Yesterday I delivered the ID lists, and prototyped a thing that will do that automatically. I went through a couple rounds of email with legal to get all the details of what data we’re supposed to be sharing with whom for the individual survey responses, and it was a good thing I did because there were a bunch of changes nobody had thought to tell me about. We had the big lab meeting, and R had a real actual agenda and a plan for what happens next and how we do it without burning everybody out. It includes hiring PMs and engineers and hopefully some customer support folks too, so if you know anybody looking, hmu. I met with F, and nobody had told them about the changes in the individual responses release process either. It was an extremely productive and extremely short meeting, very much my style. In the afternoon I looked at the status of the IRB modification, and they had flagged the consent form as needing a change — since we’re adding two typically-European locales, we have to be very sure that participants are currently residing in the US so that we don’t run up against GDPR. Fair, but it means F has to run the change back to the translation team. After that, I chased down an irregularity with a small number of IDs, which turned out to be F’s data team misunderstanding the change to the incremental window. I spent the rest of the afternoon and a sprinkle into the evening on crowdcast. Perhaps ill-advisedly we decided to merge the two relevant git repositories at like 3pm on deploy day, so while there weren’t any critical changes to worry about (C had FAQ changes she hadn’t tested yet but those are pretty low risk) there were (predictably) some surprises with ops that took B and D until 10 to resolve. Thankfully they’re pretty autonomous; I lurked in their Slack thread to keep an eye on things but didn’t have to fuss otherwise.

Somewhere in there, dinner and crossword and DS9, and chatting on facebook, and a long hot shower. Snack, and prepped a batch of nut and seed bars, and sleep.

Yesterday for work I updated the crowdcast group on this week’s plan, and attended an epic 2.5 hour meeting that covered restructuring the research group at large, current efforts at acquiring data, and potential new statistical models. The PIs are clearly overwhelmed, but don’t yet fully understand that reducing their own management workload involves both giving up some control over decisionmaking and reducing the amount of time the people they’re tapping to act as intermediate managers have for primary research. Normal growing pains are normal.

I had some good detailed conversations with D, the former grad student on whose work all these systems are based. He is on loan to us from Google. He was cathartically appalled at the dearth of onboarding I’d been given, and was able to dig up some ancient docs he’d made to transition the team after his thesis defense in 2016, as well as the primary research paper for the crowdcasting system. He’s also in favor of the dev system and server improvements I’ve been advocating for. I’ve been managing upwards a ton on this project so it’s good to get validation from someone with cachet.

I met with M on his masters thesis work to look over the code the text world folks finally sent, and we talked through a couple options for extracting the knowledge base for the natural language utterances it generates.

The nuts.com order came, which means we can put off getting staples like oats and polenta and nuts for a while yet.

Food was surprisingly fine; I cut 5 grams from the morning oatmeal for the pair of us and that seems to be setting a healthy tone for me snacking regularly and avoiding overeating.

Dinner, P had his evening class, DS9, crossword, snack, sleep. Dreamt of subway tram and trolley hacking as a 12 year old boy. Also something to do with french new wave records in an old timey shop.

Yesterday I pressed and pinned the next seam on the black armhole. Sorcerer to the Crown is wrapping up nicely. I received a security code for equifax, and struggled to get their website to accept it. Eventually it went through, but it’s still kicking back my login. I will probably have to call. At work, I set up an environment on one of our servers for a small user study, and maybe solved a weird heisenbug with NFS. If it’s still working later today I might believe I actually fixed it. V came to see me about an interesting parallelization problem she’s working on with W. It’s not well-suited to the obvious methods because the bottleneck is in the wrong spot, but there are some possibilities she’s going to prototype/check out/run more numbers on and then come back to me next week. I spent the rest of the afternoon reloading state into my brain so I can fix the knowledge base display on the website. Home, dinner, trash, crossword, then Netflix. There’s no more She-Ra, and I tried about ninety seconds of a new prospect about a chef who gets tapped as the Chosen One in a network of assassins, but it was way too focused on how bad everyone’s injuries were. I switched to iZombie, of all things, but it seems to be sticking. Sleep.

( Read more... )