Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Grant Slatton Profile picture
Grant Slatton
@GrantSlatton
Recent writing at https://t.co/lCWFZXt72C Formerly built the world's fastest filesystem at AWS, now the fastest spreadsheet at https://t.co/hLkbCuJG7H
Mar 30 • 4 tweets • 1 min read
many working in finance pay $25K/yr for access to Bloomberg Terminal

will probably start to see this level of spend on AI use for devs in the next year or two

chatgpt pro is already $2.5k/yr

as it gets better and more useful, people will bid that price up The number of responses that assume that price is a somehow a function of profit margin is astounding

A Toyota Corolla vs Honda Civic are essentially interchangeable commodities in a low-margin, competitive market — they still cost tens of thousands of $$
Mar 25 • 51 tweets • 13 min read
tremendous alpha right now in sending your wife photos of yall converted to studio ghibli anime Image
Image Ghiblification will continue until morale improves
Feb 28 • 4 tweets • 1 min read
parents who really want grandkids need to properly align incentives

have N kids yourself

then tell them when you die, your estate will be split between them in proportion to how many kids they have non-linear is even better

Jan 24 • 5 tweets • 2 min read
Met a guy who runs a single-serving website that’s just a tiny wrapper around a python lib

User googles niche problem, finds his site, pays $5 to fix their problem

Makes $10K/mo completely passive, niche enough to have no competitors

Site could be cloned with 1 day work To get an idea of what it is, imagine a normie-friendly wrapper around ffmpeg or imagemagick or some other common data transform toolkit

Told him I wouldn’t reveal obviously, so will not elaborate further

But there is truly low-hanging fruit all around, kind of astounding
Jan 15 • 7 tweets • 2 min read
It's remarkable how consistently understaffed and off-the-ball most recruiting departments are

E.g. recently have known 2 hyper-competent hardcore backend engineers be having conversations with OpenAI recruiting and just get ghosted randomly

It's really all bigtechs though When I joined AWS I gave a list of ~5 former schoolmates who would be a great fit and recruiting fumbled 100% of them

I was once considering going to Stripe and the hiring manager and recruiting fumbled me, ghosted while scheduling interviews
Dec 29, 2024 • 4 tweets • 1 min read
For disclosure, here are the questions I asked at faang:

Warmup: write a function that takes a string of parentheses as an argument and returns true if the parentheses are balanced

e.g. (())() == true, ()) == false

Next: given a grid of… [cont]

Given a grid of booleans, where True is land, False is water, and adjacent lands form an island, count the number of islands

(this has been a popular leetcode medium since 2015)

And finally, if I was instructed to focus on datastructures instead of algorithms, I'd ask… [cont]
Sep 18, 2023 • 7 tweets • 3 min read
The new GPT model, gpt-3.5-turbo-instruct, can play chess around 1800 Elo.

I had previously reported that GPT cannot play chess, but it appears this was just the RLHF'd chat models. The pure completion model succeeds.



See game & thoughts below:
Image The new model readily beats Stockfish Level 4 (1700) and still loses respectably to Level 5 (2000). Never attempted illegal moves. Used clever opening sacrifice, and incredibly cheeky pawn & king checkmate, allowing the opponent to uselessly promote.

lichess.org/K6Q0Lqda

Image
Image
Dec 11, 2022 • 7 tweets • 3 min read
GPT can iteratively write, debug, and test programs to accomplish arbitrary goals.

Pictured: GPT reading snippets of HTML from HN and building a headline scraper in Python, overcoming bugs by simply reading the errors and self-judgments and hypothesizing to itself.

Thread ↓

As shown in my previous post, GPT can be embedded in a REPL to accomplish goals in an agent-based fashion by using command line tools. Today's post shows it is capable of creating novel command line tools, such as a web scraper.

1/
Dec 8, 2022 • 10 tweets • 3 min read
GPT can use a web browser to answer questions.

When embedded in a REPL environment and prompted to strategize and monologue, agent-like behavior emerges. The agent can solve multi-step problems that involve going to pages, following links, reading the next page, etc.

Thread ↓ ImageImageImage The real HN pages showing those answers: ImageImage