GateUser-ebc4af76

GPT-5.6 Sol's most unfortunate point this time:
It's not that it's weak, but that it can't be used yet.
Strictly speaking, this is more like a futures ticket.
OpenAI has already shown its cards, but because the US government hasn't approved it yet, the actual launch will have to wait a few more weeks.
But just looking at this card, Claude is under considerable pressure.
Sol's pricing is the same as GPT-5.5: $5 for input, $30 for output.
But it's no longer going up against GPT-5.5, but rather Claude's top-tier Mythos / Fable.
On Claude's side, it's $10 for input, $50 for output.
Sol is signific

GPT-5.6 is coming soon! The front-end JS files are already available
It might really be announced very soon!

Anthropic has finally released the safe version of the Mythos model, Fable-5.
Many details about the parameters have already been discussed online, so I won't repeat them.
What’s more worth noting is the real task testing done by a third-party programming tool, Augment Code.
This set of tests ran a total of 489 programming tasks, and the results are quite interesting:
Fable-5 clearly leads in overall performance and accuracy.
Overall score +0.224, accuracy +0.191, definitely the top tier so far.
But another detail is also very important: GPT-5.5 still comfortably outperforms Opus-4

An old account that Claude banned in 2024 actually got another chance for a re-review.
Hmph, they were so haughty when they banned me back then—now they’re just as painfully scrambling to make it right!
I knew this wave of Codex overtaking them was entirely their own doing.

Opus 4.8 is now live!
Tried it out, and my first impression is that it's really verbose.
Oh no, why does it look so much like GPT-5.4?

A bit shocked.
It is said that the new version of Claude Code may request camera permissions after the user reaches a certain limit, to verify that you are the person present, with the goal of preventing "programmatic use."
No, AI writes code, just writes code, and now it's even checking in on users...

Old Ye AI Transfer Station has been running for 6 days.
Today I checked the backend—the total number of registered users has already reached 5,800.
In the past few days, we’ve been adding roughly 1,000 new users per day on average.
Honestly, this pace is a bit beyond my expectations.
But even more unexpected is that I actually managed to keep it all going.
If we also count the first 10 days of internal testing, in these half a month, there basically hasn’t been a single day of normal sleep.
During the day, I handle registrations, subscriptions, plus-signing, and keep the plus pool

I have built an AI relay station, officially opening today:
When I first started this station, the idea was very simple.
Mainly because I myself consume a lot of tokens daily, and I have many friends who can’t always access stable Claude / GPT.
It’s not just subscription hassle, but also various restrictions; looking for other relay stations, I always worry about reliability and information security.
So I decided to set up my own.
Many people ask: why dare to use Lao Ye’s?
I understand this question. After all, with AI relay stations, what everyone fears most isn’t higher cost, but n

Regarding this rsETH incident, the progress of Aave's DeFi United donation campaign is worth continuous observation.
I casually created a Dune dashboard that can track the current donation status in real-time, including total ETH donated, USD valuation, number of participating wallets, and distribution across different donation amount ranges.
Dashboard link:
As of now, DeFi United has received 69,630.53 ETH, which is approximately $2,312.32 per ETH, totaling about 161 million USD; the number of participating donation wallets has reached 36,555.
Among them, institutional and large donat

Claude, are you out of your mind? To stop us, you’re starting KYC!
From now on, even the cleanest IPs are useless—you’ll have to take photos, upload your documents, and do video verification, and you’ll instantly reveal your true nature!
For those doing intermediation/agency work, it’s even worse—looks like the sky is falling! If they can’t complete KYC for so many accounts in the account pool, then even if the accounts are still “alive,” they’re as good as dead! Not only can they not withdraw funds, they also can’t use Claude Code.

Due to the global computing power shortage, after Claude, OpenAI is also discussing the introduction of a peak-valley billing scheme.
Some netizens suggested directly giving a 10x quota during idle times, and they are willing to adjust their biological clocks and change their sleep schedules for this.
OpenAI engineers responded that such aggressive past approaches could affect the sleep patterns of the entire country, which is not ideal!
What do you think would be a more reasonable final plan?