New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Anthropic Temporarily Removed Claude Code from Pro Plan
https://news.ycombinator.com/item?id=47855832
https://bsky.app/profile/edzitron.com/post/3mjzxwfx3qs2a
Anthropic was seemingly removing Claude Code from new users who sign up on the Pro plan (which only affects 2% of new prosumer signups according to https://x.com/TheAmolAvasare/status/2046724659039932830).
These changes were then reversed (probably due to the backlash Anthropic received), but whether they'll retry later down the line is unknown.
What alternatives would you look at if you didn't have an AI lab / the hardware to run local models?
Comments
Uh, codex?
Ive seen people say Codex is pretty (brutally) shit, especially at frontend dev
Yeah. Codex is (arguably) shit at front-end. But, from my experience it's viable for back-end. So, make the design using claude (free tier possible unless you're making complex interface), then ask codex to wire it.
RIP Vibe Coding
Removing, adding, mish mashing. They are visibly going downhill. And with opus 4.7 and 1.35x token rate increase it is bad for them. Probably the first blow was code leak. And from there it just rolls down.
At new year, I hypothesized Claude will implement changes to reduce the burn rate or even making profit.
First, they barred people from using openclaw while using the subscription. I thought they want to protect the balance. Then, they doubled the allowance after the drama with US gov. Then, enable the peak adjustments for subscribers. Now, opus 4.7 isn't really much an upgrade compared to opus 4.6.
Therefore, without meaningful change in the trajectory. I conclude that the IPO means exit liquidity for angel and other kinds of big investor. For instance, Amazon is keen to realized the profit of their investment since late January this year.
There's a good change vibe coders will soon be searching for new jobs
I suspect they are doing this to free up capacity for enterprise, not because they are struggling.
From my experience codex is a lot more capable on frontend. Claude a bit better does on backend and planning, catches more edge cases. I always run codex on high or xhigh. Claude is slow af.
As usual, we use CharGBT.
It's more than capable than Claude Code.
https://github.com/Quarmire/ndn-rs/issues
All these issues were written by CharGBT.
Claude Sonnet and Claude Opus are no match for the complexity of these issues.
Their CI stayed red for days until the maintainer ran out of tokens.
I sell tokens. Who run out? I see they need help. 50% profit share don’t tell anyone
I'm testing Hermes Agent and it is better than codex.
Indeed. But Opus 4.6 cannot do well with GUI, either. I can only trust Gemini 3.1 Pro for the design.
GPT 5.4 high is way smarter than Opus 4.7 for me. I have fully switched from Opus to GPT.
Claude 4.6 Opus is good at writing CRUD code, or anything that is precise and deterministic. However, when it comes to some creative stuff, say, designing a dialect of SQL for a specific field, it just fails. GPT 5.4 high is capable of that tho.
I was planning on trying Claude Code sometime in May or June. If it's no longer available on the $20 plan, there's no way I'm splashing out $100 just to find out if it's suitable for what I want or not.
If they want to continue growing their market, I can see a lot of potential customers feeling the same way as me and just not bothering.
I'm probably canceling my Claude $20 plan at the end of the billing period and giving GPT a try. 5.5 is out now. I can use GPT with agents, which Claude can't do. 80% of people say that 4.7 is dumber than 4.6, and I've heard that GPT 5.4 was close to as good as 4.6. I run into my Claude session limits in like an hour most of the time, and people say that GPT is quite generous.
Can this be accessed effectively in Perplexity Pro?
No
Its tied to Claude.ai subscription
Zitron says the bubble bursts as soon as all the AI companies go to per-token billing.
I get great use out of my Claude subscription and personally greatly prefer them, but yeah, if I am forced to use API token billing I think I'd probably explore the open source models more, or try out ones operated by non-US companies.
ChatGPT 5.4 and 5.5 are definitely worth a shot. I'm more than happy with GPT Pro plan, and Tibo's resetting the quota all the time.
Its still available, they ran an A/B test where they removed access