So uh it's been about 6 years since I did anything "contemporary" in Machine Learning (a period correlated with how long I've been leading a team instead of making). That feels long enough for there to have been a new ML text book. Has one emerged? Any recs?

One of the biggest lessons I've learned lately is from the BizOps team at work: in their calls everyone always uses "we" instead of "you" or "they". No matter the situation, org structure, or the cast of tribes, teams, guilds and programs: "we". It's simple but *so* powerful.

You guys I might be in love with transparent huge pages. Or the lack of transparent huge pages. Something about huge pages is bad. Or good. Or possibly both. Their transparency is important though. Or maybe not. Today has been a confusing devops day.

My bit of London smells like it's on fire. Doesn't bode well.

New sound when pulling away at a bike-lane light: the whine of all the electric devices trying to keep up with us leg-powered beings. Scooters, e-bikes, and those weird briefcase-with-wheels things. Sometimes I grunt if I left my bike in too hard a gear, otherwise am silent.

Now my excitement of the DALL-E/GPT-3 stuff is normalising, I'm starting to find what people are typing in more interesting. These algorithms are for exploring our own brains!

Norway's like being at home in the (North of the) UK excpet everything's just a bit better. Transport's a bit better, houses are a bit better, services are a bit better. Improving things just a bit seems to have a huge impact on quality of life.

So there’s a world of generative/programmatic art things that I’m loving, and a world of data viz that I have always loved, but the two seem to remain stubbornly distinct. Am I missing something/someone?

Having a lovely stroll through generative art and data viz at the moment; wasn't ready to immediately bump into NFTs though! Can't tell how i feel about them, but it's definitely making this space feel different to 4+ years ago....

ok ok FINE I'll install a python env manager thinger. I've only used my new laptop for about 2 weeks & already I've got ~700 different vers of python (and pip?) installed. Uh small question though: what does one use in '22? When I was young we were supposed to use pyenv I think?

In my 6 years of building data science product, nothing has been more important than observability. ML, feature building, high volume data processing are all hard & prone to bizarre issues. It’s surmountable if you can see what you’re doing, and utterly impossible if you can’t.

are there lots of scams on linkedin? Seems like the opening for everyone that pings me is: what's your phone number.

"this is becoming epic. deep breaths" -- Wes McKinney 11 years ago about #pandas. Spot on. Wes, congrats on the release of Python for Data Analysis, 3e! wesmckinney.com/book/

When did we start calling normal old data “tabular data”? Not that I’m annoyed by it, I just didn’t notice it needed qualifying. What other fun qualifiers do we use now?

New product idea: corporate training software where they give you all the multiple choices up front and if you get 100% you don't have to sit through an hour of animated videos.

moving back to Python from Go is oddly miserable? I've gotten so used to clicking on the function name in the docs to get straight into the code; suddenly that being more than one click makes libraries seem obtuse and sorta defensive.

They've trained an AI to explain AI papers. It's almost taking the piss explainpaper.com/papers/attenti…, but not quite. Man Carrying Thing, as always, comes to mind: "The poem must resist the intelligence Almost successfully." - Wallace Stevens

If expired certs are your only problem then you have ALL THE PROBLEMS.

moving over to mastadon - I'm @[email protected]. Weird thing: scrolling through twitter for about 6 minutes looking for posts from people I actually follow and there are almost none. It's all trash. Mastadon is better almost immediately.