The fast.ai courses are exceptionally good for getting started and learning the concepts, but, once you've mastered an intuitive understanding of the techniques, I highly encourage people to dig deeper into the concepts - specifically, reverse SDEs, stochastic calculus, optimization, score matching.
Jeremy does an excellent job making the content approachable, but, if you want to go beyond being an excellent practicioner, you cannot skip the theory bits.
FYI this course does cover a lot of intuitive understanding of techniques, but it also does cover this theory. Everything is done from the foundation level, and all of the stuff you mentioned is taught and implemented from scratch as part of the course.
Of course there's loads more to do on gaining intuition, and loads more theory to be learned once the course is done! Only so much can fit in a single course :)
It's a vicious cycle, but, every time one of these internal town halls gets leaked to the internet, the execs lose the ability to speak candidly, and the workers feel like execs are completely aloof and detached from the reality on the ground.
Unfortunately, it's incredibly difficult to maintain a certain level of trust once an org gets over a certain size.
It's just as well. Seeing into the twisted workings of Zuck's mind isn't going to reassure any normal people
"Oh, you saw an opportunity to save money by firing people randomly and threatening the rest, so you can get almost as much work out of us without paying bonuses or promoting people this year? Friedman[1] would be very proud of you sir!"
Leakers also provide a vital window into company culture for people who may need to decide whether to work there. You never get the truth about what the company and leadership is really like during the Potemkin village tour of recruiting and interviewing.
I completely agree re: Potemkin village tour of recruiting, but, those townhalls are not a window into company culture at all exactly because they get leaked.
The only way to get a candid view into company culture is to have insider friends that can tell you what it's really like to work there.
Some VCs were definitely better than others: the very worse was probably the All In Crew who were trying to spread a bank run to tie the government's hand. Truly despicable.
Agreed, I‘m a big fan of the all in podcast but this really showed them (especially Sacks) from their worst side. They clearly pressured the regulators by saying that everyone who has more than the amount insured by the FDIC in their account at a regional bank is stupid and reckless if they don‘t transfer it to one of the top four banks on Monday. The situation at SVB seemed very manageable but even a small chance of this „mind virus“ spreading would be so devastating that they decided to backstop the situation.
Yes, this and his Solana pump (and dump?) really showed his true face to a point where I have to start questioning his qualities as a capital allocator
I absolutely think that depositors should be made whole - and - they will.
At the same time, I think a lot of the VCs who systematically talked shit about government regulations while trying to shill crypto as a solution should very publicly eat crow. I won't lie: it will give me a great deal of personal satisfaction, but, it will also make any "bailout" (or, whatever you want to call it ) a much easier lift politically.
Absolutely: the challenge is that any signal that you use to identify "good websites" from "bad websites" will be adversarially optimized by incredibly motivated people.
You are dealing with a moving target that has a huge financial incentive. It's a very difficult problem.
I don't agree. You start whitelisting good content manually. If babygearlab is the best result for baby gear, you start hardcoding it. If seriouseats is the best result for recipes, you hard code it. If someone better comes along, they get moved up the priority list.
You figure out a way to crowdsource certain decisions and establish who you can trust. Ask them questions with right and wrong answers. You start to tackle it one product category at a time. Instead of pagerank, which was a web of who linked to who" you start figuring which voters you have who consistently turn in good feedback.
This is some form of metamoderation that slashdot tried to implement.
If you are going to be a tastemaker, stop hiding behind "the algorithm" having some mind of its own that cant be controlled.
I distinctly remember a prescient thread from Yishan Wong (former Reddit CEO) about how trying to run Twitter would break Musk - and - it looks like it was spot on.
He paid 44 billion dollars to be a mod - I hope he is getting his money's worth.
Remove the trackers (everything behind ? in the link) and you will see the thread. I was logged in but couldn't see the thread until I removed the tracker.
He had a different reasons however. He said it would be because he can't make any political side happy. We see here different behavior which is tyranny and unrelated to any political aspects. Probably most entrepreneurs have some level of egoistic, self-righteous, tyrant behavioral aspects but private control of social media, gateway to all people, really allows one to exercise and amplify it without any controls. This is why democracy was invented.
There are a few key differences. The main one is that our AI generates an explanation to answer your question directly; we go significantly further than they do in terms of synthesizing explanations. Same goes for snippets; on You.com, you usually need to click on a button (.e.g "Open Side Panel") to get a code snippet which adds friction. Furthermore, they seem to be simply showing the full Stack Overflow page; our approach is to find and rank the most relevant code snippet while offering a "See Reference" button to make it easy to go to the original page.
Overall, our goal is to have the highest signal-to-noise ratio of any search engine when it comes to developer searches.
Beyond the criticism of the format, I found the explicit clout chasing to be completely offputting, and I think I used Clubhouse a grand total of 3x before never turning it on.
I would genuinely love to read an anthropologist dive into what causes social networks to take that turn. At its best, Twitter is full of smart nerdy people talking about science, mathematics, history, etc..
Clubhouse feels like LinkedIn, completely detached from resumes, so everyone wanted to post thinkpieces without being in any way tethered to their concrete achievements.
But that begs the question: why would anyone of note use Clubhouse, if not as an continuation of the mostly "one-way megaphone" strategy that made them big on Twitter or elsewhere?
I wish this was on the GoodNotes/SublimeText model: pay once for a premium version of the app, and then never worry about subscriptions. The moment something requires subscription my threshold for buying it goes up 1000 fold.
I'm usually not too fond of subscriptions either, but I have to say that Muse has the most ingenious subscription model I've ever come across: If you're using it casually, the free version is plenty (Remember that you can always archive stuff by exporting it, and 100 cards will go a long way. Plus, you can stop paying and still have read-access and even limited editing for your existing boards).
If you use it heavily as a daily driver, then the $3.99 or $9.99 should be a drop in the bucket. It scales very naturally between casual use and the paid tiers. Personally, I could probably do with the $3.99, but I'm choosing to pay the $9.99 since I want to support their ongoing research. Their podcast shows the enormous amount of thought they're putting into this, and it shows! Together with Blink (an SSH app with similar focus on UX design) this is by far the most productive iPad app I have.
I mean, yeah, "freemium" is somewhat common, but in this case I find it particularly graceful in that the freemium doesn't limit the features, just the complexity of your board
I have documents I wrote 30 years ago. Looks like this allows export as PDF, at least, but there is no planet on which I will voluntarily rent access to my own documents.
You'll be happy to know that Muse is designed to only gate creation of new documents on the membership, not access to anything you've already created. Lapsed memberships go into alumni mode where you can navigate, move, copy, and export everything.
Agreed. Subscriptions make sense if it's a "business" app, i.e. something you are using to run your business, e.g. Adobe CC, Office 365.
But this one looks like personal app, with little in the way of corporate functionality (Kanban boards, todos, etc). Why not charge a flat fee for 1 year of updates, and make updates optional after that?
I have the same beef with the makers of MindNode, a mind-mapping app for MacOS. It's $2.99/mo which is not much, but I chose to go with SimpleMind Pro because it was a flat 25 EUR.
Developers know the term "write once, deploy anywhere". How about "charge once, use forever"?
I get your metaphor, but I think it falls down because "write once, deploy anywhere" was a slogan for Java and it never really fulfilled that promise.
And yes, like anything in life there are top-quality options which tend to be more expensive and lower-quality options which fit a smaller budget. MindNode is a lovely app and worth paying for IMO.
