Instead of just retrieving from raw documents at query time, the LLM incrementally builds and maintains a persistent wiki — a structured, interlinked collection of markdown files that sits between you and the raw sources.
【启发】这句话从根本上重新定义了 LLM 与知识的关系：从「查询时召回」升级为「持续编译」。RAG 是每次临时拼凑，而 LLM Wiki 是把知识「编译」成可积累的中间层。对 AI 产品设计者的启发是：真正有价值的 AI 工具不是搜索引擎，而是「知识编译器」——每次交互都在为下次交互铺路，而不是从零开始。
two small flaws in the Dictionary’s compilation process
It is incredibly difficult to plan in advance what to collect for any zettelkasten, even when its scope is tightly defined, like it would have been for the Oxford English Dictionary.
TruffleRuby is ahead-of-time compiled to a standalone native executable
https://www.academia.edu/community/VDDN6V
Evina Stein has a nice list of digital manuscript databases
Alternative approach to consider: don't rebuild older posts. Transform them from Markdown into HTML in situ and then never worry about recompiling that post again. You could do this by either keeping both the Markdown source and the output document around, or by crafting the output in such a way that the compilation process is reversible—so you could delete the Markdown source after it's compiled and derive it from the output in the event that you ever wanted to recover it.
going back from the tangled document to the LP input format
if there is occasion for it, and a heading^ in anycase; to multiply cross-references and indices; tokeep a record, on a separate set of slips, of all thesources utilised, in order to avoid the danger ofhaving to work a second time through materialsaheady dealt with. The regular observance of thesemaxims goes a great way towards making scientifichistorical work easier and more solid.
But it will always be well to cultivate the mechanical habits of which pro- fessional compilers have learnt the value by experi- ence: to write at the head of evey slip its date,
Here again we see some broad common advice for zettels and note taking methods: - every slip or note should have a date - every slip should have a (topical) heading - indices - cross-references - lists of sources (bibliography)
Another thing I’ll do when I’m King of the World is declare that executable code will always be inseparable from it’s source code. Using a script language solves that, of course, because a script is it’s source. For compiled code a second-best solution is the executable always contain pointers to where the source originated, but since that source may disappear the better solution is for language compilers to simply tack the original source code (and related make files and resources) onto the end of the executable in a way that can be commonly extracted. The law will be known as the “All boots shall contain their own bootstraps” decree.
symmetric
traditional compilation foils "symmetric authoring"
readable
Readable how? I think the better approach over what we do now, where we run source code through what are essentially compilers for making websites and then treat the output similar to object files—i.e. opaque blobs that are unsuitable for anything other than either (a) executing/experiencing, or (b) publishing to others—would be to pursue non-destructive compilation. So after the Markdown is compiled (if there is any compilation step at all), you don't have to keep the original sources around. The tooling should be sufficiently advanced to work with the publishable form as input, too, and not demand only Markdown. In instances where the Markdown files are kept around because the spartan experience of opening a plain text file where the content is almost entirely unadorned by formatting concerns is the preferred way to get things done, the tooling should be able to derive the original Markdown (or a good enough rendition) from the output itself. HTML is rich enough to be able to encode the necessary structure for this on the first Markdown-to-HTML pass. Systems that implement this kind of thing could be said to support a sort of "reversible Markdown", making the publishable form a suitable source of truth—a treatment that is right now reserved only for the originals that are collected and kept off-site.
Make the writing and editing experience more like Word or Google Docs and less like LaTeX.
in an ideal LP system, the (or at least a) source format would just simply be valid files in the target language, with any LP-related markup or whatever in the comments. The reason is so that LP programs can get contributions from "mainstream" programmers. (It's ok if the LP users have an alternative format they can write in, as long as edits to the source file can be incorporated back.)
(NB: the compilation "object" format here would, much like triple scripts, be another form of human readable source.)
She frequently cites authors second-hand(“as quoted by”, “see,” etc.) rather than primary texts, and in some instancesthis practice results in the kinds of errors for which earlier compilations werecriticized. The most egregious of these occurs when Blair cites Ann Moss on
Guarino da Verona when making the unlikely claim that note taking begin in earnest with Francesco Sacchini in the seventeenth century rather than a hundred years earlier with Erasmus and Vives.
I almost feel like I've arrived as I noticed this error in the text myself.
Interesting that he calls her out for making a compilation error, something which is very meta with respect to this particular text.
Judging from the copies now extant, the number of compilations, especially florilegia and encyclopedic compendia, continued to grow as more writers engaged in selecting and summarizing for their own use and that of others.16
There is a parallel between these practices and the same sort of practices seen in social media posting, annotating, and bookmarking, however in the digital realm the user interface is so simple that one needn't put very much thought into the process and the results become almost instantaneously meaningless. Was this the case in the medieval context as well, or did the readers/compilers get more out of their practices?
Rouse and Rouse (1982), 165–68 for the medieval titles, based on the flower metaphor or others, including liber scintillarum (book of sparks) or pha-retra (quiver).
In addition to florilegium, the descriptors liber scintillarum (book of sparks) and pharetra (quiver) in addition to other flower metaphors were also used in the medieval period to describe the genre of books in which the best passages from authoritative sources were compiled.
A number of ancient compilations, like those of Pliny, Diogenes Laertius, and Stobaeus, were indeed valued as both sources and models in the Renaissance, and authors of miscellaneously arranged compila-tions invoked Aulus Gellius as the founder of that genre.
While there are ancient compilations by writers including Pliny, Diogenes Laertius, and Stobaeus, many authors in the Renaissance credited Aulus Gellius as the founder of the genre.
In Bound to Read, Jeffrey Todd Knight excavates this culture of compilation—of binding and mixing texts, authors, and genres into single volumes—and sheds light on a practice that not only was pervasive but also defined the period's very ways of writing and thinking.
all his javascript is unminified so you can see how he implemented the dynamic examples in his essays
As it should be. (This should be the default. The NodeJS/NPM's Webpack addiction needs to be curbed—or confronted with a reality check.)
The Latin noun ‘compendium’, and the phrase ‘via compendiaria’ were used assynonyms for the noun ‘methodus’ during the sixteenth and seventeenth centuries.23ByLinnaeus’s time, the word was used in Latin book titles to denote a compilation of collocatedtexts that had previously existed as separate works on their own, or which, if removed and distrib-uted separately, could be read without recourse to other parts of the book.
Reviews
Python-igraph 0.8.3. (2020, October 8). Igraph Support Forum. https://igraph.discourse.group/t/python-igraph-0-8-3/473
Panda, A., Gonawela, A., Acharyya, S., Mishra, D., Mohapatra, M., Chandrasekaran, R., & Pal, J. (2020). NivaDuck—A Scalable Pipeline to Build a Database of Political Twitter Handles for India and the United States. International Conference on Social Media and Society, 200–209. https://doi.org/10.1145/3400806.3400830
Uribe-Tirado, A., del Rio, G., Raiher, S., & Ochoa Gutiérrez, J. (2020). Open Science since Covid-19: Open Access + Open Data [Preprint]. SocArXiv. https://doi.org/10.31235/osf.io/a5nqw
VA: Sing meinen Song – Das Tauschkonzert Vol. 4
VÖ: 9.6.2017
Charts seit 16.6.2017