Most of my focus right now is on a multi-year solo project. While I don’t intend to change this, I very much miss the feeling of working alongside motivated collaborators. I also have a long list of project ideas that are waiting for collaborator/problem fit.
I’ve decided to publish some of these, as conversation starters if nothing else. If you feel like we’d work well together and these general problem spaces excite you,Overall, I’m most excited to develop eclectic educational resources in computing, math and science, for all ages. I feel that it should be fairly straightforward to move beyond traditional expository material towards more enjoyable, engaging, interactive styles of learning. I’m also interested in bringing more historical context to technical topics. just reach out and we can jam.
LLM-assisted interactive tutorials
Most of the “LLMs in education” explorations I’ve seen have been too ambitious, trying to jump straight to a Young Lady’s Illustrated Primer style universal tutor interface, through prompting. This is not within current capabilities: a good teacher can diagnose misunderstandings and ask just the right probing questions to nudge the learner to a depth of knowledge far different to that achieved by blurting out an explanation.
The main idea I’d like to try in this space is an interactive tutorial with interleaved explanations and challenges, where the feedback on each challenge is determined by an LLM. This would allow for thoughtful instructional design beyond the current capabilities of LLMs, but a dynamic, adaptive response. A simple way to start playing with this would be to prompt an LLM to categorize responses into one of a set of distinct categories, then provide a follow up prompt constraining the style of feedback that ought to be given to each.
Current status: tinkering with something for a Hack Club project. At this point I’d be interested in conversations, pointers to prior art, maybe an enthusiastic collaborator to help with prototypes.
LLM tutor roulette
This idea came up in conversation with Brandon Hendrickson. LLMs may be better at search engines than providing an explanation of a concept, but the result is often a kind of insipid explanatory prose you’d expect from a bottom quartile textbook on the topic. This is a shame, because it wastes the versatility of LLMs: why not try asking for a limerick about a topic, like Brandon does? Or a passage with random cloze deletions, or for it to assess your brief answers to quiz questions it generates for you? Perhaps there are 100 different prompts with dramatically different interaction modes, and you may not know which one is going to be most valuable for any given topic. This could be fun to do as a website that randomly picks a prompt for the topic of your choice.
Overall I’m not as interested in this as I am in using LLMs for more flexible feedback to challenges (above), but it feels like it would be worth building out.
Current status: I’d be interested to hear from someone who would lead implementation; I can give guidance and maybe rope Brandon into writing a bunch of prompts.
De motu cordis: the video game
Harvey’s 1628 De motu cordis is surely one of the most important books ever written. Through simple experiments and calculations, Harvey convincingly overturned 1.5 millennia of misguided Galenist thinking about how the body works. Any work to make De motu cordis more accessible would be valuable, in my opinion, but most of all I feel the anatomical experiments are just begging to be made into a video game. I imagine retaining the art style of the original, and doing some fun storytelling around defeating the ghost of Galen. Or something. Think Silicon Zeroes for anatomy.
Gilbert’s De magnete for young mythbusters
Another seminal 17th century work that deserves re-enactment is De Magnete by Gilbert. I’m most interested in his experimental debunking of misconceptions around magnetism, as well as his simple exploration of static electricity with. I imagine a booklet that can be downloaded for free, or ordered in a little package with a “lodestone” and “amber”. We would have to strike a balance between having a fun sequence of challenges, and exposing the original context in an interesting, historically faithful manner.
Revitalizing Hooke’s Micrographia
While on the topic of 17th century science, I feel that Hooke’s stunning 1665 Micrographia ought to be better known and more accessible, particularly to kids. I’m not sure of the best way to do it, but the main goal would be to recreate the wonder of seeing the world at this scale for the first time.
Perhaps a fun way to do this would be a microsite with some of Hooke’s best illustrations, alongside comparable screenshots that kids take with their cheap USB microscopes. Wouldn’t it be cool if they found their own fleas and compared them to Hooke’s? Or something. Give me your ideas.
Optics through observation, obviously!
There are so many fun, simple experiments in the history of optics, that it’s frankly ridiculous that I was taught the topic via lecture and textbook. Even as early as 1011 AD, Alhazen explained in Kitab al-Manazir why he believed that light travels in a straight line from objects to the eye, using camera obscura experiments. Most of these demonstrations could be recreated quite simply, with some guidance.Sadly one observation by Almazen which would be harder to replicate is that if one carefully scrapes away the back of a bull’s eyeball, one would see a flipped image of what was in front of the eye, demonstrating that our eyes works like a camera obscura.
Then, consider that before Newton, it was generally believed that light traveling through a prism was somehow adulterated to form the emanating colors. Newton demonstrated that light is in fact composed of the individual colors observed, simply by using a second prism to recombine the components into a single shaft of white light.
We could come up with a sequence of simple experiments like this, annotated with historical context, which like the De magnete idea above could be downloaded for free or purchased in a little kit with a few prisms and lenses.
Micro-games for computer science concepts
I would like to see more little games in the style of Deadlock Empire and Flippy Bit. The basic formula would be to take a CS concept that is typically explained through a lecture or two, and think how best to make it into a micro-game where victory in effect requires you to fundamentally understand the thing. Very few students retain much from being told about two’s complement, but challenge them to come up with a signed integer encoding and present them with a sequence of problems with it, and they may come to appreciate two’s complement through the process of reinventing it! I could generate dozens of topics like this.
Playable computer science papers
If there’s something you think Every Computer Scientist Should Know About Floating-Point Arithmetic, there may be a better way to actually achieve this than to simply write a paper by that name. These are great papers though; in my view they are begging to be turned into a sequence of playable challenges. There’s Goldberg on floating point, but then also Drepper on memory, Hoare’s Communicating Sequential Processes, and I’m sure many others.
Current status: I’d be interested to hear from people who have particular papers they really love, but who are not sure how best to present those pedagogically.
An index of motivating historical problems in mathematics
I feel that we could do a much better job of teaching mathematics through the lens of the original motivating problems. Mathematics often developed as a side effect of grappling with highly engaging problems; in the interest of expediency students are presented only with the final conclusion, thereby removing the fun and ruining the ending.
Long term I imagine writing a book or online reference indexed by topic, so that a teacher or student can look up the motivating context. This could start as a blog, and would mostly be an exercise in summarizing the existing literature in a style that’s fun and broadly accessible. It may also be interesting to make some of these playable, e.g. try to cross all the bridges of Konigsberg on a map, etc.
Current status: will probably need to wait to be a primary project for me at some point. If somebody is already undertaking this and looking for a collaborator, I may be able to make some contributions.
A history of breakthroughs in algorithms
Sometimes I look at an algorithm or a proof and wonder how on earth they were discovered, as if the authors ventured off into the woods confident that they’d find treasure, and lo and behold, an “impossible” result. Breakthroughs like Karatsuba integer multiplication, Strassen’s matrix multiplication (and integer multiplication, again!), quicksort… all of these are just astonishing achievements, seemingly the product of immense courage and ingenuity. Looking more closely could reveal some insight into how we can all approach challenging problems and achieve “impossible” things ourselves. If nothing else, it makes the study of algorithms much more interesting.
Current status: I’d be happy to co-author a longform piece on this, ideally with somebody excited to dig into archives and/or conduct interviews.
Teaching math with spreadsheets
There has been a lot of work around using computers to teach math, much of it quite good, but there is more that we could do and avenues that seem underexplored. One is to consider how we could help even young students learn math topics more interactively by having them write simple programs in the form of spreadsheets. I have not actually tried doing this, so the project would involve first teaching some kids this way, then publishing any discoveries.
Current status: Alec Resnick has mentioned that there’s a vein of literature on this already; I’m yet to check it out.
Stripe Press for children’s books
I’m impressed by how effectively Stripe Press has been able to revivify important out of print titles like Dream Machine and The Art of Doing Science and Engineering. There are a number of children’s books that have sadly become collectors items that I believe deserve the same treatment, such as some hard to find books by Mitsumasa Anno, and many of the Mir titles.
Current status: playing with some ideas with Charlie Harrington; interested in hearing from folk who might help us understand the publishing industry better.
Phylogenetic tree explorer
My kids are constantly asking how certain animals related to others. We can answer them by trawling Wikipedia, but these investigations don’t seem to give a holistic sense of the tree of life, nor do they do much to make it easier to answer similar questions in the future. It would be nice to have a kid friendly version of something like onezoom, perhaps with both an “explore” mode and challenges (find a particular species, find the closest common ancestor of two, etc). I haven’t thought much more about this one.
Spaced repetition for kids
I am not sure how excited I am about this one, but there may be something there: I think at least some kids would enjoy a cute, game-like spaced repetition app with the simplest possible UI, that is accommodating of typical failure modes of spaced repetition systems. I could imagine a super simple voice-based app where you could add words throughout the day and get fun questions about them later. With a little creativity, you could have the questions be interesting and dynamic in a way that would be impossible with a more general SRS app, e.g. you could ask to use the word in a sentence, what’s the opposite of that word, what’s the difference between that word and a related word, etc.
Current status: one friend is interested, but not yet actively working on this.
Nature treasure hunts
I’d like to teach botany through a sequence of increasingly challenging “nature treasure hunts”, E.g. you could have a “leaf” sequence where the first goal is to find as many different “kinds” of leaves as you can, but then each hunt introduces another aspect of morphology… maybe you’re looking for leaves with serrated leaf margins, or lobes, or different venation patterns etc. Then do a sequence with flower parts, and so on.
Understanding English spelling
In teaching my kids to read and write, I always cringe when I need to apologize for counter-intuitive orthography. On the one hand, I see English etymology and orthography as fascinating windows into history and culture; on the other, this too much for a young kid. So I would like a short handbook or microsite that covers some of the major gotchas of English orthography, designed for those helping their kids learn to read, that they themselves can read in a single sitting and perhaps refer back to as needed. I am aware of the Structured Word Inquiry community… I don’t want a whole methodology, just a quick guide.
VR bird museum
I’d like a VR (or AR) app where I can simply look at bird models, bird skeleton models, feathers etc in 3D, including arbitrarily zooming in, doing side by side comparisons etc. One way to think about this is bringing a museum’s bird collection into your home; another is allaboutbirds.org but in 3d. There are generic model viewers already of course, so this would be optimized for those who are specifically interested in birds. This would be overall a huge endeavor but any one piece of it would be interesting in itself.
Spacial audio bird song thing
I’d love to be able to pick a spot on Earth and be transported into a realistic soundscape with spacial audio. Well, we could skip the traffic sounds etc and make it just about birds: find the local birds via the eBird database, and play recordings from xenocanto. I’d personally like a game where the objective is also to ID the vocalizations, but others may find value simply in the relaxing vibes or sonic tourism.
Learning bird ID
I have thought a lot about learning to identify birds, both by eye and ear. I don’t accept the convention wisdom that it simply requires a lifetime of patient observation: I believe that we could develop a system for presenting birds side by side, pointing out key features, then quizzing specifically on those differences, rather than expecting species level identification from the beginning. An interesting technical aspect of this would be to find an appropriate sequencing for a given location and level of prior exposure. This could be combined with the spacial audio thing above in the case of ID by ear, or just a website/app for photo/video ID.
Ok that’s it for now! email me if any of this resonates with you.