My general impression of kids from elite colleges are that they're very good at finding some sort of loophole in the system to exploit, and they get lauded for it. And if they balk, for whatever reason, they feel like they're falling behind those that do. So then there's a feedback loop for everyone to take advantage of some kind of exploit to stay competitive. "You'd be stupid not to do X also, if everyone else is." with no consider to morals or character--because they're not easy to measure.
You just described Sam Altman to a T. Never forget "Founder Naughtiness," a quality pg prizes in founders, which he originally ascribed to sama:
> 4. Naughtiness
> Though the most successful founders are usually good people, they tend to have a piratical gleam in their eye. They're not Goody Two-Shoes type good. Morally, they care about getting the big questions right, but not about observing proprieties. That's why I'd use the word naughty rather than evil. They delight in breaking rules, but not rules that matter. This quality may be redundant though; it may be implied by imagination.
> Sam Altman of Loopt is one of the most successful alumni, so we asked him what question we could put on the Y Combinator application that would help us discover more people like him. He said to ask about a time when they'd hacked something to their advantage—hacked in the sense of beating the system, not breaking into computers. It has become one of the questions we pay most attention to when judging applications.
Loopt, BTW, eventually became a shady gay hookup (not even "dating") app--the digital equivalent of the men's rooms in the Port Authority Bus Terminal--before getting acquired for barely more money than it raised in VC funding, and even then, only because one of its VCs was also on the board of the acquirer (Greendot). None of this is mentioned in Loop's sanitized wikipedia entry: https://en.wikipedia.org/wiki/Loopt
Lately, I've been writing more on my blog, and it's been helpful to change the way that I do it.
Now, I take a cue from school, and write the outline first. With an outline, I can use a prompt for the LLM to play the role of a development editor to help me critique the throughline. This is helpful because I tend to meander, if I'm thinking at the level of words and sentences, rather than at the level of an outline.
Once I've edited the outline for a compelling throughline, I can then type out the full essay in my own voice. I've found it much easier to separate the process into these two stages.
And yet, Will, with all due respect, I can’t hear your voice in any of the 10 articles I skimmed. It’s the same rhetorical structure found in every other LLM blog.
I suppose if to make you feel like it’s better (even if it isn’t), and you enjoy it, go ahead. But know this: we can tell.
The essays go back a couple years. How did I use LLMs to write in 2021 and 2022?
If you're talking about something more recent, there's only two essays I wrote with the outlining and throughline method I described above. And all of essays, I wrote every word you read on the page with my fingers tapping on the keyboard.
Hence, I'm not actually sure you can tell. I believe you think I'm just one-shotting these essays by rambling to an LLM. I can tell you for sure the results from doing that is pretty bad.
All of them have the same rhetorical structure...probably because it's what I write like without an LLM, and it's what I prompted the LLM, playing a role as a development editor to critique outlines to do! So if you're saying that I'm a bad writer (fair), that's one thing! But I'm definitely writing these myself. shrug
Mainly: I wanted to get all counts for Wodehouse, but didn't want to block the page load for that. So what I did was take a small subset of the data that's needed to render the page, and have that written into the file. Then after load I fetch all the other counts.
In the tutorial I also used Bun.xxHash3, but to make sketches run on the client, I had to replace that with a library.
The actual sketch implementation that backs the demos is here:
The caption reads like an AI prompt. The last instruction, "think of something to add" resulted in the AI adding a pterodactyl on the house, dropping bird poop on it. Probably not what the prompter meant, but technically correct.
I hear people talk like this on the phone. The one I hear a lot is: "It's not about X, it's about Y1. It's about Y2. It's about Y3." Where Y is usually something humanizing.
There is the Handmade and Better Software contingent that is pushing against this trend. I don't know if they'll succeed, but they at least care about quality in a way that a lot of software engineers aren't incentivized or don't know how to do anymore.
Despite trying to instill a customer-centric culture, as soon as Bezos let his foot off the gas, his company just isn't as customer obsessed. Or, they changed their definition of customer from the buyer to the seller.
I dramatically lowered my buying from Amazon about 8 years ago, when I noticed that listings had reviews on items that were completely different than what was being listed. Apparently, sellers sell a known good product that gets good reviews, and then swap it out for something else, so that the new product can piggy back off of the good karma. Amazon just didn't shut this down for years. Also, when Fulfillment services by Amazon mixed the the official provider's inventory with 3rd party distributors and reseller inventories. Sometimes, people would get knock-offs. I knew then, Amazon would coast for at least a decade before the decline would be apparent.
I thought I'd buy more Shopify stock as a result. Dunno if I ever did.
> Or, they changed their definition of customer from the buyer to the seller.
s/seller/shareholder/
I almost never buy from Amazon any more. For certain things it is difficult because Amazon has destroyed so much logistics and has such a stranglehold that a lot small/medium sized companies only sell through Amazon now. I ordered some kitchen gadget a few months ago from the company's own website, thinking I was avoiding Amazon, and it was delivered by an Amazon driver.
Another mind boggling aspect of Amazon’s review system is that it categorises multiple products / variations under one. So if you want to buy products X with variations X1, X2 and X3- the review page for product X1 will also show X2, X3 ratings bundled together. You don’t quite know the rating of X1 individually. You can filter reviews out by the overall rating is the aggregated view. I can’t believe how this is helpful for customers.
I was trying to buy a copy of the Iliad and found that they combined the reviews of EVERY TRANSLATION AND EVERY EDITION - I found people talking about 30+ versions on the same listing!
Reviews for Fagles's Iliad were combined with Pope's Iliad and Lattimore's Iliad and so on and so forth.
Navigation is also borked for books with many different versions - if you play around with the 'hardback', 'paperback', 'audiobook' buttons at the top of the page you'll find there's no consistency about what edition they lead you to.
Maybe that's to combat the strategy where sellers add an irrelevant category or hyper specific one, in order to get the "Best Seller" stamp on their thumbnail.
I slowed down my Amazon purchases drastically when everything I shopped for seemed to be flooded with these thousands of identical products have seemingly randomly generated all-caps brand names like BHHSRE, VHYXZY, XIOU, DAUGHE, JXMOX, LANMU, IBERLS, GMJYC… (yes, these are all actual brands I’ve seen on Amazon).
Jeez, the least they can do is make it look like they are trying to curb abuse.
Bezus checked out long before he moved to emeritus. Wanted to be a movie producer and shoot his rockets. Andy took over with all of the vision of a corporate franchiser running a Dollar Store.
reply