This is true of course and I don’t think these heavily subsidized plans will be around forever, but at the same time OpenAI is just less compute constrained than Anthropic right now as well so they’re in a stronger position to be able to offer these subsidies.
GPT models are also generally more token efficient right now and that helps too — you can go a lot further on a $20 subscription with Codex than Claude Code as a result of this.
Ultimately I think many day to day tasks just need to shift away from the latest frontier models towards models that are faster, cheaper, and still perform well enough & you can phase out subsidies while keeping total cost reasonable.
It’s gotten better within the last month or so but historically there’s been an excessive amount of anti-OAI and pro-Anthropic activity on this site as well and I’ve seen numerous posts get downvoted and almost instantly flagged for calling this out more politely than you have here.
So at least anecdotally I really don’t think it’s fair to portray this as OAI doing some sort of social media psyop as if others aren’t engaged in similar behavior.
It’s also very possible that this user just has opinions and tends to think OAI is more developer friendly / that Anthropic is hostile to developers (which is common sentiment I’ve seen from many real people who are definitely not paid OAI shills or something)
HN did a massive 180 in the last month or two, and nearly every post or comment related to Anthropic is just a hate post.
The amount of anger against Anthropic on HN doesn't reflect anything I see in reality (and I work at a pretty big FAANG with Codex and Claude Code, both are great) so I do suspect that OAI is doing some guerrilla marketing here, while Anthropic isn't really marketing or doing PR at all.
> I do suspect that OAI is doing some guerrilla marketing here, while Anthropic isn't really marketing or doing PR at all.
That is a very HN-minded comment.
Sure, there's probably some accounts that are more or less controlled by the big AI labs here.
But looking at how humans have been acting for the last 20 years, you'll see that you don't need to pay people to promote things.
They'll do it freely, because they identify with it and they can't fathom other people not agreeing with them.
Do you really thing that the weekly posts about people dropping AWS for Hetzner are paid by the German company?
No.
People have limited time and money.
Some picked Claude, others picked Codex.
Claude seems to be the most popular in terms of content produced about it.
So some people probably picked Codex just because they don't want to be like everyone else.
Then they obviously have to talk down about Claude, because if Codex is not better, then they are not.
Simple.
And from my POV that's not a good thing because HN was the place where people didn't act like this.
It was pragmatism and honest debate.
Now it's becoming: my agent is better than X, my stack is better than Y...
on twitter it is pretty clear that openai employees engage in coordinated messaging in a way that I haven't seen from other frontier labs. i say that as someone who prefers codex/gpt-5.5
Honestly I expect it's just annoyed devs getting annoyed about the ratelimits on plans and post-hoc justifying. Now that Codex has far more capacity and their slot machine makes better outcomes (note: I am a heavy LLM-assisted coder) they feel like they have to justify their felt animosity towards these companies
Personally i've just been using Claude Code with a coding agent UI (vibe-kanban) that has wrapped over "claude -p" for more than half a year without problems. I'd only been coding interactively and well within the terms of their subscription plan. I'm not even that much of a heavy user, I'm only hitting 10-40% of my weekly quota on a given week, and I basically only use the subscription outside of what Anthropic considers peak hours.
And then I got caught in the collateral damage a few days ago when Anthropic announced changes to their subscription plan billing, just like every other user of that tool and similar tools like Conductor and Zed. So in a month I won't be able to use my Claude sub quotas for these tools, all because some other people are ruining it for everyone by using "claude -p" to run openclaw, hermes agent and autonomous dark factories that burn billions of tokens a day.
I would have been fine with the change, except Anthropic's messaging was very slimy. They tried to spin their change as a positive change even though it was clearly not for anyone who was using a "claude -p" wrapper over Claude Code for better UX. They're within their rights to change their subscription billing, but they still couldn't be honest to their own users about it. Evidently, this kind of gaslighting and PR stunts is something they've done over and over in the last few months. It just didn't impact me until this time.
I care about AI safety and it would take a lot for me to switch from Anthropic to OAI, but I just wish they were less arrogant and cared about their users more. Right now their behavior is at best selfish (or overly consequentialist, and I don't mean that in a good way), and at worst actively hurting their AI safety efforts by pushing people to open-weight model alternatives which are way more dangerous than closed models due to people being able to remove their safeguards easily.
> They tried to spin their change as a positive change even though it was clearly not for anyone who was using a "claude -p" wrapper over Claude Code for better UX.
I feel like they were always fairly consistent (at least since OpenClaw came out) that wrapping claude -p in a non-Claude Code harness is disallowed by the subscription and requires using the API.
The lock-in to Claude Code is the price you pay for the subsidized tokens. If you don't want lock-in, that is what the API is for.
If you’re letting Claude code just handle secrets like this you’re already fucked from a security standpoint so I don’t really see the big deal here
Today it was the Vercel plugin but if you’re letting an LLM agent with access to bash and the internet read truly sensitive information then you’re already compromised
Israel has a disproportionately large amount of tech companies for its size and he took one photo with their leader.
I have no idea why everyone on the internet wants to endlessly seethe about this & personally attack Guillermo for it as if he’s endorsed their foreign policy or something
I would switch to Cursor 3 in a heartbeat if it supported Claude Agent SDK (w/ Claude Max subscription usage) and/or Codex the way that similar tools like Conductor do
And I would happily pay a seat based subscription fee or usage fees for cloud agents etc on top of this
Unfortunately very locked into these heavily subsidized subscription plans right now but I think from a product design and vision standpoint you guys are doing the best work in this space right now
If things continue to get worse I really worry how many people might give up on life entirely. A lot of people in this industry don’t have a whole lot else going on for them, myself included.
I grinded my 20s away trying to have a successful career and if that just gets pulled out from under me I’ve got absolutely nothing.
I know a very senior engineer who took their life the day after Trump was elected. He was unemployed for a while.
While I think a lot more was going on with him than being unemployed, I'm convinced AI hitting the scene had a bit to do with it. They were an older dev 50+.
These are kind of unrelated issues. You’re right that it used to be companies just didn’t want to be involved in war at all, & generally speaking that isn’t going to cause issues.
The core of the issue here is having a private company which is trying to dictate terms of use to the military, which is not really something that has been done before afaik
Originally this contract was signed with these terms included, and it wasn’t until Anthropic started investigating how its tech was used by Palantir in the Maduro operation that this became an issue.
On a surface level it seems like Anthropic is doing the right thing here but this is really at the root of this & the outcome of the case (and whether or not Anthropic is a legitimate supply chain risk) depends entirely on the details of those conversations they had with Palantir.
They would never do this because the entire point of the company is to try and control what AI is allowed to do, who is allowed to use it, and what they’re allowed to do with it. The overarching philosophy of Anthropic is explicitly opposed to open models. If it were up to them it would be illegal to inference them in the U.S.
There’s plenty of straightforward reasons why OpenAI would want to do this, it doesn’t need to be some sort of malicious conspiracy.
I think it’s good PR (particularly since Anthropics actions against OpenCode and Clawdbot were somewhat controversial) + Peter was able to build a hugely popular thing & clearly would be valuable to have on the team building something along the lines of Claude Cowork. I would expect these future products to be much stronger from a security standpoint.
I suspect Anthropic was seeing a huge spike of concurrent model usage at a too fast of a rate that claude code just doesn't do, CC is rather "slow" at api calls per minute. Also lots and lots of cache, the sheer amount of cache that claude does is insane.
It’s hard to say exactly what prompted the decision but they banned people paying $200/mo without warning & without any reasonable appeal system in place. It’s a Google form that is itself reviewed by some automated system that may or may not ever get back to you.
This was already an ongoing issue prior to 3rd party tools using Claude subscriptions, there are reports of false positive automated bans going back for several months.
I have not seen or heard of this happening w/ Codex, and rather than trying to shut down 3rd party tools that want to integrate with their ecosystem they have worked with those projects to add official support.
I’m more impressed with Codex as a product in general as well. Their new desktop app is great & feels an order of magnitude better than Claude’s.
Overall HN crowd seems heavily biased in favor of Anthropic (or maybe just against OpenAI?) but IMO Anthropic needs to take a step back and reset. If they keep on the current path of just making small iterative improvements to Claude Code and Claude Desktop they are going to fall very far behind.
GPT models are also generally more token efficient right now and that helps too — you can go a lot further on a $20 subscription with Codex than Claude Code as a result of this.
Ultimately I think many day to day tasks just need to shift away from the latest frontier models towards models that are faster, cheaper, and still perform well enough & you can phase out subsidies while keeping total cost reasonable.