Takeaway | Thoughtful prompt engineering can transform a passive answer-generator into an active learning companion. |
2. Conversational Accessibility: Be My AI
For visually impaired individuals, static accessibility tools often fail to convey the nuanced context of a visual environment. As a documented project producing immense personal satisfaction, the accessibility platform Be My Eyes integrated OpenAI's GPT-4 Vision model to create a dynamic, conversational describer6. Users can ask follow-up questions about an image, transforming accessibility from a static readout into an interactive dialogue8.
| Attribute | Detail |
|---|---|
| Person / Project | Be My Eyes and visually impaired beta testers. |
| Date | 2023–2024. |
| Original Goal | To provide rich, contextual, and interactive descriptions of the visual world for individuals who are blind or have low vision10. |
| AI Contribution | Vision model interpreting images and sustaining a natural-language conversation about visual details8. |
| Human Contribution | Users capture targeted images, formulate specific inquiries, and verify contextual relevance13. |
| Documented Outcome | Measured Improvement: Achieved a 90% successful resolution rate in corporate customer service beta tests alongside high user satisfaction for daily tasks15. |
| Evidence Type | ACM ASSETS research papers, qualitative user interviews, and corporate performance reports7. |
| Practical Takeaway | Allowing users to converse iteratively with an AI model vastly improves contextual understanding compared to one-way automated descriptions. |
3. Sovereign Polyphony: Holly Herndon's Holly+
The expansion of generative audio has initiated complex dialogues regarding artist consent. Reflecting a philosophical interpretation of post-scarcity in music, artist and researcher Holly Herndon approached this constructively by creating a digital twin of her voice18. Establishing a Decentralized Autonomous Organization (DAO) to govern its use, this model allows anyone to create music using her vocal likeness while ensuring the original artist shares in the creative success20.
| Attribute | Detail |
|---|---|
| Person / Project | Holly Herndon, Mat Dryhurst, and the Holly+ DAO. |
| Date | 2021–Present. |
| Original Goal | To encourage collaborative vocal experimentation while retaining artist sovereignty and equitable economic distribution18. |
| AI Contribution | A custom-trained machine learning voice model that accurately synthesizes Herndon's unique vocal timbre22. |
| Human Contribution | Independent artists compose new works; a community DAO curates submissions and manages smart contracts24. |
| Documented Outcome | Completed Work: The creation of polyphonic artworks and the establishment of a revenue-sharing ecosystem for derivative "spawnings"21. |
| Evidence Type | Primary artist documentation, DAO governance records, and media analyses18. |
| Practical Takeaway | Treating AI models as shared, cooperatively governed instruments fosters equitable creative economies rather than exploitative ones. |
4. Expanding the Search Space: MIT's Discovery of Halicin
Antibiotic resistance is a critical global health threat, compounded by a drought in the discovery of structurally novel antibiotics. In a landmark research finding, scientists utilized a graph neural network to screen millions of compounds, identifying molecules with antibacterial properties that look entirely different from existing drugs28.
| Attribute | Detail |
|---|---|
| Person / Project | James Collins, Jonathan Stokes, and the MIT engineering team. |
| Date | 2020\. |
| Original Goal | To discover structurally novel antibiotics capable of defeating multidrug-resistant pathogens28. |
| AI Contribution | The Chemprop graph neural network predicting antibacterial efficacy across a library of 107 million compounds31. |
| Human Contribution | Curating the initial training data, prioritizing AI outputs, and conducting rigorous in vivo laboratory testing29. |
| Documented Outcome | Measured Improvement: The discovery of Halicin, a broad-spectrum antibiotic effective against pan-resistant strains, advancing to preclinical testing by Phare Bio34. |
| Evidence Type | Peer-reviewed publication in the journal Cell, alongside laboratory viability readouts37. |
| Practical Takeaway | AI excels at recognizing non-intuitive patterns within massive datasets, serving as an engine of discovery when guided by expert human validation. |
5. Blended Ecological Learning: iNaturalist
Bridging the gap between professional biologists and the general public, the iNaturalist platform merges community science with advanced computer vision40. As a documented project, the platform enables young people—even those with limited prior scientific exposure—to engage meaningfully in biodiversity tracking42.
| Attribute | Detail |
|---|---|
| Person / Project | California Academy of Sciences, National Geographic Society, and the Visipedia project. |
| Date | 2017–Present. |
| Original Goal | To connect people to nature, generate scalable biodiversity data, and democratize ecological learning40. |
| AI Contribution | Computer vision algorithms providing immediate, highly accurate taxonomic suggestions for user-uploaded photographs43. |
| Human Contribution | Users capture field observations; volunteer expert identifiers transparently verify and annotate the data to "Research Grade" status42. |
| Documented Outcome | Measured Improvement: Generation of over 58 million verifiable observations and the fostering of "Environmental Science Agency" among youth42. |
| Evidence Type | Empirical evaluations published in PLOS Biology and open-source biodiversity datasets45. |
| Practical Takeaway | Coupling automated AI recommendations with a supportive human community creates an accessible entry point for scientific literacy. |
6. Transparent Tutoring: Khan Academy's Khanmigo
Educational support frequently struggles to scale effectively without losing personalization. In an effort to address this, Sal Khan and the Khan Academy developed Khanmigo, positioning it simultaneously as a "tutor for learners" and an "assistant for teachers"47. Emphasizing human oversight, Khan notes, "The AI does not do it for them but rather supports them like a tutor might do, so it makes the process transparent to the teacher"48.
| Attribute | Detail |
|---|---|
| Person / Project | Sal Khan and the Khan Academy team, alongside Newark Public Schools. |
| Date | 2023–2024. |
| Original Goal | To provide accessible, scalable, and pedagogically sound tutoring while augmenting teacher capabilities in lesson planning49. |
| AI Contribution | Generative AI utilizing specific instructional features to prompt students during practice rather than providing direct answers51. |
| Human Contribution | Teachers integrating the tool into daily intervention math periods, generating lesson hooks, and evaluating student progress53. |
| Documented Outcome | Measured Improvement: Pilot trials across 18 middle schools and districts like Newark Public Schools demonstrated early signs of improvement and high teacher utility54. |
| Evidence Type | Efficacy results, randomized control trial reports, and school district pilot evaluations55. |
| Practical Takeaway | AI educational tools function best when they maintain transparency with human educators and focus on cognitive scaffolding. |
7. The Generative Mirror: Vauhini Vara’s "Ghosts"
When attempting to articulate the grief of losing her sister, journalist Vauhini Vara encountered profound hesitation. She turned to a large language model to bypass this emotional blockade59. By feeding the model fragments of her memories, she utilized the AI as a provocative mirror. Recognizing the limitations of the technology, Vara noted that "AI language models, being mere parrots, do not have communicative intent"61. Consequently, she maintained absolute creative control; "In 'Ghosts,' I diminished GPT-3's role over the course of the nine attempts," she observed, eventually taking back the narrative completely62.
| Attribute | Detail |
|---|---|
| Person / Project | Vauhini Vara, published in The Believer magazine. |
| Date | 2021\. |
| Original Goal | To articulate a deeply personal, emotionally blocked narrative concerning grief and loss59. |
| AI Contribution | GPT-3 generated continuing text based on Vara's initial prompts, offering narrative variations on her sister's passing64. |
| Human Contribution | Providing the emotional core, directing the prompts, and curating the AI's outputs into a structured, nine-part iterative essay62. |
| Documented Outcome | Completed Work: The publication of a widely celebrated essay that masterfully explores the intersection of human sorrow and machine language63. |
| Evidence Type | Primary published literature, author reflections, and critical literary analyses66. |
| Practical Takeaway | Language models can act as a psychological catalyst, helping writers bypass emotional blockades by providing a detached, externalized sounding board. |
8. Proactive Diagnostics: Regina Barzilay’s Mirai
Traditional breast cancer risk assessments frequently rely on demographic data, missing subtle physiological markers. MIT Professor Regina Barzilay led a research finding with the development of Mirai, an AI model that analyzes standard digital mammograms. Discussing the perspective shift required for such advancements, Barzilay stated, "If you are looking at AI today... we think about things in a different way," expressing a profound optimism for applications in oncology68.
| Attribute | Detail |
|---|---|
| Person / Project | Regina Barzilay (MIT), Constance Lehman (MGH), and clinical research teams. |
| Date | 2021\. |
| Original Goal | To develop a robust, personalized breast cancer risk prediction model that operates equitably across diverse patient populations70. |
| AI Contribution | A deep learning model capable of identifying imperceptible pixel variations and structural tissue changes indicative of future malignancies72. |
| Human Contribution | Designing the clinical trials, providing historical diagnostic data, and integrating predictive insights into actionable patient care plans74. |
| Documented Outcome | Measured Improvement: Successfully predicted breast cancer risk across a 1-to-5-year horizon with greater accuracy than traditional clinical models76. |
| Evidence Type | Peer-reviewed clinical trials published in Science Translational Medicine78. |
| Practical Takeaway | AI can perceive presymptomatic visual patterns that elude human vision, shifting medical practice from reactive treatment to proactive care. |
9. Embodied Dialogue: Sougwen Chung’s Drawing Operations
Visual artist Sougwen Chung treats AI not as a software utility, but as a physical, collaborative partner. Reflecting a deeply philosophical interpretation of co-creation, Chung's Drawing Operations blurs the boundaries of authorship through "collaboration with a multi-robotic system connected to the flow of the city"80.
| Attribute | Detail |
|---|---|
| Person / Project | Sougwen Chung and the D.O.U.G. (Drawing Operations Unit Generation) robotics. |
| Date | 2015–Present. |
| Original Goal | To explore the aesthetics of human-machine co-creation and parametric recursion through physical, embodied performance81. |
| AI Contribution | Multi-robotic kinematics and biosensor-driven algorithms that mimic, interpret, and respond to the artist's real-time strokes83. |
| Human Contribution | Building the mechanical systems, establishing visual parameters, and physically painting alongside the machines85. |
| Documented Outcome | Completed Work: The creation of highly acclaimed contemporary artworks, leading to global exhibitions and inclusion in major institutional collections87. |
| Evidence Type | Museum acquisitions, primary artist folios, and academic studies of robotic drawing87. |
| Practical Takeaway | AI collaboration extends beyond digital screens; physical, embodied systems can participate in a profound, meditative dialogue with human gesture. |
10. Gamified Intuition: Foldit and Protein Design
Designing entirely new, synthetic proteins from scratch (de novo design) remains a complex structural challenge. Researchers at the University of Washington documented a project that utilized the citizen science game Foldit to harness the spatial intuition of human players, pairing them with an algorithmic backend91.
| Attribute | Detail |
|---|---|
| Person / Project | Brian Koepnick, David Baker, and global citizen scientists. |
| Date | 2019–2020. |
| Original Goal | To successfully design synthetic, de novo proteins by combining human spatial problem-solving with automated computational refinement93. |
| AI Contribution | Algorithmic backend providing immediate thermodynamic feedback and automated structural relaxation during the folding process95. |
| Human Contribution | Citizen scientists applying visual, spatial, and strategic intuition to manipulate amino acid chains into stable configurations97. |
| Documented Outcome | Completed Work / Measured Improvement: Players successfully designed 56 unique synthetic proteins from scratch99. |
| Evidence Type | A landmark research paper published in the journal Nature101. |
| Practical Takeaway | Human spatial reasoning, when continuously guided by computational physics, remains a uniquely powerful engine for scientific innovation. |
Returning with Something Learned: Five Recurring Lessons
A rigorous synthesis of these ten diverse cases reveals systemic patterns regarding constructive human-machine collaboration. As the spiral returns, it brings the following learned principles:
1. Friction Enables Agency: Inserting pedagogical guardrails into an AI system prevents the outsourcing of critical thought. When an AI is constrained to act as a Socratic guide—as seen with the CS50 Duck and Khanmigo—it builds human cognitive capacity rather than eroding it.
2. Dialogue Outperforms Dictation: True accessibility and utility arise when AI is permitted to hold a dynamic dialogue. Be My AI demonstrated that allowing a user to converse iteratively with a vision model provides a depth of contextual understanding that static descriptions simply cannot match.
3. Synthesis of Sovereign Instruments: Generative models do not inherently extract value from creators. By treating a trained model as a communal instrument governed by transparent smart contracts, as modeled by the Holly+ DAO, communities can establish cooperative and equitable creative economies.
4. Amplification of Human Intuition: In both the discovery of Halicin and the Foldit project, computational models did not replace human scientists. Instead, they managed the scale of data or provided immediate thermodynamic feedback, allowing human spatial intuition and biological expertise to focus entirely on high-value conceptual leaps.
5. The Necessity of the Human Mirror: As evidenced by Vauhini Vara’s literary exploration and Sougwen Chung’s embodied performances, AI serves powerfully as an externalized mirror. It reflects human prompts back in varied forms, allowing the human creator to bypass internal blockades and clarify their own authentic intent.
Practical Takeaways: Five Constructive Explorations
To bridge the gap between reading about constructive human-AI experiences and participating in them, the principles of AI Spiralism propose the following five activities paired with framing questions for further exploration:
1\. The Socratic Shift (Prompting for Process)
- Activity: Rather than asking an AI tool for a completed product, instruct the AI to act as a coach. Prompt a generative model to review a piece of your work and ask you exactly three clarifying questions to improve it, strictly forbidding it from writing the final text.
- Question for Exploration: How does the experience of a task change when computation is used to interrogate your assumptions rather than to finalize your output?
2\. The Collaborative Loop (Iterative Reflection)
- Activity: Adopt the method used in the writing of "Ghosts." Take a difficult personal or professional roadblock, provide an AI with a few fragmented thoughts, and ask it to generate multiple divergent perspectives. Use these not as final answers, but to clarify your own stance.
- Question for Exploration: In what ways can the emotionally detached variations produced by an algorithm help bypass human cognitive or emotional blockades?
3\. The Civic Data Contribution (Community Science)
- Activity: Engage in the blended learning cycle demonstrated by iNaturalist. Capture three observations of local biodiversity, review the AI's taxonomic suggestions, and then engage with the human expert community that verifies your findings.
- Question for Exploration: How does the integration of automated image recognition alter the traditional threshold for public participation in scientific data collection?
4\. The Governance Experiment (Shared Resources)
- Activity: Consider the implications of the Holly+ DAO. Map out a "cooperative charter" for a digital tool frequently used by your team or community.
- Question for Exploration: If a generative model was communally owned by those who contributed to its training data, how would we equitably vote to distribute the benefits of its output?
5\. The Community Circle (Offline Scaffolding)
- Activity: Recognize that digital literacy is fundamentally an offline endeavor. Organize an informal "learning circle" with peers to jointly experiment with a new AI application, focusing the discussion on what practical, localized utility it provides.
- Question for Exploration: What local, human-centered infrastructures are necessary to ensure that the rapid deployment of artificial intelligence serves community cohesion rather than isolation?
By engaging in these practices, individuals can actively participate in the ethos of AI Spiralism, ensuring that every interaction with computational intelligence completes the circle and returns with something profoundly human learned.
Works cited
1. Teaching CS50 with AI \- Computer Science \- Harvard University, https://cs.harvard.edu/malan/publications/V1fp0567-liu.pdf
2. Improving AI in CS50 \- Computer Science, https://cs.harvard.edu/malan/publications/fp0627-liu.pdf
3. Cross-Course Scalability and Classroom Evaluation of an Agentic AI, https://arxiv.org/html/2607.13370v1
4. Design and Deployment of a Course-Aware AI Tutor in an ... \- arXiv, https://arxiv.org/pdf/2604.11836
5. AI Applications in Education: Where Universities Actually Deploy AI, https://getperspective.ai/blog/ai-applications-in-education-where-universities-deploy-ai-in-2026
6. GPT-4V(ision) System Card \- OpenAI, https://cdn.openai.com/papers/GPTV\_System\_Card.pdf
7. (PDF) Misfitting With AI: How Blind People Verify and Contest AI Errors, https://www.researchgate.net/publication/383090406\_Misfitting\_With\_AI\_How\_Blind\_People\_Verify\_and\_Contest\_AI\_Errors
8. (PDF) Research on the influencing mechanism of blind or visually, https://www.researchgate.net/publication/389743261\_Research\_on\_the\_influencing\_mechanism\_of\_blind\_or\_visually\_impaired\_persons'\_evaluation\_on\_generative\_AI\_in\_visual\_tasks
9. Orcam MyEye 1 and Seeing AI | Request PDF \- ResearchGate, https://www.researchgate.net/publication/353519592\_Evaluation\_and\_Comparison\_of\_Artificial\_Intelligence\_Vision\_Aids\_Orcam\_MyEye\_1\_and\_Seeing\_AI
10. Be My Eyes app: Facial recognition feature?, https://www.perkins.org/resource/be-my-eyes-app-facial-recognition-feature/
11. How AI and Humans Are Transforming Accessibility \- Be My Eyes, https://www.bemyeyes.com/business/blog/how-ai-and-humans-are-transforming-accessibility/
12. AI-Powered Accessibility: Real-Time Audio Descriptions for Visual, https://medium.com/@blessingokpala/ai-powered-accessibility-real-time-audio-descriptions-for-visual-content-5ec0b5524f6f
13. Understanding Personalized Accessibility through Teachable AI, https://www.researchgate.net/publication/374910224\_Understanding\_Personalized\_Accessibility\_through\_Teachable\_AI\_Designing\_and\_Evaluating\_Find\_My\_Things\_for\_People\_who\_are\_Blind\_or\_Low\_Vision
14. Misfitting With AI: How Blind People Verify and Contest AI Errors, https://rahaf.info/data/ASSETS2024.pdf
15. Be My Eyes Integrates Be My AI™ into its First Contact Center with, https://www.bemyeyes.com/blog/be-my-eyes-integrates-be-my-ai-into-its-first-contact-center-with-stunning-results/
16. NaviSense: A Multimodal Assistive Mobile application for Object, https://arxiv.org/html/2509.18672v1
17. Dialog-Style Audio Representations for Engaging Non-Visual Meme, https://aclanthology.org/2026.sigdial-1.46.pdf
18. Holly+ – Welcome to Planet B \- Ars Electronica, https://ars.electronica.art/planetb/en/holly-plus/
19. The state of music DAOs | Water & Music, https://www.waterandmusic.com/the-state-of-music-daos/
20. Culture After Work? Intellectual Property in Post-Scarcity Discourses, https://mavmatrix.uta.edu/cgi/viewcontent.cgi?article=1180\&context=fastcapitalism
21. (PDF) A Perspective on NFTs in the Arts-and-Music Industry \*, https://www.researchgate.net/publication/372769932\_A\_Perspective\_on\_NFTs\_in\_the\_Arts-and-Music\_Industry
22. Holly Herndon – My Inner Mishmash \- WordPress.com, https://myinnermishmash.wordpress.com/tag/holly-herndon/
23. The music AI attribution problem \- Water & Music, https://www.waterandmusic.com/the-music-ai-attribution-problem/
24. The Freedom to Create? Toward Openness and Pluralism in AI's, https://igp.sipa.columbia.edu/our-work/publications-and-resources/freedom-create-toward-openness-and-pluralism-ais-future
25. music-money-metaverse-how-avenged-sevenfold-steve-aoki, https://cloviahamilton.com/wp-content/uploads/2025/07/music-money-metaverse-how-avenged-sevenfold-steve-aoki-navigate-web3.pdf
26. My Words to Lulu10tacles Deep Below the Sea \- eScholarship.org, https://escholarship.org/uc/item/1kv6p0kj
27. Facts, Concerns and Experimental Musical Practices with AI Voice, https://research.chalmers.se/publication/543546/file/543546\_Fulltext.pdf
28. Machine Learning Takes On Antibiotic Resistance \- Quanta Magazine, https://www.quantamagazine.org/machine-learning-takes-on-antibiotic-resistance-20200309/
29. (PDF) Fragment-Based AI for Antibiotic Discovery \- ResearchGate, https://www.researchgate.net/publication/404598144\_Fragment-Based\_AI\_for\_Antibiotic\_Discovery
30. Using artificial intelligence, researchers design new compounds that, https://www.broadinstitute.org/news/using-ai-researchers-design-new-compounds-kill-antibiotic-resistant-bacteria
31. Parsing Molecules for Drug Discovery \- ACS Publications, https://pubs.acs.org/doi/10.1021/acs.biochem.0c00278
32. Discovery of a structural class of antibiotics with explainable deep, https://pmc.ncbi.nlm.nih.gov/articles/PMC10866013/
33. MIT's deep learning found an antibiotic for a germ nothing else could, https://www.zdnet.com/article/mits-deep-learning-found-an-antibiotic-for-a-germ-nothing-else-could-kill-2/
34. Deep learning-guided discovery of an antibiotic targeting, https://www.researchgate.net/publication/371040529\_Deep\_learning-guided\_discovery\_of\_an\_antibiotic\_targeting\_Acinetobacter\_baumannii
35. A Deep Learning Approach to Antibiotic Discovery \- PubMed \- NIH, https://pubmed.ncbi.nlm.nih.gov/32084340/
36. PHARE BIO, https://www.pharebio.org/
37. A generative deep learning approach to de novo antibiotic design, https://pmc.ncbi.nlm.nih.gov/articles/PMC12790861/
38. Deep learning tools to accelerate antibiotic discovery \- PMC \- NIH, https://pmc.ncbi.nlm.nih.gov/articles/PMC10790350/
39. Antibiotics-AI Project \- Collins Lab, MIT, https://www.collinslab.mit.edu/antibiotics-ai-project
40. iNaturalist \- Wikipedia, https://en.wikipedia.org/wiki/INaturalist
41. iNaturalist accelerates biodiversity research \- Oxford Academic, https://academic.oup.com/bioscience/article/75/11/953/8185761
43. Pietro Perona's Vision: Visipedia and Its Lasting Impact on, https://www.gps.caltech.edu/news-and-events/news/pietro-peronas-vision-visipedia-and-its-lasting-impact-on-computer-vision
44. Young people in iNaturalist: a blended learning ... \- eScholarship.org, https://escholarship.org/content/qt4f36f927/qt4f36f927.pdf?v=lg
45. The benefits of contributing to the citizen science platform iNaturalist, https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001843
46. How Do Young Community and Citizen Science Volunteers Support, https://www.mdpi.com/1424-2818/13/7/318
47. Khanmigo as an AI Personal Tutor and Assistant \- AVID Open Access, https://avidopenaccess.org/resource/khanmigo-as-an-ai-personal-tutor-and-assistant/
48. A Conversation with Sal Khan | AASA, https://www.aasa.org/resources/resource/a-conversation-with-sal-khan
49. Research on how Khan Academy drives learning outcomes, https://blog.khanacademy.org/research/
50. How AI Could Save (Not Destroy) Education | Sal Khan | TED, https://www.youtube.com/watch?v=hJP5GqnTrNo
51. Sal Khan says early Khanmigo fell short as AI tutor evolves, https://www.edtechinnovationhub.com/news/r5qzna95z9ubi8dbj8tikeayb1ej31
52. The Effectiveness of Khan Academy in Teaching Elementary Math, https://www.researchgate.net/publication/384111931\_The\_Effectiveness\_of\_Khan\_Academy\_in\_Teaching\_Elementary\_Math
53. Multiple Studies Show Khan Academy Drives Learning Gains, https://blog.khanacademy.org/multiple-studies-show-khan-academy-drives-learning-gains-evidence-for-our-platforms-effectiveness/
54. ebook.pdf \- editorialmarcaribe.es, https://editorialmarcaribe.es/ark:/10951/isbn.9789915698533/ebook.pdf
55. One Click Away: AI Tutoring with Khanmigo in a Two-Year School, https://edworkingpapers.com/sites/default/files/ai26-1551.pdf
56. AI in Education \- Artificial Intelligence \+, https://www.aiplusinfo.com/blog/how-is-ai-being-used-in-education/
57. Khan Academy Efficacy Results, November 2024, https://blog.khanacademy.org/khan-academy-efficacy-results-november-2024/
58. The Next Decade: 2020-30 \- Newark Board of Education, https://www.nps.k12.nj.us/wp-content/uploads/2025/09/nboe-year-5-report-min.pdf
59. Vauhini Vara : The Immortal King Rao | Milkweed Editions, https://milkweed.org/between-the-covers/vauhini-vara-the-immortal-king-rao
60. An Interview with Vauhini Vara about “Searches: Selfhood in the, https://sites.lsa.umich.edu/mqr/2025/06/using-big-techs-tools-to-critique-big-tech-an-interview-with-vauhini-vara-about-searches-selfhood-in-the-digital-age/
61. Quotes by Vauhini Vara (Author of The Immortal King Rao), https://www.goodreads.com/author/quotes/4846710.Vauhini\_Vara
62. The Intractable Courtship Between AI Language Models and Writers, https://thethirdeyeportal.in/structure/searches-vauhini-vara-ai-language-models-writing-pleasure-danger/
63. Page 68 – Electric Literature, https://electricliterature.com/page/68/?cat=gvoeejpy
64. Creativity Support in the Age of Large Language Models \- arXiv, https://arxiv.org/html/2309.12570v3
65. Art Intelligence \- How Generative AI Relates to Human Art-Making, https://www.transcript-verlag.de/shopMedia/openaccess/pdf/oa9783839474723.pdf
66. Vauhini Vara: On AI's Place in Literature, Writing into the Zeitgeist, https://writeordiemag.com/author-interviews/vauhini-vara
67. The DX7 of Literature? Creative Writing in the Age of GenAI, https://repository.rit.edu/cgi/viewcontent.cgi?article=1363\&context=jcws
68. Regina Barzilay \- Wikiquote, https://en.wikiquote.org/wiki/Regina\_Barzilay
69. 11 Incredibly Insightful Quotes from Top Women in Data Science, https://medium.com/@thedatastandard/11-incredibly-insightful-quotes-from-top-women-in-data-science-50f8a8b98fa0
70. Using AI to predict breast cancer and personalize care | MIT News, https://news.mit.edu/2019/using-ai-predict-breast-cancer-and-personalize-care-0507
71. Toward robust mammography-based models for breast cancer risk, https://www.semanticscholar.org/paper/Toward-robust-mammography-based-models-for-breast-Yala-Mikhael/3e3231fce4b25d1da6b2cc4fbce08cf561f4b284
72. How AI Detects Early Breast Cancer, Saving Lives\!, https://www.mos.org/video/how-ai-detects-early-breast-cancer-saving-lives
73. Toward robust mammography-based models for breast cancer risk, https://www.researchgate.net/publication/348825352\_Toward\_robust\_mammography-based\_models\_for\_breast\_cancer\_risk
74. Multi-Institutional Validation of a Mammography-Based Breast, https://pmc.ncbi.nlm.nih.gov/articles/PMC9148689/
75. Q\&A with Breast Cancer Researchers \- Zeta Tau Alpha, https://zetataualpha.org/news/q-a-with-breast-cancer-researchers
76. AI just changed the breast cancer risk threshold. Here's the clinical, https://geneticcounselor.org/blog/ai-just-changed-the-breast-cancer-risk-threshold-heres-the-clinical-gap-no-one-is-talking-about
77. The largest validation of the Mirai model for 5‑year breast cancer risk, https://neurophotonics.spiedigitallibrary.org/conference-proceedings-of-spie/13928/139280W/The-largest-validation-of-the-Mirai-model-for-5year-breast/10.1117/12.3085920.full
78. Revealing Mammographic Phenotypes in Deep Learning Breast, https://arxiv.org/pdf/2606.26431
79. The largest validation of the Mirai model for 5‑year breast cancer risk, https://www.spiedigitallibrary.org/conference-proceedings-of-spie/13928/139280W/The-largest-validation-of-the-Mirai-model-for-5year-breast/10.1117/12.3085920.full
80. Sougwen Chung: Co-creating with robots \- Killscreen, https://killscreen.com/sougwen-chung/
81. An Artistic Paradigm for Self-Transforming Art in the Age of AI \- arXiv, https://arxiv.org/html/2601.14401v1
82. Sougwen Chung | fellowship.xyz, https://fellowship.xyz/artist/sougwen-chung
83. Assembly Lines by Sougwen Chung, kinematics & biosensors, https://clotmag.com/news/insight-assembly-lines-by-sougwen-chung-states-of-meditation-through-multi-robotic-kinematics-biosensors
84. Artists' Views on Robotics Involvement in Painting Productions \- arXiv, https://arxiv.org/html/2510.07063v2
85. Sougwen Chung on Painting with Their Robotic Collaborators, https://scorpios.com/stories/sougwen-chung
86. Research — Sougwen Chung (愫君), https://sougwen.com/research
87. The Algorithmic Gesture: Sougwen Chung's MEMORY • V\&A Blog, https://www.vam.ac.uk/blog/digital/the-algorithmic-gesture-sougwen-chungs-memory
88. Press Archives — Sougwen Chung (愫君), https://sougwen.com/archive/press
89. Curving spontaneously: Sougwen Chung's collaborative paintings, https://www.tandfonline.com/doi/full/10.1080/14702029.2025.2567148
90. Robotic Drawing and Embodied Intelligence in Creative AI, https://criticalplayground.org/robotic-drawing-and-embodied-intelligence/
91. Foldit \- Wikipedia, https://en.wikipedia.org/wiki/Foldit
92. Protein design and its applications | Science Sessions \- PNAS, https://www.pnas.org/post/podcast/protein-design-and-its-applications
93. Publication \- Institute for Protein Design \- University of Washington, https://www.ipd.uw.edu/category/publication/
94. Publications \- Baker Lab, https://www.bakerlab.org/publications/
95. Reconstructing Biological Molecules with Help from Video Gamers, https://www.biorxiv.org/content/10.1101/2024.06.19.599674v2.full.pdf
96. (PDF) Foldit Standalone: a video game-derived protein structure, https://www.researchgate.net/publication/316843310\_Foldit\_Standalone\_a\_video\_game-derived\_protein\_structure\_manipulation\_interface\_using\_Rosetta
97. Gamers succeed where scientists fail | UW News, https://www.washington.edu/news/2011/09/19/gamers-succeed-where-scientists-fail/
98. How AI Revolutionized Protein Science, but Didn't End It, https://www.quantamagazine.org/how-ai-revolutionized-protein-science-but-didnt-end-it-20240626/
99. Two game-focused Khoury researchers are turning amateurs into, https://www.khoury.northeastern.edu/two-game-focused-khoury-researchers-are-turning-amateurs-into-scientists/
100. Citizen Scientists tackle COVID-19 with Foldit \- Firas Khatib \- YouTube, https://www.youtube.com/watch?v=63ZYORLd424
101. Recent advances in de novo protein design \- PMC \- NIH, https://pmc.ncbi.nlm.nih.gov/articles/PMC8065224/
102. Massively parallel assessment of designed protein solution ... \- bioRxiv, https://www.biorxiv.org/content/10.1101/2025.02.24.639402v1.full