Hidden Technical Debt in Machine Learning Systems
https://papers.nips.cc/paper/5656-hidden-technical-debt-in-machine-learning-systems.pdf

Very good one.
http://johnsalvatier.org/blog/2017/reality-has-a-surprising-amount-of-detail

Goods: Organizing Google’s Datasets '16
TLDR: Internal search engine operating over 26*10^9 *datasets* (e.g. DB tables), 5% daily churn, schema inference... interesting stuff
https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/45390.pdf

This common and unfortunate fact of the lack of an adequate presentation of basic ideas and motivations of almost any mathematical theory is, probably, due to the binary nature of mathematical perception: either you have no inkling of an idea or, once you have understood it, this very idea appears so embarrassingly obvious that you feel reluctant to say it aloud; moreover, once your mind switches from the state of darkness to the light, all memory of the dark state is erased and it becomes impossible to conceive the existence of another mind for which the idea appears nonobvious.
— Gromov

Source: M. Berger, Encounter with a geometer, Part II, Notices Amer. Math. Soc. 47 (2000), no. 3, 326--340.
via @atemerev
http://www.ams.org/notices/200003/fea-berger.pdf

Google support at its best
(TLDR: Firebase costs increased 70x because they started including SSL handshake in traffic calculation)
https://news.ycombinator.com/item?id=14356409

On the Turing Completeness of MS PowerPoint 😂
https://www.andrew.cmu.edu/user/twildenh/PowerPointTM/Paper.pdf

Neurosurgeon: Collaborative Intelligence
Between the Cloud and Mobile Edge
TLDR: It's best to split processing between device and cloud, here's how to do it automatically (for DNNs)
http://web.eecs.umich.edu/~jahausw/publications/kang2017neurosurgeon.pdf

DeepXplore: Automated Whitebox Testing of Deep Learning Systems '17
TLDR: Neuron coverage, use 3+ cross-referencing oracles to detect outliers
https://arxiv.org/pdf/1705.06640.pdf

Bash Strict Mode
http://redsymbol.net/articles/unofficial-bash-strict-mode/

> Design your construct, pick your deliverable - dsDNA, plasmid DNA or purified protein - and Serotiny will instantly price your order from various vendors.
https://serotiny.bio/notes/pinecone/
We live in interesting times :)

> Big data systems may scale well, but this can often be just because they introduce a lot of overhead
I believe I already posted this one, but somehow I can't find it in the history, so here it comes again
http://www.frankmcsherry.org/graph/scalability/cost/2015/01/15/COST.html

CNTK 2.0. Nice production-related features, note Model Compression, Keras support and impressive multi-GPU scaling.

> If your project involves sequence processing, such as speech, natural language understanding, machine translation, etc., CNTK should be your best choice. And if you are a vision researcher working on video processing, you should definitely give CNTK a try.

https://www.microsoft.com/en-us/cognitive-toolkit/blog/2017/06/microsofts-high-performance-open-source-deep-learning-toolkit-now-generally-available/

and now for something completely different
http://4dtoys.com/

Unicode nuances you probably don't need to know about :)
http://www.unicode.org/reports/tr29/#Grapheme_Cluster_Boundaries

Chronology of Math
http://www-history.mcs.st-andrews.ac.uk/Chronology/full.html

Volume Sliders by Reddit
https://www.reddit.com/r/ProgrammerHumor/search?q=volume&restrict_sr=on

Exploring LSTMs - with pretty pictures
http://blog.echen.me/2017/05/30/exploring-lstms/

the ever-important stuff
http://opentrannoscripts.org/trannoscript/coming-war-general-computation/

not only nice visualizations, but an entertaining stories too!
http://quillette.com/2017/05/26/paradoxes-probability-statistical-strangeness/

Even considering how much I generally dislike Linux — I'm very excited about userfaultfd and possibilities it provides to developers
https://lwn.net/Articles/718198/