Gabe

@reach_vb Codex for Gabe?

Codex for Open Source Codex for Students Codex Ambassadors Codex Security What should we do next?

@DavidKPiano The dryer ate the first one.

Everyone's talking about SOC-2 Do I have to see the first SOC to understand it?

@thsottiaux Aren't we all students at this point?

Codex is for students. It is what I wish I had 10 years ago, and it's hard to imagine a world without this technology now.

SOTA models, but I’m still choosing variants and reasoning levels. Why isn’t this automatic? 🤔

@minimaxir Those are for you to give to the stake holders / project managers. You're welcome... 😂

What is the purpose of agents giving planning estimates in days/weeks/months? I'm just going to ask you to do it and it'll only take an hour, tops.

When will we learn to stop telling everyone how good a model is. Keep it to yourself please, you're ruining throughput. 😅

@thdxr Sounds like a great idea to me... 🙄

ai password manager is that anything?

@mark_k Poor em dash it had a good run.

I think X might nerf posts containing emdash characters. Today I made a post that contained one emdash, and the post tanked. Then I deleted the post, and posted the exact same thing again but without the emdash, and it did much better. Perhaps some kind of bot filter? 🤷‍♂️

@mark_k I was top .001% of their users last year. Cancelled because of the highly subsidized subscription plans. Recently returned and now i'm going to attempt to go back to it exclusively. This was before composer 2 was announced. x.com/glwillen/statu…

Are you going to give Cursor another chance, now that Composer 2 is here?

Been playing with a different mental model: Treat the LLM like a CPU WASM agents WIT as ABI Grammar as ISA Function calls as opcodes So decoding becomes execution Definitely just an experiment, not production ready It uses @liquidai LFM2.5 1.2B github.com/gabewillen/avm

Next month i'm going to cancel my 4 subscriptions to various coding agents that total to around $800 a month and see if I can keep the same velocity and success using only Cursor up to that limit.

So I signed back up for @cursor_ai today. I'm now convinced that @OpenAICodexCli by @OpenAI is throttling subscription users or deprioritizing their requests. Despite having the OpenAI pro plan, responses and throughput are noticeably better on Cursor.

@thsottiaux Pair this with codex spark and you get insane throughput x.com/glwillen/statu…

What are we consistently getting wrong with codex that you wish we would improve / fix?

@ah20im x.com/glwillen/statu…

What would you like to see in Codex?

I had 2 codex pro accounts staggered perfectly where one would reset when the others weekly usage ran out. But @thsottiaux had to get all trigger happy and reset the usages 3 times and now they both reset at the same time.

@DavidKPiano Decomposition is the second coolest term that solves the coolest term.

"State space explosion" is the coolest term for the worst thing that most developers do

@AnthropicAI Are you sure the double rate limits aren't counting towards the total weekly usage? My hourly session seems to be moving slower than the weekly usage limit.

@DavidKPiano @OpenAI You got your self an AI Ménage à trois!

Thanks @OpenAI! I'm going to make Claude Code and Codex chat with each other for the next 6 months

@thsottiaux @kr0der By negative you mean positive right?

@kr0der Negative

odds?

@thsottiaux It's been a while since you reset our usage. I think you should test that functionality again to make sure it's still working. 😅