There's a number of people building various flavors of this. I'm one of them -- pulsemcp.com - we expose our data via API (pulsemcp.com/api) that is theoretically the starting point for what you're asking.
The issue with the current state of the ecosystem to do what you're saying is that server implementations all have extreme variances of quality. Some are very robust and reliable. Others have very poor and limited UX, if you can even get them to start up on your local machine.
So we're not really in a place where we can trust that a server "does what it says it can do [well]", and so any autonomous solution to "choose tools (servers)" probably won't work well today (yet) while the signal to noise ratio is poor.
Personally: I think autonomously choosing between tools is not the most compelling approach in the short to medium time horizon. I expect client application creators to get a lot of mileage by thoughtfully curating the possible servers exposed to their users, and building nice UX to integrate usage of relevant servers as-needed.
As a software engineer that spends a lot of time in marketing, I didn't find GPT 3.5 useful for anything besides writing marketing copy faster.
I have found GPT-4, on the other hand, to be excellent at writing complicated SQL queries, one-off scripts in just about any language, expanding and fleshing out content ideas from just a seed of insight, and debugging cryptic error messages. Among some other one-off use cases, but these I use on a recurring basis.
I'll throw in another wrinkle: it's not just the candidate pool to consider, but how companies are choosing to leverage - or avoid leveraging - the platform.
I wouldn't say Triplebyte imposes a significant cost on the candidate. They only have two rounds: one is a quiz that's quite straightforward to complete (about a half hour of multiple choice, if I recall correctly), and a single 1-2 hour video interview. If nothing else, it's good practice for the candidate.
Plus, they dangle some brand names as partners that would attract any candidate, including the cream of the crop. Stripe, Palantir, etc. So top tier candidates are certainly likely to be convinced to give Triplebyte's process a shot.
I'm not sure how often candidates get matched with those top tier companies on Triplebyte -- I didn't get matched in my recent job search, and ended up applying and receiving offers from several of them independently of Triplebyte -- but it's certainly plausible that Triplebyte has many top candidates at least giving the platform a try.
Regardless of which candidates are using Triplebyte, the only relevant data is which candidates are _getting offers via Triplebyte_. In my experience, I received several offers, only one of them via Triplebyte (I only accepted one onsite there) - Triplebyte has no insight into my other offers.
That one offer via Triplebyte was significantly lower - at least on base - than all of the other offers I received. It would've put me in the 50ish percentile on this post's plots. The other offers put me in ~60-97 percentile. Very anecdotal, but I wouldn't be surprised if the majority of companies giving offers via Triplebyte are generally paying less than top companies who are less likely to use Triplebyte despite having a presence on the platform.
While I appreciate the sentiment of this post, I'd like to see some real stats on "hustle porn" culture's effects. My impression is that the author is pitching doom and gloom where there might be only a small - and increasingly diminishing - problem.
I'm not constantly peppered by this phenomenon on my feeds. In fact, lately I've been find more of the opposite: the growing culture of valuing 8 hours of sleep, mental health, and the like. This trend seems to be especially strong in the world of entrepreneurs. YC's Startup School had a talk on "How To Win" [1] - all about how to stay sane and healthy as a founder. Podcasts like ZenFounder [2] are teaching more of the same and building up that field. And I haven't even read it yet, but the popularity of books like It Doesn't Have to Be Crazy at Work [3] represent a growing movement of successful people - who I'd say still work quite hard - that avoid these pitfalls and are very loud in teaching others how to do the same.
I'm sure, as the author suggests, there're people who take all the motivational talk to an extreme. But there's an alternative that I hope many of us are following: don't just blindly work hard - because working hard is important - but work smart, too.
That could be because you consume media on the bleeding edge of these trends.
"Hustle Porn" was quite popular in tech several years ago. It lost its luster and it's no surprise that the hustlers have now moved into broader markets.
Judging from Rob Walling's thoughts on this topic from his book -- the equation looks different for the kind of company they're looking for. There's no "huge risk of failure"; the focus is on businesses highly likely to succeed that will have (relatively) slower, but consistent, growth. And obviously (relatively) lower final valuations.
Bingo. Instead of 1 in 20 becoming a huge success (100x) and the other 19 fail, maybe for us it's 10 in 20 are "base hits" that return 2-5x and the other 10 fail.
Those are contrived numbers for this example, but you get the idea. I expect more of them to be singles/doubles, fewer to fail than the typical VC bets, but we'll have no home runs.
Have you seen the total returns of VC money? Too lazy to link stuff now but a perfunctory search will reveal that the returns are not what you think they are. They essentially approach 0% outside of the few powerhouses.
Can’t really use the Andreessen-Horowitz / Sequoia returns as representative to the entire industry.
This sounds like a big deal, but doesn't seem to be getting attention. Am I missing something?
Giants like Facebook, Twitter, etc. have a huge competitive advantage because of their user moats. There are a countless number of people who hate Facebook, but stay on it "because everyone is there" and use the one or two convenient features they appreciate, like Events or Groups.
If this initiative is for real, goodbye inferior products like all the stuff Facebook has stuck into its ecosystem (Groups, Events, Messenger, etc etc) and hello super-speciailized companies that can piggy-back on existing user moats (moats no longer).
This is obviously fantastic for the consumer, but seems to destroy something that I thought to be key to each of their bottom lines. Why are they doing this?
Google's motives tend to be just having as much data as possible so they can draw powerful inferences from it, in contrast to Facebook which mainly values exclusivity over their data. By playing the "good guy" card, Google:
- Generally encourages more data to be out there and available
- Pressures companies like Facebook to play along, weakening their data strongholds, lest they be painted as the bad guys
- Gets a lot of goodwill and, more importantly, user trust, which encourages people to give them even more data
While Google's interests seem to align with consumer interests more often than those of the other tech oligarchs, they don't do anything out of sheer good will.
Keep in mind, a huge company is also just a collection of people.
As far as I understood, this project is from the developers working on the gdpr-originating user export functionality. Maybe they've just been passionate about it and had a vision for a more user friendly solution and managed to push it through somehow.
To be honest, it didn't get my attention yet, since I don't know what useful thing I could do for example with exporting all my Facebook data. I would only be slightly worried with how to protect them from theft, and basically their usefulness is tied to the Facebook web and mobile app.
There’s a risk regulators will force companies to allow exporting data. If companies work on this voluntarily, regulators might be satisfied and not regulate this.
As some commenters have alluded to anecdotally, there's science behind the idea that "follow your passion" is bad advice. So it's not something you find, but rather something you build. There are probably very many things that could become your passion(s) if you build them.
I'm still working on building mine. I've found at least one (coding) that fits into the picture somehow, being that I've been doing it since I was a kid. Some others are more recent interests that I want to spend a few years diving deeper into before rendering a verdict.
Cal Newport wrote a whole book refuting the "follow your passion" hypothesis in 2012: So Good They Can't Ignore You [1].
And more recently, there's a Stanford study out that makes the same claim [2].
There is another great book written by Cal: Deep Work [1].
Don't follow your passion. Instead, become really good at something. Apply methodical approach to improve your craft skills. Once you got mastery, you might actually like it.
I've read the book. I actually disagree with Cal on the that one for the simple reason you can be really, really good at something and not passionate about it.
I think there's a trick here though - get really good at something and then use that to make someone else's life better - now that is something you can probably get passionate about.
This whole passion debate will run and run though - I don't think anyone really has the definitive solution - it will be different for different people.
I'd suspect a valid correlation that's not a causation. If a company is generally doing well, presumably the CEO isn't under as much stress and doesn't have as many firefighting responsibilities as those whose companies are doing poorly. Hence having the time to write a letter to a stranger.
Maybe those underperforming companies will bounce back and you'll find a (much delayed) response in the author's mailbox a few months down the line!
I think the correlation is real. The type of person who writes to a CEO is likely to do other things for the company. For instance they might be a undercover reporter researching a story. Responding to the letter can give the company free press, which can be far more valuable than the time a personal letter takes.
Likewise, many companies have a comments toll free number. Almost nobody will call them (who has a comment about their toothpaste), but it turns out the type of person who does call those numbers is likely the type of person who tells all their friends which is the best toothpaste, and so it is important that whoever answers that phone is an expert on what makes their toothpaste great. (honest answers are important - this type of person already knows their stuff so lies will hurt)
Also, CEOs often know they have trouble connecting with their real customers. If a real customer writes, that is real information that they cannot easily get in any other way. A good CEO will pay attention because what one person writes 100 others think but don't say at all.
All of the above are things that make a CEO who doesn't respond under perform.
The example of an undercover reporter sending a letter to a CEO under the premise that they will write a nice article about the company if the CEO responds is pretty absurd.
It doesn't have to be a nice article. A request for comment can either be responded to, where you have a chance for your words on record, or you get "<companyname> has not responded to requests for comment as of publishing." which is worse than whatever spin you could have concocted.
Reporters who are reviewing products do. They might have 10 different blenders to review, and if there is a button that doesn't make sense a good response takes you from "2/10 the buttons don't make sense" to "9/10, the buttons are a little weird but once you get used to them..."
I really like his example at the end - step-by-step deep dive into a specific topic. 30 hours of learning, direct links to resources, come out the other end as informed and aware of the topic at hand.
I usually find it difficult to wade through the mountains of possible (often outdated) resources on a topic I'm unfamiliar with - At one point I was thinking it'd be useful to crowdsource those kind of step-by-step guides across a variety of verticals. Kind of like an online class, except pulling from the best up to date blog posts and specific resources from around the internet, instead of relying on a singular POV of the guy who made some videos and called it a course.
Anyone know of anything like that? How do you approach deep diving into unfamiliar spaces?
I think this is a great idea. It's really saddened me watching the internet turn from a place where people deeply invested in a topic or passion share their work personally to a SEO-ranked shit show where millions of low-effort 'blogs' compile second hand knowledge for views and commissions.
While compiling great resources in one place is a good stop-gap I'd love to hear some ideas about how we can move towards a Web where people are rewarded for creating original content and putting effort into their work, rather than the current mess of clickbait and rehosted/paraphrased content. Doing a Google search for a technical topic is much less effective now than in the past, by my reckoning.
I have found that good high quality blogs typically link to other high quality blogs. Most of the ones I am familiar with are math/computer science blogs run by researchers.
If those topics are in your wheelhouse there is definitely a ton of high quality original content from the advanced undergraduate level to the research level.
so maybe a lot of the internet is seo ranked shit show...but not all of it.
I have found myself asking the same question. While I don’t have an answer I can share my plan. I am studying the social impact of the printing press. I don’t think you can go wrong studying history.
I'm building this right now actually. Superclass [1] is a place to submit online learning resources by topic, then upvote/comment on the most helpful ones.
The goal is eventually something just like you described, kind of a crowd-sourced curriculum builder.
This version is rough and buggy, but would love everyone's thoughts/feedback (sam@directedworks.com)
Hey, I just checked your project out and it's awesome. One thing id like to see is a way to filter submission types. Specifically it would be nice to filter out products likes apps and books.
Why is there a log in button at the top right but not a sign up? I know with the google auth thing it's the same, but as a user you don't know it.
Perhaps make an entire "contact" page, I dislike mailto: links. It opens my company's preferred Email client called IBM Notes and it sucks.
Once you have your MVP, please add transitions/animations. The site is too instant so to say, when clicking a button you expect some small delay. Even if you pre-loaded this, it's sort of important.
Honestly, this is why I just read technical books cover to cover. It’s just about the best point to point resource out there.
I think I remember seeing where Cornell offered a course structure where you’d do one course, all day, for 3 weeks and then move to the next. It was years ago, but I remember thinking I would have loved it.
I've tried that for a couple topics (Reinforcement Learning, Linear Algebra, Bayesian Stats) over the last six months and I found for myself it was much harder to retain information than spacing it out over several months and chipping away at it little by little (which I'm doing with Python, Algorithms, Control Theory). I personally prefer practicing interleaving, as I find that applying one subject to another make me remember a lot more of the concepts than block learning techniques.
Also in the grand scheme of things, three weeks isn't a lot of time. Granted, I'm not a fast learner!
This is generally called a Block Plan[1] and is a great system for some students. Looks like Cornell College offers it, you got my hopes up as I'm a Cornell University student and juggling five technical courses at a time has heavy context switching penalties. On the other hand, last year taking Linear Algebra and at the same time having to apply it in a signals processing class was a good combo.
Yeah, I used to take intersession courses back in my college days where I focused on one course for four weeks. That fit my mode of study way better than taking 4/5 technical courses at the same time. Doing a deep dive on one subject and then moving on was just way better for subject retention in my opinion.
Condensed studying feels easier, but leads to lower long-term retention. Now, there are methods used to help mitigate that, but those can also be used to increase retention of spread out studying too.
This also reminded me of some summer classes. Maybe one month, 5 days/week?
Some of the things that best remember from undergrad were learned over the summer.
I still don't know whether it was the short/intensive set up of the courses that improved learning, or if it was the fact that I was only willing to do summer courses in topics I knew I'd do well in. Either way, it's my best memory of school.
I love her Twitter. Rarely see girls post highly technical content. This is her only non-technical one I guess and all her posts are high quality and about assembly which helped me get started with it although I’ve never done anything in assembly before. I’m a web developer but recently got interested in assembly because of her account
She has meanwhile deleted her reply that showed consent with the cited tweet. I have a screenshot but that would prove nothing to you.
Actually, I’m quite happy with the outcome, I paid a little bit of karma for somebody to reconsider her actions.
Some great suggestions here already but thought I would add Metacademy [0] and Learn Anything [1] (White Paper here [2]). These are designed to create a map of skills and concepts for a given topic. I find an interactive visualisation to be really effective in understanding the broader ideas before starting out or during the early stages when it’s hard to see how the pieces fit together.
This. There are some great bits about checklists and their (underutilized) role in preventing human error in Don Norman's Design of Everyday Things (chapter 5).
The issue with the current state of the ecosystem to do what you're saying is that server implementations all have extreme variances of quality. Some are very robust and reliable. Others have very poor and limited UX, if you can even get them to start up on your local machine.
So we're not really in a place where we can trust that a server "does what it says it can do [well]", and so any autonomous solution to "choose tools (servers)" probably won't work well today (yet) while the signal to noise ratio is poor.
Personally: I think autonomously choosing between tools is not the most compelling approach in the short to medium time horizon. I expect client application creators to get a lot of mileage by thoughtfully curating the possible servers exposed to their users, and building nice UX to integrate usage of relevant servers as-needed.