dmarble

Plot twist: The * in C-00000291*.sys is "-block-ultron"

Premature deployment of Crowdstrike AGI disaster response plan.

I'm in awe at the complexity and unknowability of it all, but I also have to chuckle at the thought that some portion may be vestigial.

I'm particularly fond of the "Egg shaped object with no associated processes". :)

Very similar story here on the PM side in the US.

~15 YOE (FB/Meta most recently), ramping up a search after a sabbatical, targeted job search to roles where I have non-trivial experience and domain expertise, customized cover letters, leveraging my network, open to relocation, open to hybrid or remote, etc.

I'm seeing 3-5% response rate over the past few months. It's rough out there. No response for seemingly great matches. Slow moving recruiting process, even at early-stage startups. Rejections after screenings and first round interviews where the mutual fit seemed excellent.

Hiring manager friends and talkative recruiters tell me that in contrast to the past decade where they'd routinely screen people who met most of what they're looking for, they're now dealing with a massive volume of very qualified applicants (and trudging through a massive volume of unqualified applicants). Deciding who to screen and who to do a first round of interviews with is taking a lot more time and effort. And the first round of interviews might include 6-8 unicorn (i.e. perfect) candidates, where in the past they'd be elated to find 1 unicorn.

I've been through a couple down cycles, so I'm focused on grinding away till I find something. I think every level is feeling the pain in some proportional way. Big sympathy for early career folks. Even if we reset compensation expectations, it'd be a shame if the sector ends up losing out permanently on a range of talent. There's no way tech needs are going to decrease on a medium-term horizon (though they may shift).

India's beef exports are primarily buffalo meat ("carabeef"), not cow meat. In addition, there are states in India where slaughtering cows is permitted.

Huge congrats to you and the crew, Spencer! Well deserved. It was always a pleasure working with your product and the people behind it at prior companies, and lobbying for more adoption as a result.

1. What are you most looking forward to (product/tech-wise, impact, financially for you, financially for the organization, etc.) that may not have been possible just a few years ago?

2. What do think are the greatest challenges you will have to face in your role as a founder-CEO and as a company in AMPL's next phase?

3. If you were starting over today, what ideas might get you excited enough to go at it again for the next decade?

Don't any of the major commercial open source vendors offer custom terms to give access to the commercial source? I'd imagine they'd contemplate it for big deals. Seems like one of the only ways to keep some of these sophisticated customers onboard.

I work on identity resolution problems among other data challenges at a big retailer. We get very few details when a credit card is used in-store -- pretty much just the basic card info. We don't own the payment networks and a variety of legal and business reasons prevent us from doing what may seem to be possible to get an individual's data based on a card number (from the networks themselves or other data brokers). What's possible is mostly limited to what you can do with the name on the card and store location. We've invested significantly in the problem and the answers are a patchwork of guesses.

There's an interesting related issue here for brick & mortar businesses with CCPA and GDPR in effect: you can do some useful analytics, personalization, and fraud prevention work with probabilistic identity info, but if someone verifies they actually are Person X and wants to download or delete whatever data you have on them, what can you confidently say is actually their data?

Will companies be held to different standards based on how much money they've invested and success they've had in identity resolution, in which case this might be a factor dissuading them from doing more identification and personalization? Or if they haven't invested millions in trying to figure out who people are, but it's possible to do so, are they liable for some kind of misconduct if they don't produce all the data they have that could have been tied together for that person? Is the choice binary, i.e. either invest big in identity resolution and take it as far as possible (with parallel governance investment) or de-identify everything you can? A privacy advocate might think on first pass that it's as simple as choosing the latter, but that's mostly not possible due to requirements we face related to other regulation and business realities: fraud, anti-money laundering, age-related laws, shoplifting, intense competition in a razor thin margin industry, etc.

Data privacy is complicated.

  On Android Chrome with Data Saver turned on, elements with loading="auto" or 
  unset will also be lazily loaded if Chrome determines them to be good 
  candidates for lazy loading (according to heuristics).

I don't agree with this design decision to make the default ("or unset...") allow for lazy loading, even if limited to Data Saver enabled phones. Doesn't this mean every site on the planet that deems Android >= 7.0 web traffic important and makes use of pixel-based or iframe-enclosed tracking will have to go through testing and potential modification?

A couple summary quotes:

"[W]hen [Dr. Kent] Berridge, a dedicated young scientist who was more David than Goliath, stumbled upon evidence in 1986 that dopamine did not produce pleasure, but in fact desire, he kept quiet. It wasn't until the early 1990s, after rigorous research, that he felt bold enough to go public with his new thesis. The reward system, he then asserted, has two distinct elements: wanting and liking (or desire and pleasure). While dopamine makes us want, the liking part comes from opioids and also endocannabinoids (a version of marijuana produced in the brain), which paint a 'gloss of pleasure', as Berridge puts it, on good experiences. For years, his thesis was contested, and only now is it gaining mainstream acceptance.

. . . .

"'It's easy to turn on intense wanting,' he says, when we eventually sit down. 'Massive, robust systems do it. They can come on with the pleasure, they can come on without the pleasure, they don't care. It's tricky to turn on the pleasure.' He hadn't expected his findings to turn out this way, but it made sense. 'This may explain', he later tells his audience, 'why life's intense pleasures are less frequent and less sustained than intense desires.'"

== On mindfulness meditation: ==

"It's not that meditation makes the wanting go away. 'What it is doing', Berridge says, 'is giving the more cognitive mind a way of distancing itself from the urgency of those wants. It's a practised mental gymnastic. A want occurs, but because you're so practised, you can recognise that want, appraise it, feel it all around, focus on that, and the feeling of urgency as a feeling, without engaging in it.'"

== On controlling our desire/pleasure systems: ==

"The strength of dopamine/wanting in an addict's brain is so fierce that it is hard to conquer. ... Walter Mischel's famous marshmallow tests told children that they could forgo one marshmallow for the promise of two if they waited a while. Mischel tracked the children in later life and found a link between self-control and success. The controlled kids had resisted the marshmallow by simply making a decision and moving on without further discussion. They turned away from it, or tugged their pigtails to distract themselves from allowing it to arouse their senses. The children who deliberated, or lingered over the marshmallow, were more likely to cave in. 'It looks as though the best way of resisting is not to open the question[.]'"