@kunaljeweller I haven’t yet found a task /goal couldn’t do with good prompting and a bunch of time.
English
Edward Frazer
155 posts
Edward Frazer
@edwardcfrazer
Codex @ OpenAI | Prev. YC W24
San Francisco, CA Katılım Temmuz 2023
144 Takip Edilen192 Takipçiler
@jeffzwang Somewhat true but a bit off; I think (truly) unlimited token budgets have created new paradigms that haven’t fully emerged yet.
English
SF startups are actually ahead of the AI labs in AI adoption
AI labs have access to better models but there's more organizational friction to new AI-enabled operational patterns
Elad Gil@eladgil
People at major AI labs (using internal models) 3-4 months ahead of startup silicon valley engineers SV founders/eng 3-6 months ahead of NY NY founders/eng 6-12 months ahead of rest of world Most people have no idea how fast AI shifting as 1-2 years behind SOTA "The future is here, just not equally distributed" - Robert Heinlein
English
@_jmwebb @yacineMTB Goals make it far less likely the model will stop early. It minimizes the probability of “even if it’s wrong”.
I’m personally really really enjoying it — I can expect it to be done exactly the way I describe even if it takes 12+ hours.
English
@yacineMTB Why is goal any better than just giving it a task to complete?
English
@kunaljeweller Try using imagegen and having it clone the image! Works fairly well for me, though the latency sometimes hurts.
English
wow codex legit can't do frontend whatsoever, the leaderboards were correct. so it's codex for code and claude for design. i'm trying low on codex before i give up to see if it's for some reason able to handle frontend
English
@TheRohanVarma PMF for the in-app browser might be in gaming.
Ive seen a *bunch* of people working on games on my TL the last few days.
$100 plan will take them far
English
Today I’m attempting to hit the 5-hour limit on the Codex $100 Pro plan.
The method: build a MapleStory-like game from scratch.
30 minutes in, I already have a working game with sprites, maps, and assets generated with Imagegen.
Unfortunately, I’ve only used 5% of the limit so far.
At this pace, I may need to start building RuneScape in parallel just to make a dent 😬
English
@edwardcfrazer @GregKamradt but will it continuisly type @.codex review and iterate on the feedbacks until codex reviewer says no more major issues found?
English
When OAI employees say, “I let codex run all night…”
What framework they use?
How do you set up the task so it has enough work for 8 hours?
English
@AmirAnvarov @GregKamradt Just tell Codex to post @.codex review for you and read the github comments for responses.
English
@edwardcfrazer @GregKamradt "For example, over the weekend I had Codex watch over 8 PRs until all passed CI and had no valid P1 issues from Codex review." can u please share how did you achieve that? i'm currently annoyed that i have to ping @ codex review and than paste it's output back to Codex to fix
English
@aidenybai If using Codex, have it check the session history in .codex!
It should have its edit tool calls and hopefully reconstruction is easy.
English
@AuspiciousQE We're trying to make Codex the best place to do work for all knowledge workers!
I think that might be competitive with Cowork, but that's not how we think of it day to day :)
English
@edwardcfrazer Hey Edward - I saw the ad for 5.5, and it looks like it has the ability to work with files on your system now. Is this the cowork competitor via Codex?
English
I’m so excited about this!
Helping to build features like the in-app browser has been *so* fun, and that effort has transformed the way I work internally.
I basically never leave the Codex app for anything. It’s the best way to build software.
OpenAI@OpenAI
Codex for (almost) everything. It can now use apps on your Mac, connect to more of your tools, create images, learn from previous actions, remember how you like to work, and take on ongoing and repeatable tasks.
English
Edward Frazer retweetledi
Introducing GPT-5.5
A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done.
Now available in ChatGPT and Codex.
English
@thisisjonc Do you use the app or CLI? How can we make it better for you?
English
Codex on the $100/month plan on Extra High uses just 10% per day - that's completely reasonable.
What's not reasonable?
Any of those token-gobbling Claude regressions that Anthropic puts out like 4.7.
I'm fully sold on Codex now so long as OpenAI doesn't drop the ball.
English
@NicolasZu So happy you’re enjoying the in app browser!!
How could we make it even better?
English
Ok this is a truly insane NEW way to design for me, WITHIN Codex
If you are doing dev on a web technology, watch below
In that video:
- I am playing my game within Codex (yes)
- I use a codex-made tool to design buildings (see some tweets below it's really powerful)
- I can ask Codex to iterate and the game changes WITHOUT refreshing
- I can point at UI elements
- I can take a screenshot
All WITHIN Codex
It makes iterating on design incredible
Nicolas Zullo@NicolasZu
Everyone... Froze the game to marvel a bit, I had to share it with you Still can't believe this game is 100% AI Engineered in Codex Running in web browser, with THOUSANDS of zombies continuously trying to overcome your perfectly optimized factory Just insane!!
English
@rohitdotmittal What about Claude’s workflow would you like to see in Codex?
English
I need something with Claude's workflow, but Codex model.
I think Claude is too sycophantic and constantly changes their answers. Since they nerfed reasoning, the quality of the output has also decreased.
I need a model that thinks deeply and gives me a holistic response that it stands by.
English
@nickbaumann_ Able to explain how? Even on a high-level.
English
English
