Derek Murray

@garyfleming A talking dog is an amazing achievement, congrats.

My biggest achievement in February was teaching the dog to give me a high five, and yelling “Radical!” when he succeeds.

@THarris_Tim @EE Did one of your recent prime ministers change VAT to –33%?

@ch402 Too long indeed! Hope you’re doing well, Chris!

@mrry Thanks for sharing! (Also, so lovely to see you pop up in my feed! It's been a long time!)

It's surprisingly how little there is on "how to run a research group." Lots of great resources on being a researcher (eg. You and Your Research), and on leading other types of teams (eg. books on software engineering management). Research leadership seems like an open problem.

@stephenrkell Also I’d question whether “X is hard to use” implies “X is not influenced by an ICFP’97 paper.” 😃 Less facetiously I suspect these abstractions are favoured more by people who have to *read* UI code than those who have to write a lot of it from scratch.

@mrry Thanks Derek! Good call as always... the functional reactive abstractions always do seem much more right. I always assumed React.js was influenced by FRP, but now I look at it, nobody says this. They do say "React is hard to use"... maybe it didn't borrow those ideas after all.

I've always studiously avoided user interface programming, because it's 'obviously' very complex, and feels needlessly so: the abstractions never seemed right. Are there any good references examining this complexity? Are the abstractions any 'righter' in the web era?

@stephenrkell A bit like how one might call C# “influenced by Haskell” through features like LINQ, but probably wouldn’t consider the whole C# language a true descendant of Haskell.

@stephenrkell [Not my field at all, but…] I think there is some influence from FRP to React.js, but possibly more on the “functional” side than the full-featured “reactive” side, so it diverges from the ideals.

One of my favorite parts of working at Google was when @dave_andersen was “interning” at the next desk over. Congratulations Dave!

@lindsey @frankmcsherry in that bit forgetting that many things a computer can do are bad, and shouldn’t be encouraged.

and to be clear, I genuinely *love* this bit from the timely dataflow book, but I was still incapable of stopping myself from making this and then sharing it with you

Don’t think I can take much credit for those 221 stars… maybe some of the Issues though!

Is there a discount if your name is @dijkstracula?

@dijkstracula @cbarrett Like this? research.google/pubs/pub49008.…

Deep Learning based malloc implementation. Is this anything?

@dijkstracula Move over Phil Collins: pbs.twimg.com/media/E5h5EDlX…

@anaklimovic @jsimsa If only we could have put that example in the paper… 😃

When we started tf.data, I never expected it to be used to solve Advent of Code puzzles, but here we are! pgaleone.eu/tensorflow/202…

It's a very exciting day @Lacework. If you're interested in coming to work with us on the opportunity ahead, let me know! sprou.tt/16Vxks47F5W

@stephenrkell On (or off?) the deep end, self-adjusting computation bestows incrementality on more “interesting” types of computation: umut-acar.org/self-adjusting…

@stephenrkell On the practical end of the spectrum, there’s Vesta: grosskurth.ca/bib/2000/heydo… Although it’s “just” a build system, the caching extends to the functional language inside.

More generally, I wager that a huge % of compute work worldwide is trivially repeated unreasonably many times. What research exists on practical yet "deep" approaches to incremental computation? i.e. that could catch and obviate such issues.

How am I only just learning that @tearsforfears put out an EP of indie covers in 2014? Nostalgia really is fractal… discogs.com/release/560979…