Прибить список тем на первой странице к заметкам, которые есть - #5
Merged
Merged
Conversation
The README says what the repository is about:
Topics run across reinforcement learning, deep learning, dopamine and
reward prediction, neuroscience adjacent to learning theory, and
meta-research.
Every note already declares its own topics in front matter, and CI already
refuses a note whose front matter is wrong. Nothing compared the two, so the
sentence and the notes could only ever agree by somebody remembering.
Two directions, and only one is a failure.
A topic a note declares and the front page does not mention is drift: the note
landed, the page did not follow, and a reader is told the repository is about
less than it is. That fails, naming the topic.
A topic the page claims and no note covers yet is not drift. It is a repository
saying where it is going, which the same sentence says outright -- "the list
grows by what gets read, not by a plan fixed in advance". Reported after a
passing run, never failed: turning an intention into a red build teaches people
to delete the check, not to write the note. Today that reports `deep learning`
and `meta-research`.
The sentence going missing is a failure too, because rewriting it is how this
check would quietly stop checking.
Parsing it is an English list, so the commas separate and a leading "and"
belongs to the last item. Splitting on the word `and` as well looked tidier and
was wrong twice: it cuts "dopamine and reward prediction" into two half-topics
no note could cover, and without a word boundary it cuts inside any word
carrying those letters. Matching is on words, not phrases, so `dopamine` is
covered by "dopamine and reward prediction" and `reinforcement-learning` by
"reinforcement learning" -- the prose stays prose.
Tried three ways against a copy: a note declaring `causal-inference` that the
page does not mention (fails, naming it), the sentence reworded away (fails),
and the page updated to match a new topic (passes, and the remaining gap drops
to `deep learning` alone).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
README говорит, о чём репозиторий:
Каждая заметка уже объявляет свои темы во front matter, и CI уже отказывается принимать заметку с неверным front matter. Сравнивать их было некому — фраза и заметки могли сходиться только потому, что кто-то помнит.
Два направления, падает только одно
Тема, которую заметка объявляет, а страница не упоминает, — это дрейф. Заметка легла, страница не догнала, читателю говорят, что репозиторий уже о меньшем, чем он есть. Это падает, называя тему.
Тема, которую страница заявляет, а заметок ещё нет, — не дрейф. Это репозиторий, говорящий, куда он идёт, — и та же фраза говорит об этом прямо: «the list grows by what gets read, not by a plan fixed in advance». Сообщается после успешного прогона, никогда не роняет:
Превратить намерение в красную сборку — значит научить людей удалять проверку, а не писать заметку.
Пропажа самой фразы — тоже падение, потому что переписать её и есть способ, которым такая проверка молча перестаёт проверять.
Разбор фразы
Это английский список, так что разделители — запятые, а ведущее
andпринадлежит последнему пункту.Делить ещё и по слову
andвыглядело аккуратнее и было неверно дважды: это режет «dopamine and reward prediction» на две полутемы, которых не покроет ни одна заметка, а без границы слова режет внутри любого слова с этими буквами. Я прошёл через оба варианта, прежде чем понял.Сопоставление по словам, а не по фразе:
dopamineпокрывается «dopamine and reward prediction»,reinforcement-learning— «reinforcement learning». Проза остаётся прозой.Проверено
causal-inference, которого нет на страницеthe sentence listing the topics is gonedeep learning