Justin Reidy (@VegetablesReidy) header
Justin Reidy (@VegetablesReidy) profile picture

CTO, No Diet Dietitian. Board, Stern Center for Language & Learning. Building Mathblast, math facts for my kids. Ex-Webflow, Loom, Airbnb. Aspiring mensch.

BTV
Joined July 2007
Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)Photo by Justin Reidy (@VegetablesReidy)
After a lot of requests - here it is. Some extensive thoughts on Alpha. Also with a section on why it matters even if you don't want it to matter. substack.nomoremarking.com/p…
15
36
13
163
33,700
Peculiar observation (n of many) from building an adaptive learning algorithm for math facts: +2 is slower than +3 or +4. My hypothesis is that brains do the minimum the task demands. +0 and +1 get memorized almost for free. +3 and +4 are slow enough to count that memorizing pays off. +2 sits in between. Counting it is fast enough that nothing ever forces the memory. If you want a fact to be instant, something has to *ask* for instant.
1
69
Intrinsic motivation is great and all, but I haven't met a kid who is intrinsically motivated by 7x8. Sometimes you just have to do things. Gamification is the usual fix, but it has its own problems. Game mechanics have their own rules, and they can fight learning science. The thing you always need to ask is, what is the game paying out for? Playing more, or remembering more?
8
8
885
I've been wrestling with this question a lot as I've been building a math facts app. Lean too hard into learning science and you kill motivation. Lean too hard into gamification and the facts don't stick. Coming up with an engine that solves for gaming and learning in a way that works for ALL learners is... really freaking hard
1
87
One of the key things we must do as educators is making "the push" doable and building a tolerance for effort. The truth is that we can't make every student love every subject. But we CAN make effort habitual. And we can build an environment that reinforces the work.
Left to ourselves, we avoid effort. But with some push, we can learn to want it: On acquiring A Taste For Effort in life & school: ↓
1
143
Fable 5.2 was enough to put me back on Ultra from Max. Well played, Claude, well played.
1
96
Periodic reminder that velocity is a vector. Going faster in the wrong direction is worse than doing nothing.
Software got a lot easier if you know what you wanna build, and a lot harder if you don’t. With clarity, you can straight-shot to the solution. Without it, you’ll ship a pile of junk and spend a lot of money doing it. Pre-AI era you would ship nothing, which is an improvement.
5
305
After this week's blue wave, the conventional wisdom will be that educational reform is over. But even as some changes will wither, a thousand flowers are about to bloom. Parents know what their kids are missing now. They won't accept it, especially when the alternative to an engaged kid is cognitive surrender to AI. With building easier than ever, the teachers (and parents!) who know what kids need can ship it themselves, straight to the families who are desperate for it. Change is coming. One way or another.
2
107
In endurance training there's this idea of "junk miles" – going further just to increase a number. It's work without an outcome. Learning, like training, needs to ask: What's the outcome? And is the stimulus right for that outcome? Otherwise it's just junk.
There's a million ways to do active learning wrong. The way to do it right is "deliberate practice": mindful repetition on performance tasks just beyond the edge of one's capabilities. Deliberate practice is about making performance-improving adjustments on every single repetition. Any individual adjustment is small and yields a small improvement in performance. But when you compound these small changes over a massive number of action-feedback-adjustment cycles, you end up with massive changes and massive gains in performance.
1
167
Love to see a learner as well-known as @patrickc sharing his observations here. Models are already at a place where they can perfectly manage curriculum delivery. The problem is that there's no ideal harness, and the UX is wildly jagged. But the future is *very* bright.
The models are very good at creating custom spaced repetition apps with algorithms adapted to you and the problem domain ("review my learning and improve the efficiency of your scheduler") plus tailored explanations for anything that you consistently get wrong.
2
180
I told my (almost) 10 year old a couple nights ago that learning to love learning will be THE most important thing in our new AI enabled world.
the most important enabling belief is that you can learn anything - a domain, a behavior, etc.
1
4
226
Working on spelling rules with my oldest, who kept expressing them as math equations ("ie=k" for start /k/). Now we're learning them as decision trees and I have no idea why this isn't more of a thing:
1
2
110
The realization that "you can just do things" is liberating once you internalize it.
2
120
"It's people, not programs, that are going to make a difference in the life of a kid." Or maybe better put: people supported by programs. In this case, the program was ~750 reading specialists whose whole job was helping teachers teach reading. The specialists work alongside teachers in their own schools, and by law, schools can't pull them into subbing or admin work. It's individualized attention for the teacher, the same thing we know works for kids. And it helped take Alabama from last to nearly average in reading. Individualized support is what makes the difference, for students and for teachers. Hiring 750 specialists is a huge lift, and kudos to Alabama for doing it. Even so, a specialist can't be in every classroom every day. AI can fill that gap, day to day, for every teacher.
1
1
195
> I'm a long-time EdTech skeptic. And I'm here to say that advances in AI have changed the game. 100%. Traditional EdTech was a disaster (and still is in most classrooms). And I know "this time it's different" is what people always say. But seriously: this time it's different
Replying to @johnarnold
I'm a long-time EdTech skeptic. And I'm here to say that advances in AI have changed the game. When I started at Alpha 10 months ago, I wasn't ready to believe this would be the case. I was absolutely an Alpha skeptic—mostly because I was an ardent EdTech skeptic. My Alpha journey has been eye-opening (and humbling). I just started writing about it yesterday, to capture the some lesser-known aspects of Alpha (rapidly-improving quality of the apps and the expertise of the growing Learning Science team). I also applaud @jliemandt for the entrepreneurial approach he has brought—and especially for understanding that a Learning Science-led approach could be revolutionary in the AI era. The skepticism is only natural, given the legacy of disappointment for tech and education. I appreciate everyone keeping an open mind where Alpha is concerned. Like you, I think it's more interesting to judge Alpha by its potential in 3 years than even its current state today. It's a loud theme in yesterday's Sstack. karenvaites.org/p/alpha-scho…
8
3
6
34
6,929
Justin Reidy retweeted
i'll say it. this is the most braindead direction in AI right now the tech is insane, yes. it sounds human, it laughs, it stops mid sentence, it drops the “uuhh” that's EXACTLY what's wrong with it we spent decades making computers fast and precise and now we're paying billions to make them hesitate like a nervous dude when i talk to a model i want answers AS FAST AS POSSIBLE, AS SHORT AS POSSIBLE. i don't want a fake friend that talks over me if you want a human experience go talk to a REAL human with REAL emotions. touch grass bro please and honestly if you're so lonely that you need a machine to fake pauses and laughs for you, that's not a product problem, that's a you problem this app is cooking people's brains and this is the proof
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa
197
273
33
2,684
121,797
Justin Reidy retweeted
I don't think you understand the gravity of this situation, here listen to this.
Holy fucking shit.... This is actually crazy, and you can run it locally!?? It's almost indistinguishable compared to a real person. Dessa is a speech-to-speech AI model, which speaks like an actual human. YES, AN ACTUAL HUMAN!!.
217
334
167
5,573
541,211
Justin Reidy retweeted
Raise your hand if you work in a 2010s SaaS company that has built their own agent platform that no one really wants to use.
Replying to @ajkemps
To be direct about it: every company that considers themselves a “tech” company is currently building a much worse version of something they could buy off the shelf, and furthermore they often have several teams competing on this front Then they launch it after 3-6 months of building in a silo and can’t get their colleagues to use it for entirely predictable reasons Coding agents and their adjacent products are among the most competitive markets of all time. The benefits of a highly competitive market accrue to customers by virtue of this intense competition among suppliers. Econ 101 Is the internal team working on your agent platform really keeping pace with the market? Even if they put in the same hours (they don’t), they don’t have the breadth of customers and hard contact with the market to get the feedback necessary to make the product any good Big corporates that call their devs “IT” don’t delude themselves in this way. They buy the best, recognize their relative advantage (vendors will build almost anything for you because the competitive pressures demand it), and get back to work
32
5
3
401
34,641
if steve jobs saw you reserving 15% screen width for a single icon and then putting a left side arrow back button on the unreachable right side of the screen he would've personally taken you behind a dumpster and beat you with a metal pipe
327
1,084
150
24,791
651,227
Justin Reidy retweeted
Dear god it's inventing German from first principles
GPT-6.1-Sol reasoning traces totally legible, nothing to worry about
161
996
50
17,255
459,809