Cercerilla @cercerilla@octodon.social

Reading ars-Technica’s editorializing about ad blockers & playing the worlds smallest violin. Oh boo-hoo publishers might need to innovate and figure out how to sell content without mercilessly exploiting readers for ever last minuscule scrap of data that can be mined and mindshare that can be exploited.

Ad-tech is killing the world and I haven’t the smallest shred of sympathy for any entities that won’t move beyond that broken exploitive model.

@sharp that does make sense, we keep our house quite a bit colder overnight.

Is there any scientific reason why smoke detectors always start the chrip of doom in the middle of the night? Observation bias? A deep seeded hatred in REM sleep? Also why do they still use 9v batteries that nobody has kept around since the 90s so you have to go slowly insane as the tell tale smoke detector slowly chirps away at your very soul minute by minute as you wait for any store in the world to open since it’s New Year’s Day.

@b sure, all 12 of us 😅

I just ripped my last pair of pants, but I work from home, so who'll know? 🤷🏻‍♀️

I'm spending my New Years eve researching which types of humidifiers are safest for avian respiratory systems. How about you?

It's been a slow burn, but there's no putting the fire back in the matches now. We're in a world where ad-tech is pervasive in nearly every communication form, and getting more and more effective at massive scale individual targeting and understanding how to affect individual people. There's no material difference between the tech that makes you buy a Starbucks Latte and the tech that makes you start to support Nazis.

One thing that isn't discussed enough is how the current dominance of ad-tech on the internet actively facilitates propaganda, e.g. of the sort used by Russia or US super-PACs to influence elections. We've built a system that is incredibly efficient and finely tuned to follow individuals around and target them with information specifically tailored to each demographic to maximally affect their feelings and actions, and commoditized this access to human minds

@b We can go back to the days of web-rings so help readers discover content without having to go through a search engine or news aggregator. 👵🏻

@b more or less yeah; not much. I'll still have to consume some things out of necessity, but I'll limit and carefully control how I write into the system to minimize my footprint.

@b I'll continue to use some of them, like reddit, out of a level of necessity, but wherever possible (like with reddit) will do so anonymously or with purpose-created specific temporary identities designed to limit the amount of information that can be deduced about me.

@b New Years resolution: slowly extricate myself from all online services that are monetizing my data, time, or attention.

@vector I don't have a video recording but slides here: https://speakerdeck.com/rebeccaskinner/a-brief-survey-of-machine-learning-in-haskell

I'm currently auditing the source for Shreddit https://github.com/x89/Shreddit before using it to clear my data from reddit; has anyone who's more comfortable reading Python than I am gone through and audited the source recently? I don't want them to backchannel any of the data I'm deleting into an external database.

@fardog @craigmaloney Even with auto-fill disabled LastPass seems to have removed the option to manually copy the username & password for a site, so you're stuck either going into your vault and viewing the password in plaintext to get at it, or auto-filling everything on the page, which could potentially include any advertisers scripts. Seems like they should allow you to fill in a specific field with a specific bit of data from your vault.

I needed a mental break from the #haskell book so tonight I decided to pick back up with #converge and start trying to get a 0.7.0 release out. Mostly bugfixes and migrating from glide to dep, but the major feature will be a total rewrite of the HCL parser, with the goal of eventually allowing us to migrate toward a more powerful configuration language.

I'm presenting on Strongly Typed Deep Learning with #haskell in a few weeks to the #StLouis machine learning and data science meetup.

I wish that #haskell's #tensorflow story was a bit better. I see a lot of potential for some really great capabilities to emerge from the combo of a language great for DSLs and with a lot of correctness guarantees, but the current solutions seem to require too much #categorytheory for a gentle intro to non-haskellers

Any suggestions for ways to make it easier?

Having to write #javascript makes me realize why so many startups have kegs in the office.