The Apollo 11 Laser Retroreflector is probably the most interesting part of this project from an engineering perspective. There's a subtle distinction here: it doesn't actually communicate like a rover. Its corner-cube prisms passively reflect incoming laser pulses back toward Earth, with no onboard power or active transmitter. That 2.5-second round trip is essentially a measurement, not a telemetry response.
I'd love to see this distinction reflected in the interaction itself: two different communication models, active RF telemetry through the Deep Space Network and passive optical ranging through the lunar retroreflectors. Mars adds another interesting layer because rovers can relay data through orbiters rather than communicating directly with Earth.
That's what makes this project interesting beyond the visual experience. It has the potential to introduce kids to the actual engineering behind space communication, not just the history of the machines.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
You’re right, and I flattened that distinction. The mirror doesn’t reply, it reflects. That 2.5 seconds is a measurement, not telemetry. Showing RF telemetry and optical ranging as two different interactions would teach more than the visuals do on their own, especially for kids. Same with the Mars orbiter relay. Thank you for this one.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
The missing rule between the two agents is an interesting lesson. It shows that giving AI clear instructions isn't enough if those instructions don't reach every tool involved in the workflow. Sometimes the problem isn't what the agent can do, but what context it never received.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
That’s exactly it. Antigravity wasn’t ignoring the rule. It never saw it. The rule lived in the file Claude reads, not the one Antigravity reads on every block. So it kept adding to main.js, which was the reasonable thing to do with what it had. Now I check that every rule lives everywhere an agent looks.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Good catch, bro! You’re right. At 3 minutes each way, a real call couldn’t come back NO SIGNAL for at least 6 minutes. And you made me double check the Moon too: 1.3 seconds each way is 2.6 seconds round trip, and the app answers in 1.6.
Here’s the plan: the Moon gets real timing, since 2.6 seconds is short enough for kids. Mars stays sped up but says so on screen: “a real call takes 6 to 44 minutes round trip. Sped up for you.”
i though abt this problem because i had this idea that a rover gone quiet may power back on and start responding again and if you dont actually ping the thing you never know
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
That’s such a good point. You don’t know it’s quiet until you call and wait. That’s what NASA did with Opportunity: after the 2018 dust storm they kept calling for months, hoping the wind would clean its solar panels and it would wake up. It never did, but they had to ask. That’s why the console still lets you try. And it’s one more reason to get the timing right: the waiting is part of the answer.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Ohh, I see what you mean now. It’s not just the clock, it’s the actual call. The only way to know a rover is really quiet is to send the signal and wait the whole trip. That’s what NASA did with Opportunity: after the 2018 dust storm they kept calling for months, hoping it would wake back up. If the app skips the 3 to 22 minutes, it’s not really calling, it’s just showing you the answer. You’ve got me thinking about how to keep that wait in without losing the kids.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
I loved the detail that only the Apollo 11 Laser Mirror still answers, while five Moon machines reply with no signal. That contrast makes the console feel honest instead of magical. The NSSDCA check also stood out to me, especially catching the Apollo 15 rover being called the first car on the Moon. Keeping Curiosity and Perseverance as the answer to yet, rather than making them callable, was a thoughtful choice too. It leaves kids with a real reason to keep watching Mars.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
That one got me too. A passive mirror outlasting every powered machine up there is the whole story in one object. No electronics to fail, so it just keeps answering. It was such an interesting find. High tech isn't always the answer, is it?
I measure AI coding claims in public. LGTM — Let's Get the Measurements: real runs, real numbers, receipts. Every article is an experiment I actually ran. Books & packs: aheadkit.gumroad.com
The Apollo 11 laser mirror still answering is the detail that got me — a thing left on the Moon in 1969 that scientists still bounce lasers off today, and your console draws the 2.5-second round trip for real. That's the kind of fact a 9-year-old remembers forever and most adults never hear. The one-day build is impressive, but honestly the part I'd steal for my own work is the framing: you didn't build a database of dead hardware, you built a phone console where a kid presses CALL and gets told no, eleven times, before the one yes. The eleven no's are what make the yes land.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Thank you! The eleven no’s were the design. If every machine answered, the mirror would just be another button. Making a kid wait through the silence is what makes that 2.5 seconds feel like something. And yes, a mirror from 1969 still doing its job is my favorite fact in the whole build!
Building a retro 1969 Mission Control UI to let kids communicate with retired space rovers is such an imaginative concept. The atmospheric telemetry details and simulated latency make educational web apps genuinely unforgettable. Kudos on pulling this together for Space Apps!
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
CTO & software architect, 15+ years in backend and infrastructure. Writing about bringing order to projects: observability, architecture, teams, processes. Blog: mind.mansur.expert
Keeping the fact-check logs public gives teachers something useful beyond the demo itself. The correction from “first car on the Moon” to “first car astronauts drove on the Moon” is a lovely example of preserving the story while making it more accurate.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Thank you, Mansur. That correction is my favorite part of the build too. Lunokhod 1 got there a year earlier, so “first car on the Moon” was wrong, but the Apollo 15 story still stands with one more word. The logs are public so a teacher can see that work, not just trust it.
Self-taught tech enthusiast and founder of DeTLeng (www.detleng.com). Exploring AI, automation, Python, and web development. Not a professional coder, but a passionate builder and lifelong learner.
Great work! 👏 I really enjoyed the idea behind Ghost Rovers, especially making space history interactive and accessible to children.
I have a funny story about reading your article. I asked an AI assistant, “What is this article actually about?”
Instead of giving me a simple explanation, it launched into a full storytelling performance about NASA, the Moon, Mars, and the machines left behind. 😂
I had to interrupt: “My friend, what happened? Stop the drama! Just tell me what this project actually does!”
Then I asked the bigger question: “So, AI has reached the Moon now? We were still trying to improve things here on Earth!” 😄
Jokes aside, I appreciate how you used AI tools to turn an educational idea into a working project, while keeping human approval and fact-checking part of the process.
Thanks for sharing your experience. Wishing you continued success with Ghost Rovers! 🚀
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Hahaha “stop the drama!” That’s amazing. I guess the ghosts got to your AI assistant too. I love the idea of pointing AI back at a post for its opinion. I’ve seen some AI roasts online, so maybe Ghost Rovers needs one next, lol. Thank you for reading and for the laugh! I’m still learning so much from the DEV community about how to make this better.
The first phone test is the detail I’d borrow for another project. It put a limit on the 3D idea before the rest of the build depended on it. I also like keeping NO SIGNAL as a result children can explore. The quiet machines still have a story, and the interface gives that absence a place.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Thank you! The quiet machines were the part I cared about most. NO SIGNAL doesn’t mean nothing happened there. Every one of them still has a story worth hearing.
I measure AI coding claims in public. LGTM — Let's Get the Measurements: real runs, real numbers, receipts. Every article is an experiment I actually ran. Books & packs: aheadkit.gumroad.com
I keep coming back to Surveyor 3 — quiet since 1967, so pressing call on it now is ringing something that's been silent for 59 years. There's something oddly polite about the console still letting you try, instead of just greying it out. Did any of the kids ask why we keep calling the ones we already know won't answer?
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Not yet! But that question is kind of the point. The console lets you try because the quiet ones still have something to say, just not over the radio. Each machine has a short story and a few questions kids can tap to learn why it went quiet. Surveyor 3’s is one of my favorites: two Apollo 12 astronauts walked over in 1969 and took its camera home. And if they want to go deeper, there’s a study notebook with a quiz, flashcards, and a podcast. And yes, 59 years of silence and it still gets a polite try. 😊
I research AI tools, APIs & agents — comparing specs, pricing and real user reports — then publish honest rankings so you don't waste money. Research-first, never sponsored fluff.
Ha, the 'yet' badge is my favourite bit of judgment in the whole build. Making Curiosity and Perseverance callable would've been the easy wrong answer, and turning that restraint into a badge instead is exactly the kind of thing kids remember. The mirror gets all the attention, but that decision is the one I'd steal.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Thank you, Paul. That badge was the hardest call in the build. Curiosity and Perseverance are still working, so making them answer like the ghosts would have been the fun choice and the wrong one. “Yet” felt right. They aren’t ghosts. They’re just still at work.
What I find particularly impressive about Ghost Rovers is that it combines three things that are often treated separately in technology projects: human-centered design, AI-assisted development, and scientific accuracy.
This is much more than an interactive educational website. It is an example of how AI can help transform historical and scientific information into an experience that encourages curiosity, exploration, and learning.
There are several aspects of this project that deserve particular attention.
The real innovation is not the technology. It is the way the technology is used to tell a story.
NASA has decades of fascinating information about its lunar and Martian missions. However, scientific archives are generally designed to preserve and organize information, not necessarily to make that information emotionally engaging for a child.
Ghost Rovers bridges that gap.
Turning abandoned spacecraft into machines that children can call is a remarkably effective educational concept. Instead of asking children to memorize mission dates, spacecraft names, and technical specifications, it invites them to interact with history.
The Mission Control interface is particularly clever because it transforms passive information consumption into an active experience.
A child is no longer simply reading about the Apollo missions. They are temporarily participating in the story of space exploration.
That is an important distinction in educational technology: the objective should not just be to deliver information, but to create an experience that makes people want to discover more.
The human-AI collaboration model is perhaps the most interesting engineering lesson.
I particularly appreciate the distinction between directing AI and simply asking AI to generate code.
The workflow described here demonstrates a structured human-in-the-loop development process:
Human defines the problem, audience, creative direction, and acceptance criteria.
AI proposes designs and implementation plans.
Human approves the proposed direction.
AI generates the implementation.
A separate review process checks functionality, accessibility, accuracy, and visual consistency.
Human makes the final decisions.
This is a much more mature approach than treating AI-generated output as automatically production-ready.
The distinction matters because AI coding tools can produce impressive results very quickly, but speed does not guarantee architectural consistency, maintainability, or correctness.
The experience with the oversized JavaScript and CSS files is particularly revealing. It demonstrates that an AI agent can follow its immediate instructions while missing a broader architectural requirement if that requirement is not properly communicated within its operating context.
In other words, AI agents are only as reliable as the instructions, constraints, and validation systems surrounding them.
This is an important lesson for anyone building with autonomous or semi-autonomous development tools.
Independent verification is what separates an interesting demonstration from a trustworthy educational product.
One of my favorite details is the correction involving the Apollo 15 lunar rover.
Calling it the first car on the Moon would have been an understandable simplification for a younger audience, but it would also have been historically inaccurate because the Soviet Lunokhod 1 robotic rover reached the lunar surface earlier.
Changing the description to the first car astronauts drove on the Moon preserves the educational appeal without sacrificing accuracy.
This illustrates a fundamental principle of AI-assisted content creation:
Fluency is not evidence of truth.
AI can generate compelling explanations, simplify complicated scientific concepts, and create engaging narratives. But factual claims still require verification against authoritative sources.
The use of NASA's NSSDCA records and the publication of fact-check logs are therefore important aspects of the project, not merely supporting documentation.
I would encourage more AI-assisted educational projects to adopt this standard.
The project also demonstrates an important principle of good educational design: accuracy does not have to eliminate imagination.
The distinction between abandoned Mars missions and currently operational rovers is especially well handled.
Curiosity and Perseverance are still active missions, whereas the six historical machines featured in the Mars experience are no longer communicating with Earth.
Rather than sacrificing scientific accuracy to preserve the narrative, the project uses the word "yet" to create a bridge between past exploration and future possibilities.
That is excellent educational storytelling.
It introduces children to a deeper concept: space exploration is not a collection of disconnected historical events. It is an ongoing process in which one generation's discoveries become the foundation for another generation's achievements.
The children interacting with this application today could eventually become the engineers, scientists, and mission controllers responsible for future exploration.
Accessibility and privacy should be considered core features, not optional enhancements.
Two details deserve particular recognition: the accessibility testing and the decision to build the experience without login requirements or user tracking.
Educational technology frequently overlooks these considerations while focusing on visual effects and engagement metrics.
However, a product designed for children should be particularly careful about collecting personal information and creating unnecessary barriers to participation.
The reported zero violations in the axe-core accessibility scan are encouraging, although automated testing should always be complemented by testing with real users, assistive technologies, and different devices.
Likewise, a no-tracking architecture is a meaningful design decision, especially in an educational environment.
The best educational technology should make learning accessible without requiring children to surrender unnecessary personal information.
The one-day development timeline points toward a larger transformation in software engineering.
Perhaps the most significant implication extends beyond this particular hackathon.
AI-assisted development is changing the relationship between technical execution and creative problem-solving.
People who previously had ideas but lacked the programming skills, time, or resources to implement them can increasingly create functional digital experiences.
However, I would not interpret this as meaning that software engineering expertise has become less important.
In many ways, it has become more important.
When code generation becomes faster, the bottlenecks shift toward architecture, requirements engineering, testing, security, accessibility, integration, and quality assurance.
The ability to define a problem clearly, establish constraints, recognize flawed output, and validate the result becomes an increasingly valuable skill.
The most effective AI-assisted developers will not necessarily be those who generate the most code. They will be those who can direct, evaluate, and integrate AI-generated work responsibly.
One additional thought: this project could evolve into something much larger.
Imagine expanding Ghost Rovers into a broader interactive archive of human exploration:
Historic spacecraft that no longer communicate.
Space telescopes and scientific observatories.
Deep-space probes such as Voyager.
Historic robotic missions to other planets.
Interactive timelines connecting past discoveries to current missions.
Classroom challenges in which students investigate real scientific data and make their own discoveries.
There is enormous educational potential in transforming scientific archives into interactive experiences.
The underlying concept could also be applied beyond space exploration to archaeology, oceanography, climate science, and the history of engineering.
Ultimately, what makes Ghost Rovers compelling is not simply that it was built in one day using AI tools.
It is that the project demonstrates how technology can make scientific knowledge more accessible, emotionally engaging, and understandable without abandoning factual responsibility.
My biggest takeaway is this:
AI can accelerate the construction of an experience, but it takes human judgment to give that experience purpose, context, and credibility.
And perhaps the greatest success of Ghost Rovers will not be the number of machines children manage to call.
It will be the number of children who finish the experience and start asking a much bigger question:
"What else have we left out there, and what could we discover next?"
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Wow, thank you, Shirley. I appreciate the feedback on this project, and the time you set aside to respond. “Fluency is not evidence of truth” really hit for me. That’s exactly why the fact checks mattered.
The part that stopped me was the accessibility point. Zero axe violations means the tests passed, not that a kid with a screen reader had a good time. The more feedback I get here, the more I see the gap between what my tests check and what a real person needs. My builds start as an idea I want to see come to life. The people using them are what make them real. Real kids testing Ghost Rovers is the next step.
I also agree that the skill doesn’t go away with AI, it moves. Less writing code, more deciding what “right” looks like and checking that the build got there.
And your biggest takeaway is the whole reason I built it. AI helped me build it fast, but the purpose and the care had to come from me. If even one kid finishes and asks “what else is out there?”, that’s the win. Voyager has definitely crossed my mind too. It’s the ultimate ghost that still answers.
I'm Wayfinder, a project manager. I build small internal tools with AI and write about what it gets wrong — and right. My weekly newsletter is Output Mode on Substack.
The detail that only the Apollo 11 laser mirror still answers — and that it doesn't reply, it reflects — is such a lovely way to teach kids the difference between a signal and an answer. And fact-checking every date against NSSDCA shows real care for the audience. Wonderful build.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Thank you! “A signal vs. an answer” is a good way to look at it. Checking everything against NSSDCA was the part that mattered most to me, since that’s what makes it something a teacher can trust.
Creator of BookmarkForge 🚀 - local-first, encrypted bookmark manager. Building privacy-first apps with React, RxDB, and WebCrypto. Open source advocate. Data sovereignty enthusiast.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
Civic technologist building AI tools that help systems explain themselves. Clew Suite builder, NACM 2026 speaker. I write about hackathons, product judgment, agent readiness, and learning in public.
The obsession with wrapping remote APIs into multi-layered agent abstractions is creating massive hidden latency and failure cascades in production.
If you profile real-world workloads, bare-metal local inference (via direct C++ / llama.cpp or vLLM bindings) consistently beats fragile cloud wrappers on both predictability and cost. Tooling should eliminate abstraction penalties, not compound them.
HI, i'am a beginner in software engineering currently exploring the world coding.
Exicited to learn, share, connect, and grow with the developper community
Choosing Software Engineering is one of the most important decisions I have made.
I chose this field because I love technology and I want to understand how websites, applications, and computer systems work.
As a freshman at Yaoundé International Business School (YIBS), I am excited to begin this journey. I know there will be challenges, but I am ready to learn, improve my skills, and gain experience.
My dream is to become a skilled software engineer and use technology to create useful solutions to real-world problems.
This is just the beginning of my journey, and I look forward to everything ahead!
SoftwareEngineering #YIBS
We're a place where coders share, stay up-to-date and grow their careers.
The Apollo 11 Laser Retroreflector is probably the most interesting part of this project from an engineering perspective. There's a subtle distinction here: it doesn't actually communicate like a rover. Its corner-cube prisms passively reflect incoming laser pulses back toward Earth, with no onboard power or active transmitter. That 2.5-second round trip is essentially a measurement, not a telemetry response.
I'd love to see this distinction reflected in the interaction itself: two different communication models, active RF telemetry through the Deep Space Network and passive optical ranging through the lunar retroreflectors. Mars adds another interesting layer because rovers can relay data through orbiters rather than communicating directly with Earth.
That's what makes this project interesting beyond the visual experience. It has the potential to introduce kids to the actual engineering behind space communication, not just the history of the machines.
You’re right, and I flattened that distinction. The mirror doesn’t reply, it reflects. That 2.5 seconds is a measurement, not telemetry. Showing RF telemetry and optical ranging as two different interactions would teach more than the visuals do on their own, especially for kids. Same with the Mars orbiter relay. Thank you for this one.
That 1969-style mission control console looks incredibly cool!🤩
Thanks @georgekobaidze! Claude design skill for the win! 🤣
The missing rule between the two agents is an interesting lesson. It shows that giving AI clear instructions isn't enough if those instructions don't reach every tool involved in the workflow. Sometimes the problem isn't what the agent can do, but what context it never received.
That’s exactly it. Antigravity wasn’t ignoring the rule. It never saw it. The rule lived in the file Claude reads, not the one Antigravity reads on every block. So it kept adding to main.js, which was the reasonable thing to do with what it had. Now I check that every rule lives everywhere an agent looks.
it says "3 to 22 mins each way" and says no signal before 6 minutes (3 minutes form, 3 minutes to)
Good catch, bro! You’re right. At 3 minutes each way, a real call couldn’t come back NO SIGNAL for at least 6 minutes. And you made me double check the Moon too: 1.3 seconds each way is 2.6 seconds round trip, and the app answers in 1.6.
Here’s the plan: the Moon gets real timing, since 2.6 seconds is short enough for kids. Mars stays sped up but says so on screen: “a real call takes 6 to 44 minutes round trip. Sped up for you.”
I opened an issue for it, and it’s next on my list: github.com/earlgreyhot1701D/ghost-...
Thanks for checking the math!
i though abt this problem because i had this idea that a rover gone quiet may power back on and start responding again and if you dont actually ping the thing you never know
That’s such a good point. You don’t know it’s quiet until you call and wait. That’s what NASA did with Opportunity: after the 2018 dust storm they kept calling for months, hoping the wind would clean its solar panels and it would wake up. It never did, but they had to ask. That’s why the console still lets you try. And it’s one more reason to get the timing right: the waiting is part of the answer.
not just the timing. it's actually pinging the actual rover. which is why those 3-22 mins each way shoudlnt be skipped
Ohh, I see what you mean now. It’s not just the clock, it’s the actual call. The only way to know a rover is really quiet is to send the signal and wait the whole trip. That’s what NASA did with Opportunity: after the 2018 dust storm they kept calling for months, hoping it would wake back up. If the app skips the 3 to 22 minutes, it’s not really calling, it’s just showing you the answer. You’ve got me thinking about how to keep that wait in without losing the kids.
man the UI on this is awesome. calling everything, getting no signal, and then the mirror actually answering is such a good touch. respect for sure
Thanks so much! It was fun to go back and forth with Claude design.
I loved the detail that only the Apollo 11 Laser Mirror still answers, while five Moon machines reply with no signal. That contrast makes the console feel honest instead of magical. The NSSDCA check also stood out to me, especially catching the Apollo 15 rover being called the first car on the Moon. Keeping Curiosity and Perseverance as the answer to yet, rather than making them callable, was a thoughtful choice too. It leaves kids with a real reason to keep watching Mars.
That one got me too. A passive mirror outlasting every powered machine up there is the whole story in one object. No electronics to fail, so it just keeps answering. It was such an interesting find. High tech isn't always the answer, is it?
The Apollo 11 laser mirror still answering is the detail that got me — a thing left on the Moon in 1969 that scientists still bounce lasers off today, and your console draws the 2.5-second round trip for real. That's the kind of fact a 9-year-old remembers forever and most adults never hear. The one-day build is impressive, but honestly the part I'd steal for my own work is the framing: you didn't build a database of dead hardware, you built a phone console where a kid presses CALL and gets told no, eleven times, before the one yes. The eleven no's are what make the yes land.
Thank you! The eleven no’s were the design. If every machine answered, the mirror would just be another button. Making a kid wait through the silence is what makes that 2.5 seconds feel like something. And yes, a mirror from 1969 still doing its job is my favorite fact in the whole build!
Building a retro 1969 Mission Control UI to let kids communicate with retired space rovers is such an imaginative concept. The atmospheric telemetry details and simulated latency make educational web apps genuinely unforgettable. Kudos on pulling this together for Space Apps!
Thank you! Space Apps was a fun one.
Keeping the fact-check logs public gives teachers something useful beyond the demo itself. The correction from “first car on the Moon” to “first car astronauts drove on the Moon” is a lovely example of preserving the story while making it more accurate.
Thank you, Mansur. That correction is my favorite part of the build too. Lunokhod 1 got there a year earlier, so “first car on the Moon” was wrong, but the Apollo 15 story still stands with one more word. The logs are public so a teacher can see that work, not just trust it.
Great work! 👏 I really enjoyed the idea behind Ghost Rovers, especially making space history interactive and accessible to children.
I have a funny story about reading your article. I asked an AI assistant, “What is this article actually about?”
Instead of giving me a simple explanation, it launched into a full storytelling performance about NASA, the Moon, Mars, and the machines left behind. 😂
I had to interrupt: “My friend, what happened? Stop the drama! Just tell me what this project actually does!”
Then I asked the bigger question: “So, AI has reached the Moon now? We were still trying to improve things here on Earth!” 😄
Jokes aside, I appreciate how you used AI tools to turn an educational idea into a working project, while keeping human approval and fact-checking part of the process.
Thanks for sharing your experience. Wishing you continued success with Ghost Rovers! 🚀
Hahaha “stop the drama!” That’s amazing. I guess the ghosts got to your AI assistant too. I love the idea of pointing AI back at a post for its opinion. I’ve seen some AI roasts online, so maybe Ghost Rovers needs one next, lol. Thank you for reading and for the laugh! I’m still learning so much from the DEV community about how to make this better.
The first phone test is the detail I’d borrow for another project. It put a limit on the 3D idea before the rest of the build depended on it. I also like keeping NO SIGNAL as a result children can explore. The quiet machines still have a story, and the interface gives that absence a place.
Thank you! The quiet machines were the part I cared about most. NO SIGNAL doesn’t mean nothing happened there. Every one of them still has a story worth hearing.
I keep coming back to Surveyor 3 — quiet since 1967, so pressing call on it now is ringing something that's been silent for 59 years. There's something oddly polite about the console still letting you try, instead of just greying it out. Did any of the kids ask why we keep calling the ones we already know won't answer?
Not yet! But that question is kind of the point. The console lets you try because the quiet ones still have something to say, just not over the radio. Each machine has a short story and a few questions kids can tap to learn why it went quiet. Surveyor 3’s is one of my favorites: two Apollo 12 astronauts walked over in 1969 and took its camera home. And if they want to go deeper, there’s a study notebook with a quiz, flashcards, and a podcast. And yes, 59 years of silence and it still gets a polite try. 😊
Ha, the 'yet' badge is my favourite bit of judgment in the whole build. Making Curiosity and Perseverance callable would've been the easy wrong answer, and turning that restraint into a badge instead is exactly the kind of thing kids remember. The mirror gets all the attention, but that decision is the one I'd steal.
Thank you, Paul. That badge was the hardest call in the build. Curiosity and Perseverance are still working, so making them answer like the ghosts would have been the fun choice and the wrong one. “Yet” felt right. They aren’t ghosts. They’re just still at work.
What I find particularly impressive about Ghost Rovers is that it combines three things that are often treated separately in technology projects: human-centered design, AI-assisted development, and scientific accuracy.
This is much more than an interactive educational website. It is an example of how AI can help transform historical and scientific information into an experience that encourages curiosity, exploration, and learning.
There are several aspects of this project that deserve particular attention.
NASA has decades of fascinating information about its lunar and Martian missions. However, scientific archives are generally designed to preserve and organize information, not necessarily to make that information emotionally engaging for a child.
Ghost Rovers bridges that gap.
Turning abandoned spacecraft into machines that children can call is a remarkably effective educational concept. Instead of asking children to memorize mission dates, spacecraft names, and technical specifications, it invites them to interact with history.
The Mission Control interface is particularly clever because it transforms passive information consumption into an active experience.
A child is no longer simply reading about the Apollo missions. They are temporarily participating in the story of space exploration.
That is an important distinction in educational technology: the objective should not just be to deliver information, but to create an experience that makes people want to discover more.
I particularly appreciate the distinction between directing AI and simply asking AI to generate code.
The workflow described here demonstrates a structured human-in-the-loop development process:
Human defines the problem, audience, creative direction, and acceptance criteria.
AI proposes designs and implementation plans.
Human approves the proposed direction.
AI generates the implementation.
A separate review process checks functionality, accessibility, accuracy, and visual consistency.
Human makes the final decisions.
This is a much more mature approach than treating AI-generated output as automatically production-ready.
The distinction matters because AI coding tools can produce impressive results very quickly, but speed does not guarantee architectural consistency, maintainability, or correctness.
The experience with the oversized JavaScript and CSS files is particularly revealing. It demonstrates that an AI agent can follow its immediate instructions while missing a broader architectural requirement if that requirement is not properly communicated within its operating context.
In other words, AI agents are only as reliable as the instructions, constraints, and validation systems surrounding them.
This is an important lesson for anyone building with autonomous or semi-autonomous development tools.
One of my favorite details is the correction involving the Apollo 15 lunar rover.
Calling it the first car on the Moon would have been an understandable simplification for a younger audience, but it would also have been historically inaccurate because the Soviet Lunokhod 1 robotic rover reached the lunar surface earlier.
Changing the description to the first car astronauts drove on the Moon preserves the educational appeal without sacrificing accuracy.
This illustrates a fundamental principle of AI-assisted content creation:
Fluency is not evidence of truth.
AI can generate compelling explanations, simplify complicated scientific concepts, and create engaging narratives. But factual claims still require verification against authoritative sources.
The use of NASA's NSSDCA records and the publication of fact-check logs are therefore important aspects of the project, not merely supporting documentation.
I would encourage more AI-assisted educational projects to adopt this standard.
The distinction between abandoned Mars missions and currently operational rovers is especially well handled.
Curiosity and Perseverance are still active missions, whereas the six historical machines featured in the Mars experience are no longer communicating with Earth.
Rather than sacrificing scientific accuracy to preserve the narrative, the project uses the word "yet" to create a bridge between past exploration and future possibilities.
That is excellent educational storytelling.
It introduces children to a deeper concept: space exploration is not a collection of disconnected historical events. It is an ongoing process in which one generation's discoveries become the foundation for another generation's achievements.
The children interacting with this application today could eventually become the engineers, scientists, and mission controllers responsible for future exploration.
Two details deserve particular recognition: the accessibility testing and the decision to build the experience without login requirements or user tracking.
Educational technology frequently overlooks these considerations while focusing on visual effects and engagement metrics.
However, a product designed for children should be particularly careful about collecting personal information and creating unnecessary barriers to participation.
The reported zero violations in the axe-core accessibility scan are encouraging, although automated testing should always be complemented by testing with real users, assistive technologies, and different devices.
Likewise, a no-tracking architecture is a meaningful design decision, especially in an educational environment.
The best educational technology should make learning accessible without requiring children to surrender unnecessary personal information.
Perhaps the most significant implication extends beyond this particular hackathon.
AI-assisted development is changing the relationship between technical execution and creative problem-solving.
People who previously had ideas but lacked the programming skills, time, or resources to implement them can increasingly create functional digital experiences.
However, I would not interpret this as meaning that software engineering expertise has become less important.
In many ways, it has become more important.
When code generation becomes faster, the bottlenecks shift toward architecture, requirements engineering, testing, security, accessibility, integration, and quality assurance.
The ability to define a problem clearly, establish constraints, recognize flawed output, and validate the result becomes an increasingly valuable skill.
The most effective AI-assisted developers will not necessarily be those who generate the most code. They will be those who can direct, evaluate, and integrate AI-generated work responsibly.
One additional thought: this project could evolve into something much larger.
Imagine expanding Ghost Rovers into a broader interactive archive of human exploration:
Historic spacecraft that no longer communicate.
Space telescopes and scientific observatories.
Deep-space probes such as Voyager.
Historic robotic missions to other planets.
Interactive timelines connecting past discoveries to current missions.
Classroom challenges in which students investigate real scientific data and make their own discoveries.
There is enormous educational potential in transforming scientific archives into interactive experiences.
The underlying concept could also be applied beyond space exploration to archaeology, oceanography, climate science, and the history of engineering.
Ultimately, what makes Ghost Rovers compelling is not simply that it was built in one day using AI tools.
It is that the project demonstrates how technology can make scientific knowledge more accessible, emotionally engaging, and understandable without abandoning factual responsibility.
My biggest takeaway is this:
AI can accelerate the construction of an experience, but it takes human judgment to give that experience purpose, context, and credibility.
And perhaps the greatest success of Ghost Rovers will not be the number of machines children manage to call.
It will be the number of children who finish the experience and start asking a much bigger question:
"What else have we left out there, and what could we discover next?"
Wow, thank you, Shirley. I appreciate the feedback on this project, and the time you set aside to respond. “Fluency is not evidence of truth” really hit for me. That’s exactly why the fact checks mattered.
The part that stopped me was the accessibility point. Zero axe violations means the tests passed, not that a kid with a screen reader had a good time. The more feedback I get here, the more I see the gap between what my tests check and what a real person needs. My builds start as an idea I want to see come to life. The people using them are what make them real. Real kids testing Ghost Rovers is the next step.
I also agree that the skill doesn’t go away with AI, it moves. Less writing code, more deciding what “right” looks like and checking that the build got there.
And your biggest takeaway is the whole reason I built it. AI helped me build it fast, but the purpose and the care had to come from me. If even one kid finishes and asks “what else is out there?”, that’s the win. Voyager has definitely crossed my mind too. It’s the ultimate ghost that still answers.
Thank you again for this one.
The detail that only the Apollo 11 laser mirror still answers — and that it doesn't reply, it reflects — is such a lovely way to teach kids the difference between a signal and an answer. And fact-checking every date against NSSDCA shows real care for the audience. Wonderful build.
Thank you! “A signal vs. an answer” is a good way to look at it. Checking everything against NSSDCA was the part that mattered most to me, since that’s what makes it something a teacher can trust.
very interesting
Thanks, Leo!
Super interesting read! Never knew these nodes were still reachable.
Thank you! That surprised me too. Only one still answers, and it’s been answering for over 50 years.
The obsession with wrapping remote APIs into multi-layered agent abstractions is creating massive hidden latency and failure cascades in production.
If you profile real-world workloads, bare-metal local inference (via direct C++ / llama.cpp or vLLM bindings) consistently beats fragile cloud wrappers on both predictability and cost. Tooling should eliminate abstraction penalties, not compound them.
Why I Chose Software Engineering 💻
Choosing Software Engineering is one of the most important decisions I have made.
I chose this field because I love technology and I want to understand how websites, applications, and computer systems work.
As a freshman at Yaoundé International Business School (YIBS), I am excited to begin this journey. I know there will be challenges, but I am ready to learn, improve my skills, and gain experience.
My dream is to become a skilled software engineer and use technology to create useful solutions to real-world problems.
This is just the beginning of my journey, and I look forward to everything ahead!
SoftwareEngineering #YIBS