4 ms·

On slightly off topic note: Codex is absolutely fantastic right now. I'm constantly in awe since switching from Claude a week ago.

lifty 6mo ago
I made this switch months ago, ChatGPT 5.4 being a smarter model, but I’ve had subjective feelings of degradation even on 5.4 lately. There’s a lot of growth in usage right now so not sure what kind of optimizations their doing at both companies
- CamperBob2 6mo ago
  Agreed. Watching the intermediate "Thinking about X ... Now I'll do Y" text on GPT 5.4 lately has been like watching a hypothetical smart drug wear off. All of the major models have been getting worse lately, not just Opus.
  - philpem 6mo ago
    Makes me wonder if the output is starting to get back into the training input and we're seeing the first signs of model collapse.
    - CamperBob2 6mo ago
      Business model collapse, maybe. Can't wait, I need to buy some RAM for my local model server.
vidarh 6mo ago
Codex has been good quality wise, but I hit limits on the Codex team subscription so quickly it's almost more hassle that it is worth.
lores 6mo ago
I would switch to Codex, but Altman is such a naked sociopath and OpenAI so devoid of ethical business practices that I can't in good conscience. I'm not under any illusion that Anthropic is ethical, but it is so far a step up from OpenAI.
- simianwords 6mo ago
  [dead]
  - DonHopkins 6mo ago
    [dead]
  - emaro 6mo ago
    There's not one thing that stands out, but he abandoned the entire core principles of OpenAI (took a 180), constantly lies to people and doesn't plan to stop. https://www.newyorker.com/magazine/2026/04/13/sam-altman-may-control-our-future-can-he-be-trusted?currentPage=all https://www.newyorker.com/magazine/2026/04/13/sam-altman-may...
- nh2 6mo ago
  Cannot you use Codex (which is open source, unlike Claude Code) with Claude, even via Amazon Bedrock?
  - embedding-shape 6mo ago
    Codex with Anthrophic's models is not as good as using the models with the harness it was trained in mine for. Same goes vice-versa too.
- bob1029 6mo ago
  I'm with you on the ethical part, but everything is a spectrum. All the AI leadership are some shade of evil. There's no way the product would be effective if they weren't. I don't like that Sam Altman is a lunatic, but frankly they all are. I also recognize that these are massive companies filled with non shitty engineers who are actually responsible for a lot of the magic. Conflating one charlatan with the rest of it is a tragedy of nuance.
  - subscribed 6mo ago
    Yeah, but there's distinct difference between "risks their company because they refuse to help with killing little kids" and "happily helping with genocide". One of these is better.
toenail 6mo ago
I have also switched from claude to codex a few weeks ago. After deciding to let agents only do focused work I needed less context, and the work was easier to review. Then I realized codex can deliver the same quality, and it's paid through my subscription instead of per token.
onion2k 6mo ago
I use Codex at home and Opus at work. They're both brilliant.
yukIttEft 6mo ago
I'm currently "working" on a toy 3d Vulkan Physx thingy. It has a simple raycast vehicle and I'm trying to replace it with the PhysX5 built in one (https://nvidia-omniverse.github.io/PhysX/physx/5.6.1/docs/Vehicles.html https://nvidia-omniverse.github.io/PhysX/physx/5.6.1/docs/Ve...) I point it to example snippets and webdocumentation but the code it gens won't work at all, not even close Opus4.6 is a tiny bit less wrong than Codex 5.4 xhigh, but still pretty useless. So, after reading all the success stories here and everywhere, I'm wondering if I'm holding it wrong or if it just can't solve everything yet.
- lukan 6mo ago
  " or if it just can't solve everything yet." Obviously it cannot. But if you give the AI enough hints, clear spec, clear documentation and remove all distracting information, it can solve most problems.
  - shdh 6mo ago
    Most simple problems with plenty of prior art, sure
- seba_dos1 6mo ago
  It works somewhat well with trivial things. That's where most of these success stories are coming from.
  - shdh 6mo ago
    Exactly this, the SNR is polluted by this anecdata because someone was able to implement a CRUD backend they couldn’t before
- wg0 6mo ago
  Most of the folks are building CRUD apps with AI and that works fine. What you're doing is more specialized and these models are useless there. It's not intelligence. Another NFT/Crypto era is upon us so no you're not holding it wrong.
glerk 6mo ago
Codex/GPT5.4 is just superior to Opus4.6 for coding. I swear it costs me 1/2 of the tokens to achieve the same results and it always follows through the plan to completion compared to Opus that takes shortcuts and sweeps things under the rug until I discover them through testing. I'm not accusing anyone of foul play and I don't have financial interests in either company, but it feels like "something" within Code Claude/Anthropic models is optimizing to make you spend more tokens instead of helping you complete the task.

lifty 6mo ago

I made this switch months ago, ChatGPT 5.4 being a smarter model, but I’ve had subjective feelings of degradation even on 5.4 lately. There’s a lot of growth in usage right now so not sure what kind of optimizations their doing at both companies

CamperBob2 6mo ago
Agreed. Watching the intermediate "Thinking about X ... Now I'll do Y" text on GPT 5.4 lately has been like watching a hypothetical smart drug wear off. All of the major models have been getting worse lately, not just Opus.
- philpem 6mo ago
  Makes me wonder if the output is starting to get back into the training input and we're seeing the first signs of model collapse.
  - CamperBob2 6mo ago
    Business model collapse, maybe. Can't wait, I need to buy some RAM for my local model server.

vidarh 6mo ago

Codex has been good quality wise, but I hit limits on the Codex team subscription so quickly it's almost more hassle that it is worth.

lores 6mo ago

I would switch to Codex, but Altman is such a naked sociopath and OpenAI so devoid of ethical business practices that I can't in good conscience. I'm not under any illusion that Anthropic is ethical, but it is so far a step up from OpenAI.

simianwords 6mo ago
[dead]
- DonHopkins 6mo ago
  [dead]
- emaro 6mo ago
  There's not one thing that stands out, but he abandoned the entire core principles of OpenAI (took a 180), constantly lies to people and doesn't plan to stop. https://www.newyorker.com/magazine/2026/04/13/sam-altman-may-control-our-future-can-he-be-trusted?currentPage=all https://www.newyorker.com/magazine/2026/04/13/sam-altman-may...
nh2 6mo ago
Cannot you use Codex (which is open source, unlike Claude Code) with Claude, even via Amazon Bedrock?
- embedding-shape 6mo ago
  Codex with Anthrophic's models is not as good as using the models with the harness it was trained in mine for. Same goes vice-versa too.
bob1029 6mo ago
I'm with you on the ethical part, but everything is a spectrum. All the AI leadership are some shade of evil. There's no way the product would be effective if they weren't. I don't like that Sam Altman is a lunatic, but frankly they all are. I also recognize that these are massive companies filled with non shitty engineers who are actually responsible for a lot of the magic. Conflating one charlatan with the rest of it is a tragedy of nuance.
- subscribed 6mo ago
  Yeah, but there's distinct difference between "risks their company because they refuse to help with killing little kids" and "happily helping with genocide". One of these is better.

toenail 6mo ago

I have also switched from claude to codex a few weeks ago. After deciding to let agents only do focused work I needed less context, and the work was easier to review. Then I realized codex can deliver the same quality, and it's paid through my subscription instead of per token.

onion2k 6mo ago

I use Codex at home and Opus at work. They're both brilliant.

yukIttEft 6mo ago

I'm currently "working" on a toy 3d Vulkan Physx thingy. It has a simple raycast vehicle and I'm trying to replace it with the PhysX5 built in one (https://nvidia-omniverse.github.io/PhysX/physx/5.6.1/docs/Vehicles.html https://nvidia-omniverse.github.io/PhysX/physx/5.6.1/docs/Ve...) I point it to example snippets and webdocumentation but the code it gens won't work at all, not even close Opus4.6 is a tiny bit less wrong than Codex 5.4 xhigh, but still pretty useless. So, after reading all the success stories here and everywhere, I'm wondering if I'm holding it wrong or if it just can't solve everything yet.

lukan 6mo ago
" or if it just can't solve everything yet." Obviously it cannot. But if you give the AI enough hints, clear spec, clear documentation and remove all distracting information, it can solve most problems.
- shdh 6mo ago
  Most simple problems with plenty of prior art, sure
seba_dos1 6mo ago
It works somewhat well with trivial things. That's where most of these success stories are coming from.
- shdh 6mo ago
  Exactly this, the SNR is polluted by this anecdata because someone was able to implement a CRUD backend they couldn’t before
wg0 6mo ago
Most of the folks are building CRUD apps with AI and that works fine. What you're doing is more specialized and these models are useless there. It's not intelligence. Another NFT/Crypto era is upon us so no you're not holding it wrong.

glerk 6mo ago

Codex/GPT5.4 is just superior to Opus4.6 for coding. I swear it costs me 1/2 of the tokens to achieve the same results and it always follows through the plan to completion compared to Opus that takes shortcuts and sweeps things under the rug until I discover them through testing. I'm not accusing anyone of foul play and I don't have financial interests in either company, but it feels like "something" within Code Claude/Anthropic models is optimizing to make you spend more tokens instead of helping you complete the task.