Can't wait until somebody makes a game hit in Steam using unscii as every UI in the game.
1,902 karma · joined June 15, 2014
They are still training with less than 1/10 of what the US companies do own, and yet having similar results.
China is killing it.
Some other people has to grind harder and even be better than you to get half of your success, that doesn't mean that they are wrong, or that the book is wrong.
I believe a lot, if not all advice there in the book is necessary. Other people might not work at Google, but as I've said before, might need to grind different gears in order to be successful, if you don't -- good for you!
A lot of your suggestions would get you fired very quickly on many companies, it's good that it all works for you.
With its pointy edges, even in a very slow accident hitting a pedestrian, the outcome will make any Tarantino movie look soft, in terms of blood being spilled around.
Don't even get me started on those huge American cars, they are the absolute terror in terms of pedestrian safety.
But I have a hard time believing ruby-core will want to hear community feedback... people have been talking about this for ages... Ruby is omakase?
RBS and Sorbet suck. One is very limited, the other isn't part of ruby-core and makes you rewrite the function arguments again, similar to Java's annotations... Doesn't look like Ruby at all, or DRY, mostly like a workaround!
LowType is what it should have been -- hard to believe we are in 2025 and we still don't have a decent, programmer-friendly solution in ruby-core.
Meanwhile Python has it right since a long time. No wonder it is so stagnated with people going for other stacks.
Ruby is slowly becoming what Perl did, a very niche language.
First, lots of server-side code is IO-bound, writing it in Rust vs. Java/C# would barely show any difference in a Monitoring tool, in a real-life scenario.
His authorization system is very limited in scope, of course it can be fast! Get real users and we will see if that will still be fast.
When you are running it in production, even if using Zanzibar's approach of loading everything into memory, you'd still need to handle many aspects he didn't think of, like updates to such permissions, and dealing with sharding etc. Things are always more complex in real life.
And last not but the least, Notion is really fast as it is. I never knew it was slow.
Without bringing any new concept to "Notion", I find it hard to believe this will ever work.
I hope he finds happiness building it though, building is fun!
Wish we'd see more action on RBS, perhaps getting Sorbet to core or something. At least having some sort of consensus in the community.
Or even some considerable project in the JIT, dunno!
Python is definitely ahead in types, work on removing the GIL... list is huge.
No suicides created by ChatGPT Today. Billions of dollars in GPU will sit idle. Sudden drop of Linkedin content...
World is a better place
We have to spend a good amount of time organizing the requirements, as it never comes perfectly from product management, as well alignments and weighting tradeoffs from the architecture.
Reviewing code also became a very daunting and time-consuming task.
Python and others have similar issues, with them having limitations as well
Which use-cases needing Qt WebKit is an issue?
Of course, you'd still need interfaces for DI, mocking etc, but that's pretty most of my interface usage.
I also write lots of Typescript, and the furthest I go is to use 'Omit' and other utility types, but already feel like it's too much.
It's also well-suited for that. Of course, you won't end up with a tiny Go docker image, but this doesn't matter.
Of course, if you expect a full FE+BE 'omakase' framework like Rails there isn't anything with the same weight. I began to see this as a plus, you actually don't need it all, and nowadays it's very modern to delegate auth to a service etc. I know it isn't DHH's PoV, but it makes it much easier to maintain, so you focus on writing business logic and do the FE in a widely supported framework like React, or use Microsoft stuff, your choice.
The DTOs/DI and the typical .NET developer stuff isn't bad or hard to learn, most of it comes naturally when you think "What would a statically typed language need?"
It's what allows C# code be very clean and easy to follow, where you know exactly what is available unlike Ruby that a lot of things are implicit and can get very nasty. After so many years debugging and improving Rails apps performance, I got sick of it and C# feels fresh.
Then there's LINQ and a lot of language sugar that makes C# code really beautiful. I've done also some Java, and can easily vouch for C#. It's the Ruby of statically typed langs.
And the speed, don't get me started. It's so fast.
Why C# doesn't have first class functions and can't go fully functional (and likely never will): Scala tried and its compiler is slow, even after so many iterations and new novel compiler ideas.
I like the way it is, and hope it doesn't change. Unless they could make this possible without making the compile process extremely slow.
C# is a great language, it's now very modern and has the best parts of Typescript, while leaving out the bad ones.
It's also extremely fast and multi-platform.
It also doesn't have the fragmentation that Java or JVM langs has.
And it's also open source nowadays. I think Sillicon Valley hasn't caught up with those recent changes, I bet more startups would be using C# if they knew.
I hope Microsoft delivers though, we need more alternatives to React + RN or Flutter.
And also it has extreme limitations that only world models or RL can fix.
Meta can't fight Google (has integrated supply chain, from TPUs to their own research lab) or OpenAI (brand awareness, best models).
The US had 6.3% GDP deficit in 2024, meanwhile it's 3% in the EU. The US is growing on debt, and on its way to become Japan.
Even on EU's worst debt crisis times, it didn't run such a big deficit as the US.
Add in the fact that the EU has accepted the Meta/Google tax in advertising, among other US' interests businesses. Which it can change its mind and get such businesses in trouble.
Not to forget about shale oil. The US became the biggest oil producer BY A WIDE MARGIN.
But it's all based on a $60 barrel, if the oil barrel goes back to $40, you'd see most of those companies filing bankrupcy and trillions disappearing. Here's another industry that the US government will need subsidize further with more US debt and taxpayers money.
There's a lot of risk attached to the US' recent growth, people would be naive to not consider them. We'd have to see if that trend lives for longer, and what are the tradeoffs.
The EU could decide to also destroy its country, feed people a lot of corn syrup and diabetes, and provide them GLP-1 drugs to increase its GDP.
Or just make the european believe that big cars are great, and they need a 5 bedroom house for 3 people.
All this would push that holy productivity number UP, but does it make rational sense to do that? Nope.
Waymo can just add the cameras exactly the way Tesla has, and train based only on that information.
Now it has tons and tons of data, they could gradually remove the Lidar on cities that they've driven over and over again. IF driving without Lidar is worth it... maybe it isn't even worth it and we should pursue using Lidars in order to further reduce accidents.
Meanwhile people use Tesla sporadically in a few spots they consider safe, they will always have data that isn't useful at all, as it can already drive on those spots.
--
Another thing, we can definitely afford to have Lidars on every car, if that would make our cars safer.
Imagine if China does a huge supply chain of Lidars, I bet the cost would be very tiny. And this is supposing there aren't any more automation and productivity gains in the future, which is very unlikely.
Lidar production just doesn't have that big scale, because it's a very tiny market as of now. With scale, those prices would fall like batteries and other hardware have fallen with the years.
TLDR: Tesla lost badly
I'm from a western country and I agree with your statement and have a similar fear. My country is doomed because of the pension system.
BUT this doesn't apply to China. Their system isn't structured this way, therefore this is mostly irrelevant for them.
As long as China have a working population bigger than most countries, as well such amazing universities, they will perform better than all those countries.
Even with the population decline, they'll still have more able workers than all western countries for at least the next 100 years.
Let that sink in.
It isn't worth it, because those that already using it, will likely pay or migrate to a different solution.
I wasn't a .net dev before, so I was surprised that it 'abstracts' RabbitMQ, Azure Service Bus etc. This doesn't exist in the 2 other languages I've done a lot of work with -- even though we also have message bus. This is such a useless feature!
There's no need to use it, you can just use RabbitMQ with its original package, or Azure Service Bus etc.
People used it because it was famous, not because it was stellar in some way. There are so many other alternatives.
For O11y or other features, those are easily implemented or already even existing if you use Datadog / something using Opentelemetry.
So as I've said before, it's useless. It's something worth maintaining only for the people charging for it and those that are stuck with it, new projects will use something else.
Tell it to create an API for the LLM data ingestion, then integrate with it on your software.
BTW, this is far from what an LLM Observability tool will offer you. You are a bit confused what O11Y is.
This will let you tune your application being run.
Also, this might change in a time of burst traffic, and if you don't have such tools like Prometheus, DD etc, you aren't able to tune accordingly.
The thing is that tuning a production setup is a bit of an art, there are many tradeoffs for what you do (typically, cost vs. benefit), so you need to make those decisions yourself.
If the alert is constantly ringing and you are satisfied with how the system is running and tradeoffs, you should disable it.
They are way ahead in terms of execution from OpenAI.
How powerful the LLM model is, is a bit irrelevant for that usecase. The most important part is well integrated it is, how it fits certain user desires (anime, real life, kinks).
This doesn't seen to be something a big company like OpenAI would be great at executing upon.
To me, that's a huge bubble sign. This is a small market. Having OpenAI wanting to do something there makes me thing they are running out of usecases.
- Open source LLM models at most 12 months behind ChatGPT/Gemini; - Gemini from Google is just as good, also much cheaper. For both Google and the users, as they make their own TPU; - Coding. OpenAI has nothing like Sonnet 4.5
They look like they invested billions to do research for competitors, which have already taken most of their lunch.
Now with the Sora 2 App, they are just burning more and more cash, so people watch those generated videos in Tiktok and Youtube.
I find it hilarious all the big talk. I hope I get proven wrong, but they seem to be getting wrecked by competitors.