uncoverLAB

@aliszu LFG!

Branding for Gatekeep - an AI tool that turns math questions into personalized video explanations . Target: kids, teenagers, students. The challenge: math is scary. The brand can't be. The Problem Is Real → 39% of students get "very nervous" doing math problems (PISA 2022). → Math anxiety peaks at ages 14-16 - exactly when concepts get harder. → Students with high math anxiety take 7.86 fewer STEM courses over four years. Not because they can't do it. Because they avoid it. What Anxiety Does to Learning Anxiety hijacks working memory, limiting problem-solving. High-anxiety students rely on basic counting instead of math thinking. A bad experience can lead to avoidance that affects career choices. So here's come the design Logo A "G" that's also a smile. Friendly face = perceived safety. Research shows round, face-like shapes reduce formality and build trust instantly. Colors: Pink, orange, yellow, green, purple. Not random. Each one does something: 💛 Yellow - attention, memory, optimism 🧡 Orange - critical thinking, reduces anxiety, anti-fatigue 🩷 Pink - calming, lowers heart rate, reduces stress 💚 Green - easiest on eyes, concentration, mental clarity 💜 Purple - creativity, wisdom, still calm Warm, saturated colors + round shapes = positive emotions + better retention. That's not opinion → it's from Plass et al. (2014) research on emotional design in learning. Why It Works Duolingo proved this model. 90% completion rates for gamified content vs 25% for standard e-learning. Gamified approaches show effect sizes of 0.46-0.82 in meta-analyses. That's substantial. The psychology Playful design satisfies autonomy (you control the pace), competence (visible progress), and relatedness (you're not alone). The Point Traditional educational branding is serious. Formal. High-stakes feeling. That's exactly what triggers anxiety in the first place. Gatekeep's brand says: this is different, but it's is safe. Math can actually be fun.

@aliszu 🐝

deeper research and complete 100% correct answers higher quality references / examples every ai starts failing on slightly out of distribution tasks, there needs to be market for more efficiency more bespoke ui patterns bespoke engineering patterns it should know the high quality code twitter users output eg shaders.paper.design/liquid-metal if i want to make a self driving car it should look into codebases written by @realGeorgeHotz and meet me at my current skill level, keeps me going until i get a self driving car at my parking lot if i want to train my own transformer it should be able to read the research papers and get the info there this is a simple repository of high quality resources that the model can choose from, complete with full stack code snippets and examples if i want good modern design it taps into @SprrrintDesign @uncoverLAB_app @Ibelick @motiondotdev components if i want good ai integrations it taps into @OpenAI @AnthropicAI @togethercompute @pinecone full stack examples / cookbooks (not just docs) if i want to do fronteir science it better research fronteir algorithms and compare and contrast even if it doesnt write code and it just provides me relevant things to read thats equivalent of a human coworker latency doesnt matter because humans can take hours to do the same work ai can real millions of lines of code for you if parallelized in seconds or minutes therefore latency doesnt matter what matters is quality if theres an update on a source it should know it if tailwind v5 drops it should know how to update your code repo currently if i ask for an ai sdk integration it outputs v3 code patterns my solution for that would be a constantly updating tap of information the model can crawl through