When you lose the ability to write, you also lose some of your ability to think

This page summarizes the projects mentioned and recommended in the original post on news.ycombinator.com

Our great sponsors
  • InfluxDB - Power Real-Time Data Analytics at Scale
  • WorkOS - The modern identity platform for B2B SaaS
  • SaaSHub - Software Alternatives and Reviews
  • nitter

    Alternative Twitter front-end

  • snippets

  • > Totally agree, I also pay for api access and I am still trying to figure out how I can benefit from using it when it gives me blatantly incorrect commands or wrong answers.

    So here's one that saved me a bunch of time:

    https://gitlab.com/-/snippets/2535443

    Basically, I knew the package could do what I wanted; I knew it was almost certainly in GPT-4's data set; I could do it myself, but it would involve searching through all the documentation and figuring out the right bits. It just did it for me.

    Now there were a few minor bugs: It duplicated a buggy error message, and at some point later it called a "Subtree" method rather than a "Tree" method. But those were a lot easier to fix than writing the code from scratch.

    Once I had a list of 27 book names I wanted put into "canonical order" and in the form of a golang array. I could have done it myself but it would have been tedious; I just pasted the list into GPT-4, asked it what I wanted, and out popped the result.

    Here's another place it was helpful recently; I prompted:

    "We're hiring a new community manager for the $PROJECT. I'd like you to help me develop some interview questions. What information to you need about the role to help you develop more useful questions?"

    The questions it asked me about the role were really good; just answering the questions was quite a useful exercise, and I'm sure the resulting document will be a good intro to whomever we hire. I wouldn't say the resulting interview questions were brilliant, but they were solid, and I used a couple of them.

    There are times when "the form" is there for a reason; if you want something re-written in a specific form, GPT-4 can do a good job. I wrote an email recommending something to somebody's managers in a different company in a different country; then I pasted it into GPT-4 and asked if it had any suggestions. It did a fair amount of rewording, of which I decided to take about half. In this case, the "polite form" is there to avoid offense, and it's exactly what I wanted.

    I've also asked it to write some Tweets highlighting some specific aspects of an upcoming conference I'm planning. It did a good job coming up with the sort of punchy, tweet-length messages which seem to do really well.

    Connecting it to the context of the article: My day job is basically arguing with people on the internet. :-). I do read and write all day every day; but I don't write messages where diplomacy is critical, nor do I write tweets. Perhaps I could get better at those, but I don't think it's worth the effort. Am I the worse off for that? Probably not in the way the author thinks; I don't think being diplomatically polished would change my thinking that much -- much less being able to write punchy tweets.

    If I started relying on it for the core writing, however, I'd certainly be selling myself short.

  • InfluxDB

    Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.

    InfluxDB logo
  • clownfish

    Constrained Decoding for LLMs against JSON Schema

  • https://github.com/newhouseb/clownfish

    Structural Alignment: Modifying Transformers (like GPT) to Follow a JSON Schema

  • axcell

    Tools for extracting tables and results from Machine Learning papers

  • Do you have a link to the repo? How does it compare to axcell (which is very domain-specific) [0].

    [0] https://github.com/paperswithcode/axcell

NOTE: The number of mentions on this list indicates mentions on common posts plus user suggested alternatives. Hence, a higher number means a more popular project.

Suggest a related project

Related posts