HN user

ravroid

97 karma
Posts4
Comments36
View on HN

One strategy I've been experimenting with is maintaining a 'spec' document, outlining all features and relevant technical notes about a project. I include the spec with all relevant source files in my prompt before asking the LLM to implement a new change or feature. This way it doesn't have to do as much guessing as to what my code is doing, and I can avoid relying on long-running conversations to maintain context. Instead, for each big change I include an up-to-date spec and all of the relevant source files. I update the spec to reflect the current state of the project as changes are made to give the LLM context about my project (this also doubles as documentation).

I use an NPM script to automate concatenating the spec + source files + prompt, which I then copy/paste to o1. So far this has been working somewhat reliably for the early stages of a project but has diminishing returns.

GPT-4o 2 years ago

In my experience so far, GPT-4o seems to sit somewhere between the capability of GPT-3.5 and GPT-4.

I'm working on an app that relies more on GPT-4's reasoning abilities than inference speed. For my use case, GPT-4o seems to do worse than GPT-4 Turbo on reasoning tasks. For me this seems like a step-up from GPT-3.5 but not from GPT-4 Turbo.

At half the cost and significantly faster inference speed, I'm sure this is a good tradeoff for other use cases though.

I too have been noticing a deterioration in recent months. Things that generally worked smoothly before have random bugs now. For example, turning off shuffle and resuming a song will sometimes just stop playing Liked Songs after the song ends, shuffle will turn off/on seemingly at random, etc. Also there seem to be odd inconsistencies in the UX between mobile and desktop such as liking songs and seeing Smart Shuffle songs in playlists.

I've been generally very happy with the Spotify app for years and it's disappointing to see the quality slipping.

I've found that, at least on the mobile app, results are filtered by "Featured" which fills the results with irrelevant sponsored products. So each time I search for an item, I then have to go under Filters and select Best Rating, Highest Selling, etc. It's a bit tedious but seems to be a shortcut through all of the BS results they show you by default.

This is an excellent piece of advice.

Of the five companies I have worked for, the two where I was disconnected from leadership were where I felt the least valued (and consequently less motivation). Of the three where I worked directly with leadership, I was treated with significantly more respect and felt a much stronger sense of purpose.

That was the only raise I received unfortunately. But once I realized that the work I was doing produced the same value as salaried employees, I learned to use leverage to negotiate better compensation. I haven't waited around for a raise ever since.

When I first started out as a software dev for an agency, I was offered $9/hr. I took it because it was more than the minimum wage I was making at my restaurant job.

After a few months they rewarded me for my hard work by giving me a raise of $1. At the time I was grateful but now I realize what an insult that was.