curo

big files, i'll shrink, thank you!

Made a little "lofi"/classical reading room for my wife to binge Jane Austen before the Sense & Sensibility and Pride & Prejudice releases.

The rooms are based off the original text. 18 rooms from 6 novels. Music for each room, many mentioned in the book.

This page embeds the exact payload of the original including campaignVersion: "v5", candidateSetVersion: "mendelsohn-2025", all eight private excerpts, contexts, profile copy, and internal IDs. And it doesn't cite the original. I.e., infringement.

There's a section in the report "what's the most accurate translation" - tldr it's complicated. Presumably the top 8 translations by popularity pass some threshold of credibility.

Nice lesson on animation performance, i learned a lot by reading this.

As someone trying to build animated, edutainment-rich courses I'm a big admirer of Josh's work. Another banger

An interactive Greek & Roman Mythology course: https://www.scrivium.com

If we can nail this one, then an entire Oxford-grade education in the same style.

This is really fun — have you played with making the tile position opinionated (not agnostic)?

i wonder if have the clues point to a starting square (e.g., "E5") would be better than the current "reveal" aid. The spatial information would become more helpful toward the end when the player is dealing with the words they need help on.

It's at the end of the interview. He wants to build Starfleet Academy for technical fields. Physical with a digital equivalent. Thinks education will become like a gym (self-educate to look sexy) by the time AGI gets here.

We're using LLMs to cram a complete, 2500-year education into a Duolingo-like experience: https://www.scrivium.com

Previously, you could only build Duolingo-style apps for fields with concrete answers and progressions (e.g., Brilliant, Peak, Duolingo...).

LLMs unlock the ability to turn unstructured text into structured text. That means it's now much more cost effective to turn biographies, poetry, history, etc into individual game-like lessons.

We're running a Greek & Roman Mythology class — all interactive and free. It's great for creators and self-educators. I'd love any feedback: https://www.scrivium.com

Building a Duolingo-like app to "learn from the greats, daily." (https://www.scrivium.com)

Until recently, it was cost-prohibitive to gamify topics with indefinite answers and progressions. As a result, "left-brained" topics have been gamified for years, but "right-brained" topics have resisted gamification. LLMs and generative AI unlock game economics for unstructured text.

Casual gaming is only category outpacing passive, social media. The most direct way to elevate humanity's media appetite is to turn human greatness into casual games.

Real GDP per capita growth for US vs UK was almost identical until 2008. The last 3 years have been terrible for the UK, but if you're looking for the start of UK's stagnation you have to go much farther back than Brexit.

The fact that this thread is full of people who envisioned a physical recall is evidence enough the word is misleading. If reuters was an industry journal, sure.

Or it's just a misleading headline

Not scripture. Passages from the "Great Books" (literature, philosophy, poetry...)

We tried to partner recently to co-create interactive books with a well-known classical education influencer on X.

His job was to select passages from books and provide some commentary. Ours was to turn that material into an interactive title.

We tried tirelessly to look up the passages he'd send to us in the original text. The only quotes that matched were the top 5% short and famous quotes. The rest was made up completely, presumably by AI.

His 1 million+ followers consider him a world-class subject matter expert. But he doesn't read any of the books he's teaching. Eye opening.

- Location: US (willing to move anywhere, currently eyes on NYC) - Willing to Relocate: Yes - Technologies: Next, Typescript, Postgres, React, React Native (and Expo), Tamagui, Tailwind, AWS, Vercel, Python, PHP, ML/LLM model pipelines - Resume/CV: https://www.linkedin.com/in/cureau/ - Email: [email protected]

Focus is on edtech. Life mission is to make the best character education. I've been a CTO, VP of E, and an IC. I've built million dollar companies from scratch. I'm currently building "Crone: Feel Smarter" which you can download on the app store.

Best fit for those in edtech that need interim senior leaders or heavy hitter ICs.

Would "pragmatic-universal" and "pragmatic-consensus" count as a fifth and sixth? I.e., "what's useful for the group?" and "what the group thinks is useful for them?"

An example of the first category might be a parent deciding it's better for their kids to believe in Santa Claus. An example of the second is to create a society on some maxim that's rendered true by group-consensus, "everyone should have an equal say."

I'm curious how education or entertainment companies fit into this framework, or how they think about PMF in general.

Was Duolingo targeting customers who accepted the "hard fact" that passive audio was the only way to learn a language? Was MasterClass a "future vision" because people didn't believe celebrities would spend their time teaching? Or is it that we NEED education and entertainment, so these two providers just differentiated from a crowded market.

I like this idea, although preference-tuning for politeness might negate this effect

I'm bootstrapping https://www.crone.ai and making a small amount of money.

I've bootstrapped a few companies. B2B apps are easier since you can do custom work for business clients while you perfect your software.

Crone AI is much harder. It's my first edtech consumer app. It's been live for 3 months and we've made $1,500 so far (lol).

My previous business startups made $200k ARR and $1M ARR. Iterating on consumer apps is more fun and you can still extrapolate growth from small numbers.

are there others? (besides Project Gutenberg)

I've had a lot of fun working with both:

1. Lexical

2. ProseMirror/Tiptap

Both of these are very extensible, actively developed, and well-built. Tiptap is indy and pushes its pro product now (fair). Lexical is supported by meta. Pros and cons to both of these FOSS types.

One alternative. If you live in VS Code, consider Foam which gets you Roam/Obsidian like flows alongside an LLM (Github Copilot).

I have no affiliation to Foam. It's just awesome.

Ramana Maharshi says anything transient is false

CBT and other psychotherapies challenge X or Y as false

It seems they're using the term to distinguish "meditation for emotional regulation" (v1 research) from "meditation for itself" (v2 research)

> the pursuit of "states" can be a trap in of itself

Fully agree although a lot of devotees of Ramana Maharshi said they fell into silence just sitting in his presence. Improbable but it would be hilarious if scientists could make little "mouna wifi hubs" where practitioners got a silence handicap by sitting around it.

In the end though, I'm also skeptical that anything about self-inquiry can be replicated by an additive approach.

Fyi, CNBC (a news site) is citing a figure in a white paper by Envoy, an employee analytics platform, that leaders regret not having more employee analytics.

> "73% workplace leaders believe that easier access to data would enable them to drive smarter decisions about their space, programs, and policies... "

> "80% of executives say they would have approached their company’s return-to-office strategy differently if they had access to workplace data to inform their decision-making.

Here's the source:

https://envoy.com/wp-content/uploads/2023/08/Workplace_Data_...

OpenAI open sourced their evals framework. You can use it to evaluate different models but also your entire prompt chain setup. https://github.com/openai/evals

They also have a registry of evals built in.