I was getting tired of summarizing long articles & threads on HN/Reddit with ChatGPT so I made a simple little Chrome/Firefox extension to do it for me:
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.
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.
You're equating this to censorship, when I think it's more like Google adding security measures so you can't break their search engine rather than removing unfavorable results.
Cool concept. Not relevant to me in its current state being limited to skin care products, but would love to use something like this for things like supplements or other products where I otherwise have to sift through Amazon reviews & reddit threads.
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.
I think it's because software varies vastly in complexity and spans many domains. Every software engineering role may use a generalized skillset to a degree, but I feel like there can never be one all-encompassing bar that can effectively cover the requirements of every role.
I bought a cheap $60 or $70 drone on Amazon and eventually gave it a 3-star rating. I was hardly being critical of the product at all, just gave an honest review. The next day I received an email from the company offering me $100 Amazon giftcard to remove the review.
You made a great thing. It's so simple and effective.
I've read a handful of workout plans, considered hiring personal trainers, and sifted through YouTube videos without ever being able to establish a good workout routine. This is EXACTLY what I needed.
This is hilarious. Paying money to passive-aggressively insult somebody, I think you may have invented a new market. Has this product generated any revenue yet?
I use another GPT to turn that spec into a development plan for Codex that I include in AGENTS.md. (https://chatgpt.com/g/g-698a6ee58aec8191ba1e3b520b13b5e7-dev...)
I'm curious what advantages this product offers vs. using a prompt?