2,155 karma · joined April 8, 2018
Furthermore, within China, alternative conspiracy theories claiming that the virus came from the US were allowed to circulate. A significant number of people in China actually believe that theory.
These combinations are questionable and very easy to filter, probably with Luna/Haiku tier of models that are able to tell things might get fishy here, and it would likely escalate to heavier checks and trigger KYC or straight banning.
Those are only my guess and probably aren't how the system works, but I think these rules are fairly easy to come up with for developers who have any idea of anti-abuse. I've seen too many Chinese posts mourning their accounts and communicating that their setups there would be similar mechanisms, I would say.
In previous months, there was news that Claude Code uploading a special signal for the Chinese timezone is pretty evident. I probably got away from having serious insomnia, using PST on my computers, and exclusively speaking English with those models lol.
Especially for Claude because Anthropic is very good at identifying mainland Chinese and getting them banned in hours. There are many of them who are willing to pay more than the original rate for a stable experience.
It's very hard for them because they'll need a legit phone number and bank cards that are not issued in China, and a clean enough IP, etc. and those better match together to make sense. (Back in the day, ChatGPT required resident IPs, which made it worse, but they worry about growth more now). Obviously, they have to use a VPN to access the real Internet, and most of the IPs they can find are shared with bots and abusers.
It's an art to come up with a great test suite that covers just enough and minimizes overlap, not only survives but also helps with refactoring.
It's essentially punishing players for playing well, and often could be gamed by playing badly intentionally.
Theoretically and conceptually I agree. But in practice there are a lot of programming languages aren’t as expressive. People prefer codebases with duplications rather than visitor patterns everywhere. In essence, visitor pattern is a tool to solve multi-dimensional abstraction problems, just like type classes in Haskell or CLOS in Common Lisp. But it’s so verbose and non-straightforward so more often than not it’s not worth it even conceptually it’s a legit case for “single source of truth”.
The current implementation isn't command-line, but a re-implemented GUI in disguise, awkward and even more buggy. Why should I use that over a GUI? I would prefer Electron over those TUI unless I have to SSH.
A great QA can understand the features of a product quickly, turn those concepts into some sort of grid or matrix in their mind, then pull a bunch of paths and scenarios with estimated priorities and probabilities at a fast and efficient pace, all with great coverage. They can also identify features contradicting each other more quickly than product people.
I think a good QA is capable of being a great vibe coder nowadays, too. If you can write great test suites (write names only), agents nowadays are able to turn those specs into decent codebases. Comparatively, I know a lot of decent dev having not very good taste in testing, who often write overlapping tests or missing important paths.
Source: trust me bro
However, when I tried out the SuperPower skill and had multiple agents working on several projects at the same time, it did hit the 5-hour usage limit. But SuperPower hasn't been very useful for me and wastes a lot of tokens. When you want to trade longer running time for high token consumption, you only get a marginal increase in performance.
So people, if you are finding yourself using up tokens too quickly, you probably want to check your skills or MCPs etc.
[0] https://openai.com/index/our-approach-to-advertising-and-exp...
Probably because it's a legacy and disappearing slowly? Modern Mandarin only has four tones left and has already lost tone patterns.
Do you know there's a "robot tone" in Chinese? It's simply swap every character to the flat or the first tone. Though it's under the stereotypical false assumption that robots have troubles with tones, kids in the late last century often communicated in that tone for fun without issues.
At the end of the day, vocal Chinese is always ambiguous with or without tones and in practice heavily relies on context. It requires written language to truly fix that.
Even though OpenAI has a lot of cash to burn, they're not in a good position now and getting butchered by Anthropic and possibly Gemini later.
If any major player in this AI field has the power to do it's probably Google. But again, they've done the Flutter part, and the result is somewhat mixed.
At the end of the day, it's only HN people and a fraction of Redditors who care. Electron is tolerated by the silent majority. Nice native or local-first alternatives are often separate, niche value propositions when developers can squeeze themselves in over-saturated markets. There's a long way before the AI stuff loses novelty and becomes saturated.
On MacOS is much better. But most of the team either ended up with locked in Mac-only or go cross platform with Electron.
Even for non-Mandarin/Guanhua, such as the Shanxi dialect, I can understand them because the pronunciation is much closer to mine, just the tones are completely novel.
Though, as a guy who speaks perfect mandarin from Beijing, I’m struggle even to pass the easy ones… So it can definitely used some improvements. The example 你好吃饭了吗 returns hào → hǎo, fān → fàn, le → liǎo. The first two are the model listen my tone mistakenly, and the last one should be le instead of liǎo in this context.
Also I see in the comment section people are worry about tones. I can guarantee tones are not particularly useful and you can communicate with native speakers with all the tones messed up and that’s perfectly fine. Because as soon as you leave Beijing, you’ll find all the tones are shuffled because of every region has their own dialect and accents, which doesn’t stop people from communicate at all. So don’t let tone stuff slow your learning process down.
I somehow kept the habit of handwriting for years. But as a guy in my early 30s, I do notice characters fade away from my brain from time to time, which wasn't a thing at all in the 20s. And to my surprise, some of the characters are fairly frequently used - I was just completely stuck when I was trying to recall them.
Probably that's how brains and organs peaked and will slowly break down over the following decades just like hard drives.
Also, if you consider latency, locking does not work well because client B might do operations before he/she even acknowledges the lock from client A because of latency.
The puzzle need to be solved is how to let people rejoin the workforce later without their career wrecked (with a discontinued CV).
The thing with Tao Te Ching is it's too ambiguous because: 1) The Chinese language is very overloaded and thus very ambiguous. 2) Classical Chinese is even more so. 3) Tao Te Ching is intentionally filled with clever puns which makes it more ambiguous.
The problem with translations is the translator has to interpret source texts into specific meanings in the target languages. It's like opening Schrödinger's cat box, or unwrapping monads in Haskell and Rust, which essentially deduct multiple possibilities into a single deterministic value.
If you're really into it, you probably want to learn some basic Chinese and classical Chinese (lucky they're not so different from each other), and figure out how to look up in the dictionaries. It's probably not as difficult as it sounds - all you need to do is decrypt with dictionaries.
Maybe there should be a new form of digital translation, just like hovering texts on Duolingo and it will display all the possible meanings of the word/expression.
Remember it in your brain: It's like let the state occupying one of the 16 registers in your brain. Later it will automatically offloads in the hard disk in your brain but there's a chance it gets lost or cannot be recalled reliably. Not recommend. But I do this more recently because I realised I don't have to do everything.
The stateless approach: Do it immediately so you don't need to bear the state/variable. Even though sometime it disrupts my current tasks, I find this approach is surprisingly relieving - less debts. Just like software engineering - minimise states because they're evil.
External storage approach: Write it down on paper or an app. There are trade offs between the tool you're using, but the key is to minimise the cost of your moves.
For pen and paper I tried different configurations until I can always comfortably carry them in my pocket.
For digital approaches I'm currently shovel things to Linear. Make sure you're fluent with shortcuts so you can create tasks and jump around like a breeze. I also use Arc browser and pin it in the third slot so I can jump to Linear with <Cmd-3> without even thinking about it.
Is this "PC" processor still aiming for data centres or desktops? It's not surprising at all if it's the former one.
The last time I attempted to run at full speed after not running for years, I struggled to keep up and lost my balance. I started tilting forward slowly and eventually fell then slide on the ground for a while, resulting in multiple scratches on my face, front pelvis, etc.
Before LLM, it's perfectly possible to spin up a SaaS on 5$ Digital Ocean VM and charge $4.99 per seat monthly. If you're using low overhead techs like Go and SQLite you might get away pretty far with a decent user base.
But LLM is inheirently costly compared to those traditional apps. No matter if you're calling OpenAI or DIY your own GPU cluster it's gonna be way more expensive. Spin your own GPU might ended to be more expensive because utilization problems and upfront costs.
The subscription model was kind of the silver-bullet for SaaS but it's probably not going to work well in the AI era.
OpenAI, Elevenlabs, Runway, and Midjourney: they have subscription model but the quota is strict and tight. The "unlimited" plan is simply pay-as-you-go.
Early wave of LLM products with unlimited subscription models like Github Copilot and Notion AI are probably pricing way too low. $7 or $10 is way too low to support heavy usage.
But charging $50 might scare most user away because it exceeded people's expectation for SaaS. And probably still end up losing money. And hobby users may ended up paying too much for the core users - that will lead us back to sophisticated pricing tiers like Elevenlabs and Runway.
Are there alternatives? I dunno. Maybe implement bring-your-own-key properly? Like OAuth but for LLMs? It's definitely interesting to see how things will turn out eventually.