Author here. Some context that didn't fit the post.
Why I built it: English won the "world language" position by historical accident, and I kept noticing that almost none of its difficulty does communicative work — it's residue (spelling frozen by the printing press mid-vowel-shift, verbs that fossilized before the -ed rule generalized, articles that even advanced learners never master). I wanted to answer one narrow question rigorously: how simple could English be while still being English?
How it works: I started from the research (what learners actually struggle with, with sources), turned each pain point into a design target, and held every rule to the same acceptance bar — it must be statable without a hidden word list, and the spec's own examples must pass a mechanical sweep. That last part exists because my first design pass produced contradictions I only caught after building a linter that scans the World-English example text for abolished forms (an irregular verb, a "was", a "children") that leaked back in. Rules are not stable until they survive their own tooling.
Tech: static Astro site, tools in TypeScript/Bun. The translator is deliberately deterministic — it applies only the transforms it can prove from surface forms (verb regularization, plurals, banned words) and flags everything that needs part-of-speech or syntax (article deletion, tense changes) instead of guessing. There's also a reverse translator (WoE → standard English), a pronunciation renderer (respelling + IPA + audio via espeak-ng), and the whole thing ships as a Claude Agent Skill if you want an LLM to do the syntax-aware translation the deterministic tool won't.
The obvious limitation, stated plainly: nobody is going to adopt this, and it's not trying to be adopted. A century of prior art — Basic English, the spelling reform boards, Esperanto — is a graveyard, and there's a research page documenting exactly what each one tried and why it died. This is a design study. The output I care about is the map of where English's difficulty lives.
The most useful comment you can leave: pick one specific rule and tell me where it breaks. Every rule's spec records the problem it solves, the trade-off it accepts, and the alternatives it rejected — so "this whole idea is dumb" is less fun for both of us than "rule G2 collapses under anaphora, here's a sentence it can't handle."
Can't focus on reading anymore? ThinkTotem transforms books into conversations so you stop re-reading the same paragraph 5 times.
Phones and social media rewired our brains for quick hits, making traditional reading feel impossible. ThinkTotem solves this by turning books into active conversations with questions, examples, and 'explain-it-back' prompts. Instead of fighting your modern attention span, we work with it.
Does anyone want to help me figure out how to successfully do outreach?
I mean yes. It can be rewired back. But I think twitter (specifically for me) is required for my industry. I try to limit it to 30 min a day, but still I cannot cut it completely.
There are concentration exercises that can be done, but they are demanding...
Yes, unless it's diagnosed adhd, it's a skill that can be improved.
There are serious exercises to train the attention, which help with reading attention. but they're not fun...
I used to read half an hour before going to sleep and an hour on the commute. Now in the eve I have zero will and I feel I get carsick on the commute...
I even struggle with articles...
This conversations seems to be the only thing that works for me lately.
To guarantee that it does not illuminate and stays on track, I pass the relevant chapters in the context window, I pass the topics to cover, the ones already covered and the latest messages exchanged.
So highly relevant content as input and a refined prompt do the trick.
But context is only additive. You can give it the book content, but I don't think there's anything you can do to prevent it from bringing in unrelated info, unless you have a separate non-AI layer (e.g. connected to the RAG) that gives the user the particular snippets from the book, but even then you can't control what the AI bring to its interpretation of these snippets.
One thing that I think could be nice is if you could have it always keep the book in view, and just scroll to the appropriate place it is referring to, and then somehow highlight the specific snippet it is currently referring to, so that I could visually verify that it's giving me "grounded" information. Does your system do something like that?
It currently does not do that. It's a different way to "experience" a book.
There's an argument where, like you proposed, book and tutor are on at the same time. I will keep it in mind
The usual approach is to sum the present values of the expectations of any future cashflows.
For most cryptocurrencies, most people come unstuck at that point and just punt.
The only analytical approach I could foresee being useful would be if you could quantify the quality of the marketing of the new asset. My expectation is that that likely drives the bulk of the pricing most of the time.
This is very true. Marketing drives the perceived value a lot of the time. But this is a short/medium term trick. In the long term just the ones valuable will survive.
Do you look at specific data points to understand if the asset is overvalued or a scam?
> Do you look at specific data points to understand if the asset is overvalued or a scam?
Well yes. What you do is calculate the expectation of the future cashflows and PV them.
If there aren't any future cashflows then the fair price is zero and anything else is overvalued. That doesn't mean you shouldn't buy it but you should treat it like any other form of gambling.
If the asset creator won't release any quantifiable information about the future cashflows then the correct response is to assume it's a scam until they do.
That leaves the cases where there are quantifiable cashflows. Some smart contracts fall in this category. What you do then is analyse said cashflows and decide whether the price is justified.
Why I built it: English won the "world language" position by historical accident, and I kept noticing that almost none of its difficulty does communicative work — it's residue (spelling frozen by the printing press mid-vowel-shift, verbs that fossilized before the -ed rule generalized, articles that even advanced learners never master). I wanted to answer one narrow question rigorously: how simple could English be while still being English?
How it works: I started from the research (what learners actually struggle with, with sources), turned each pain point into a design target, and held every rule to the same acceptance bar — it must be statable without a hidden word list, and the spec's own examples must pass a mechanical sweep. That last part exists because my first design pass produced contradictions I only caught after building a linter that scans the World-English example text for abolished forms (an irregular verb, a "was", a "children") that leaked back in. Rules are not stable until they survive their own tooling.
Tech: static Astro site, tools in TypeScript/Bun. The translator is deliberately deterministic — it applies only the transforms it can prove from surface forms (verb regularization, plurals, banned words) and flags everything that needs part-of-speech or syntax (article deletion, tense changes) instead of guessing. There's also a reverse translator (WoE → standard English), a pronunciation renderer (respelling + IPA + audio via espeak-ng), and the whole thing ships as a Claude Agent Skill if you want an LLM to do the syntax-aware translation the deterministic tool won't.
The obvious limitation, stated plainly: nobody is going to adopt this, and it's not trying to be adopted. A century of prior art — Basic English, the spelling reform boards, Esperanto — is a graveyard, and there's a research page documenting exactly what each one tried and why it died. This is a design study. The output I care about is the map of where English's difficulty lives.
The most useful comment you can leave: pick one specific rule and tell me where it breaks. Every rule's spec records the problem it solves, the trade-off it accepts, and the alternatives it rejected — so "this whole idea is dumb" is less fun for both of us than "rule G2 collapses under anaphora, here's a sentence it can't handle."
Rules on one page: https://worldenglish.tugulab.org/rules/learner-grammar/ Translator: https://worldenglish.tugulab.org/translate Prior art / research: https://worldenglish.tugulab.org/research/ Code (MIT) + specs (CC BY): https://github.com/clacladev/world-english