@dwlz Ha! Yeah I loved Quizlet in school. But that's not real learning
Notes
- @astralwave 99.9% of the facts and skills I've learned was without spaced repetition
- @ben_m_somers I'm a logo kid! I'm papert obsessed! What's a good implementation of spaced repetition?
- @brian_lovin I wonder why LLMs are systemically so bad at estimates like that! Is it because that's how long it'd take a human?
- Spaced repetition gives me the ick It's like the Soylent of learning. It's a scientist's idealized form of learning, stripped of all the natural messiness that makes learning rich and beautiful Picture a mom using spaced repetition flashcards on her baby Now picture that mom speaking lovingly to her baby about whatever's on her mind as they go about their day Which world do you want to live in? Where do you think the baby is better off? The humane way to learn something is to be immersed in an environment where learning happens naturally, automatically, as a consequence of natural motivation and play Think about all your most positive learning experiences. Learning your native language. Learning to move your body at playgrounds as a child. Learning to play video games. Learning how to use a computer by messing about. All natural, without instructions, or idealized, measured doses of learning Natural learning is a beautiful human process. Spaced repetition loses all of that. We should flip our focus from learning random facts as fast as possible to crafting ENVIRONMENTS where skills and learning happen naturally Want to learn French? Go to France Want to learn math? Go to Mathland (Logo or Scratch) Want to learn programming? Go to Val Town ;)
- @raveeshbhalla @jeffreyhuber @ValDotTown nice!! why do you want it somewhere else? 😅 anything i can do to help? i could send you a pr to make it prettier! ie help with markdown to html formatting
- @nbbaier @JnBigio @geoffreylitt Linked at the bottom of this blog post https://blog.val.town/vibe-code
- @nerdymomocat @saradietschy @ValDotTown thanks! sara - if you'd like a personal tour of val town, shoot me a dm! would be fun to make your first val together
- @dsp_ Which one???
- @julesagent context: this is cool! would make it a good fit for val town, but mcp would be even better!
- @julesagent jules mcp client when?
- "as long as humans have any role in creative work, there will always be returns on deeper focus and attention." https://x.com/geoffreylitt/status/1991508961950904438
- Programming is not dead! There is so much confusion right now about the fate and future of programming. Let me clear some up some confusions: 1. Are our jobs at risk? 2. Do you have to use AI to keep up? 3. Is 'learning to code' bad advice? The answers to all three questions is a resounding NO. But I understand your confusion. AI coding is very confusing! There are three ways it's confusing you: 1. You're conflating cruise control with full-self-driving AI progress is notoriously hard to predict. The photo in this tweet is my cofounder Tom's "Over the self-driving rainbow", which shows when different car companies made confident predictions of FSD, and when those predictions came due with FSD still not deployed. Yes I know that it seems like AI coding progress is so fast and that it's just months away from putting us all out of jobs. But truck drivers still have jobs. Take a deep breath. We just don't know how long it'll be. It could be decades. And if your computer programming job is gone, you better believe every other job in society is gone too. (Because the AI programmers will automate those too!) So at least you won't be alone in figuring out what to do with yourself. 2. You're too focused on the agentic AI coding of today AI coding today is in the dial-up era. It's just so slow and expensive. Everything you do is optimized around cutting costs and waiting around for the LLM. This is why it's so frustrating and unfun. You're not in a flow state. You are waiting around. You don't know what it's doing. It will totally change once LLMs are faster and cheaper. It's the difference between dial up and broadband. The context-switching of AI coding today fools you into thinking you're more productive than you are. It's an illusion. Engineers who use NeoVim and no AI at all are just as, if not more, productive than the biggest AI engineer. They don't have to waste time figuring out how to use the new model or IDE that comes out every week. 3. You've forgotten that programming is theory building Vibe coding only works for stuff you intend to throw away. Otherwise: code you don't understand is legacy code, holding you back. The bottleneck in computer programming is what your human brain understands. This will always be true - as long as humans have opinions they want to express about software. The future of AI coding will be optimized around bringing the human brain up to speed as fast as possible, keeping the human brain in flow, automating all the drudgery, but no more Computer programming is a skill, like a physical skill, and you have to actually do it. You can't get better at surfing without getting wet. AI can't help you get better at surfing. You have to actually get on the board. Computer programming is a skill. You don't have to type every line of code to get better at it, but you gotta be deeply involved. This is why Cursor tab-tab-tab is good, it keeps you in the loop. This is why superfast models by Cursor and Windsurf are good. This is why Windsurf's focus on understanding - codemaps - are good. Anyways, I just wanted to say that you don't need to be sad or worried about your job, or that the craft of programming is dead, or that learning to code is useless. All of those things are exactly false. There has never been a better time to learn to code, or to be more excited about the positive future for the craft of coding. The future is bright and the possibilities are endless ☀️
- @jmduke Yeah perfect thanks
- https://www.val.town/x/stevekrouse/research-newsletter-bot
- PSA: Research agents are now just 10 lines of code https://x.com/stevekrouse/status/1991278146969362478/video/1
- @reelCheburashka this is awesome! thank you!
- @swyx - where can i find the best price/throughput/intelligence charts over time? i want to see the 'moores law' of llms. i want to build intuition about what kinds of improvements to expect
- oh boy https://x.com/stevekrouse/status/1991271088878244264/photo/1
- request for app: http://slugs.ai a list of all these slugs: gpt-5.1-2025-11-13 with good search, etc
- @DanielPrevoznik correct! this is exactly what i was going for!!
- VOICE MODE! IS THIS FAST ENOUGH FOR YOU???? https://x.com/holdenmatt/status/1990906902406894003 https://x.com/stevekrouse/status/1990920324322312343/video/1
- @albfresco @cerebras @relace_ai @morphllm nice!!!
- @albfresco @cerebras demo or it didn't happen! overwriting - would be fun to use @relace_ai or @morphllm to do the edits
- https://x.com/stevekrouse/status/1990913758521078224/photo/1
- ok but what if we generated a new app on every keystroke? hope my friends at @cerebras don't hate me https://x.com/stevekrouse/status/1990902186226684260/video/1
- @saimonsharif No need to change the prompt The use-case was an "data analyst agent" that could agentically query a database and then chart the output
- Just helped a customer switch from Sonnet 4 to Haiku 4.5 and the speedup was HUGE, with no obvious downgrade in intelligence 2x faster, 3x cheaper, same intelligence – crazy!
- @nfarina Thanks! Yeah we do but our templates are a bit stale. Happy to write you something custom How about lead scoring when folks fill out your form? I've been having a lot of fun with OpenAI research agents for that Chat on a call next week? I'm sure we can think of something…
Text may be partial · Read original
- @lastgoodhandle this is a very good joke
- @nfarina Thanks!!! If you want a reason to use Val Town, may I suggest: new user enrichments in Slack? It's really fun to get realtime notifications when users sign up, including where they work, their linkedin, etc, even if they sign up with a personal email Of any interest?
- @GanatraSoham Yes! This is an example of something that hit our MCP server before our API server because we can move faster with MCP https://x.com/stevekrouse/status/1989364664426487864/photo/1
- If you've read my tweets recently, you know I'm not MCP's biggest fan But there are a lot of upsides, especially for users: 1. Cheaper – Don't pay us for credits. Pay your inference provider directly 2. Better – Use whatever state-of-the-art LLM you want 3. Val Town everywhere – Get the best parts of Val Town – instant deployments, built-in SQLite, etc – in your favorite LLM coding tool https://blog.val.town/mcp
- 📢 Introducing The Val Town MCP Server Bring Val Town to your favorite LLM – Cursor, ChatGPT, Zed, wherever! For example, here we use Claude Code to make a blog in a couple prompts https://x.com/stevekrouse/status/1989329834078515583/video/1
- @bernhardsson TOMATOES! i'm obsessed with sugar bombs
- @colinhacks @charmaine_klee - you gonna fix this?
- @colinhacks it's a bad look
- THE TRUTH COMES OUT https://x.com/simonw/status/1989001665526264169 https://x.com/stevekrouse/status/1989002053176172964/photo/1
- @erikphoel I would pre-order this book so fast
- @WhoKnow50498590 @kandros5591 @ValDotTown It'd just be code. You can use or build a workflow engine on top of Val Town. We have sqlite built in so it's pretty easy to make computations more durable




