HNHacker News
TopNewBestAskShowJobs

thanhhaimai

2,530 karma · joined January 16, 2012

submissionscomments
thanhhaimai··on Claude Sonnet 4 now supports 1M tokens of context
I work across the stack (frontend, backend, ML)

- For FrontEnd or easy code, it's a speed up. I think it's more like 2x instead of 3x.

- For my backend (hard trading algo), it has like 90% failure rate so far. There is just so much for it to reason through (balance sheet, lots, wash, etc). All agents I have tried, even on Max mode, couldn't reason through all the cases correctly. They end up thrashing back and forth. Gemini most of the time will go into the "depressed" mode on the code base.

One thing I notice is that the Max mode on Cursor is not worth it for my particular use case. The problem is either easy (frontend), which means any agent can solve it, or it's hard, and Max mode can't solve it. I tend to pick the fast model over strong model.

thanhhaimai··on Cerebras Code
The issue is not about whether the limit is too high or too low. What turned me back was that they claimed "no weekly limits" as a selling feature, without mentioning that they change it to a "daily limits".

I understand it's a sale tactics. But it seems not forthcoming, and it's hard for me to trust the rest of the claims.

thanhhaimai··on Cerebras Code
> running at speeds of up to 2,000 tokens per second, with a 131k-token context window, no proprietary IDE lock-in, and no weekly limits!

I was excited, then I read this:

> Send up to 1,000 messages per day—enough for 3–4 hours of uninterrupted vibe coding.

I don't mind paying for services I use. But it's hard to take this seriously when the first paragraph claim is contradicting the fine prints.

thanhhaimai··on Animated Cursors
For a demo of the new Cursor feature in Ghostty: https://www.youtube.com/watch?v=enwDjM7pNNE
thanhhaimai··on Electric cars produce less brake dust pollution than combustion-engine cars
I'm not sure I can agree with this. I have 2 EVs, and the tire looks almost brand new after like 10k miles. I think the driving habits matter more than whether the car is EV or not.
thanhhaimai··on The IRS Is Building a System to Share Taxpayers' Data with ICE
In case you're serious about the question, the answer is a yes.

Emphasis mine.

> Seventy-five percent of the prisoners in U.S. Marshals custody are detained in state, local and _private_ facilities; the remainder are housed in Federal Bureau of Prisons (BOP) facilities.

https://www.usmarshals.gov/what-we-do/prisoners/operation/cu...

Here is another source:

> The U.S. Marshals Service detains about 60,000 people a day who are awaiting federal trial or sentencing. While it doesn't operate jails, it does partner with public and private detention facilities. The service assesses detention conditions at these facilities.

> We found shortcomings in its oversight. For example, many deputies who reviewed state and local facilities hadn't received required training. Additionally, some facilities didn't meet some standards for 3 years in a row, including food safety standards.

https://www.gao.gov/products/gao-24-106348

thanhhaimai··on Major reversal in ocean circulation detected in the Southern Ocean
I think you overestimate how rich the bottom 70% of the world is.

The bottom 70% of the world's population would have less than $X00 in the bank, and wouldn't have much control over their lifestyle.

thanhhaimai··on Copper is Faster than Fiber (2017) [pdf]
High Frequency Trading is one.
thanhhaimai··on Microsoft extends free Windows 10 security updates into 2026
The only reason for me to ever boot into Windows is for gaming. As more and more games support SteamOS, it's increasingly tempting to just drop Windows altogether.
thanhhaimai··on GCP Outage
The status page is green, but there are outages reported: https://downdetector.com/status/google-cloud/
thanhhaimai··on Did "Big Oil" Sell Us on a Recycling Scam?
Another concept comes to mind: Identity Theft.

It's a crime from a criminal to scam the bank. Instead of the bank being held accountable for mistaking the identity (or don't do enough due diligent), the concept shifts the blame to the account owner, saying it's their fault for having their identity "stolen", despite them was not involved in the scam process.

thanhhaimai··on DumPy: NumPy except it's OK if you're dum
For solo works, the terseness might work, but usually only in short term. Code I wrote 6 months ago looks like someone else's code. For team work, I'd prefer to be explicit if possible. It saves both my teammates' time and my time (when I eventually forget my own code 6 months from now).
thanhhaimai··on Wikipedia's Most Translated Articles
> Your pessimism doesn't seem to be based on any kind of facts, unless I'm missing something here

Yes, the slow ramp of anti Wikipedia rhetoric is a documented fact, in both Russia and the USA: https://gizmodo.com/trump-doj-threatens-wikipedias-nonprofit...

thanhhaimai··on xAI's Grok 3 comes to Microsoft Azure
It refers to this: https://www.reuters.com/markets/deals/musks-xai-buys-social-...

> "xAI and X's futures are intertwined," Musk, who also heads automaker Tesla and SpaceX, wrote in a post on X: "Today, we officially take the step to combine the data, models, compute, distribution and talent."

thanhhaimai··on The Tariffs Have Some Math -- Matt Levine
https://archive.is/WO6DV
thanhhaimai··on xAI has acquired X, xAI now valued at $80B
This move makes it more likely that the internal number for TSLA is not good, and Elon is expecting the price to go down.

He has access to the real revenue number. If it's going well, he wouldn't have to perform this maneuver. xAI was relatively separated from X and TSLA, and wasn't having the backlashes associated with the two. Now he risks having the xAI branding tarnished too. He wouldn't do this unless the TSLA internal numbers are bad and he has to protect himself first, at the cost of xAI brand.

thanhhaimai··on Stoicism's appeal to the rich and powerful (2019)
> For one, that all bad things are actually good is not a tenet of Christianity and is not in the Bible

Just for my own learning, if it's not considered a tenet, why do I see this line of thought so often portrayed: "This <bad thing> is not bad. God made is so to test my faith."

thanhhaimai··on The DOJ still wants Google to sell off Chrome
Your source shows the "Market Share Worldwide". I don't see it shows the "Default Browser Worldwide".

AFAIK, in the US, the default for:

- Windows: Edge

- Mac: Safari

- Ubuntu: Firefox

I think the claim that "Google just pays PC makers to make Chrome the default browser" needs a clear source.

thanhhaimai··on Show HN: Can I run this LLM? (locally)
This is where I hope HN has a downvote option. This is not erroneous to the point I want to flag, but the quality is low enough that I want to counteract the upvotes. This is akin to spam in my opinion.
thanhhaimai··on Claude 3.7 Sonnet and Claude Code
> Third, in developing our reasoning models, we’ve optimized somewhat less for math and computer science competition problems, and instead shifted focus towards real-world tasks that better reflect how businesses actually use LLMs.

Company: we find that optimizing for LeetCode level programming is not a good use of resources, and we should be training AI less on competition problems.

Also Company: we hire SWEs based on how much time they trained themselves on LeetCode

/joke of course

thanhhaimai··on Vietnamese Graphic Design
No worries at all! If the site is open sourced, I can help update the translation text.

Thanks for the site, and keep up the great work!

thanhhaimai··on Vietnamese Graphic Design
This is such a great collection!

However, many of the translations are incorrect. For example:

This text:

> Dưới ngọn cờ quang vinh của đảng, nhân dân Nghệ-An Xô Viết anh hùng

Is incorrectly translated as:

> Under the glorious flag of the Vietnamese Communist Party, the people of Nghe An will crush their American enemies.

But it should be:

> Under the glorious flag of the [Vietnamese Communist] Party, the people of Nghe An are Soviet heroes.

There was no mention of "crush their American enemies" at all.

I would exercise caution and encourage the readers to double check the translation.

Edit: after reading more of the page, I changed my post from "some incorrect" to "many incorrect"

thanhhaimai··on Carbon capture more costly than switching to renewables, researchers find
> Think about it - if these formations were not leakproof on geological scale, there wouldn't be any oil and gas for us to extract in the first place.

This is survivor bias. We only see the sites where it didn't leak too much, and extract from it. The sites that leaked, we don't hear about them (no oil to extract). In fact, a quick search shows that there are many leaking sites (both on land and under the oceans). The argument you present is a case of survivor bias, especially on the time scale of multiple millions years.

thanhhaimai··on CDC data are disappearing
The question becomes: Is this crime pardonable? If yes, how likely the current administration will pardon those people?
thanhhaimai··on How to improve your WFH lighting to reduce eye strain
I'm curious what is the percentage of the population that prefers 3000k warm light. I personally disliked it. My mind constantly notices the distortion in color accuracy, and it's an unpleasant feeling. My preference is 5000k "day" light.
thanhhaimai··on Character amnesia in China
> The orthography may be inconsistently phonetic, as is the case with English spelling, or highly consistent, such as the Korean Hangul system. No writing system is perfectly phonetic. But phonetic systems enable the native speaker, with just a few dozen symbols, to reliably write whatever they can speak, and read out loud anything they can read.

I'm not sure if the author has studied Vietnamese. I'm a native Vietnamese, and I believe the language is perfectly phonetic.

If I hear a word, I can write it. If I see a word, I can pronounce it, regardless of whether I understand the meaning.

It's interesting that among the 4 countries (China/Japan/Korea/Vietnam), it's the only one that completely reinvented the language into Latin based. I think that refactor addressed the phonetic issue well enough. When I was there, there was also no TV program for "spelling bees" or something like that. Even a third grader could read/write almost any word (even when they don't understand the text yet)

Edit: adding to this original post to reply a common theme people brought up in multiple posts.

I think bringing up dialects and provincial accents is not convincing. There is one official way "gia đình" should be pronounced. It's taught in school, even in the South. Pronouncing it as "da đình" can still be understood, and it doesn't retract from the point that the language is phonetic.

In other words, assuming I know nothing about the meaning of the word, if I hear "da đình" I can correctly write down it as so. I wouldn't know that in Saigon that also means "gia đình". But I definitely can write it down exactly.

I don't think using provincial speaking accent is a good line of argument here. Otherwise, no language in the world can satisfy the phonetic requirements. Any group of people can have different accents, different tones, different sound length and pauses.

thanhhaimai··on Synchronization Is Bad for Scale
These are a lot of extraordinary claims. I would love to learn more about the evidence and what make you think this way. There are many successful companies that built scalable services with golang, so the burden of proof for the above claims are high.
thanhhaimai··on Machine Learning Is Still Too Hard for Software Engineers
> There is a lot of jargon and you do have to sit down and learn how things work, but once you do, deep learning is fairly simple.

This is my realization too. I think ML for SWE courses should focus on "translation" first. Like by "kernel" they mean this specific thing, not the normal meaning of kernel. This is similar to other fields like finance (which I'm working on). After you learn the language, it's actually not too terrible to understand.

thanhhaimai··on In Praise of Difficult Children (2009)
> but also need to test the rules and see if they are really as important as everyone says.

Thank you.

Then I don't quite agree with the article on this. I don't need to "test the rules" myself to see if they are ready as important as everyone says. I can observe.

For example: I don't need to drive without a seatbelt to "test" whether seatbelt is as important as everyone says. I can observe that people without seatbelt have worse outcomes from accidents.

I think this might also be related to whether the kid grow up with parents they can trust and are able to explain the reasons behind "rules" to them. I was fortunate that my parents always tried their best to explain why I should or should not do something. I questioned them a lot, but defaulted to "trust" when they didnt have time to explain.

thanhhaimai··on In Praise of Difficult Children (2009)
> Along a similar line, I’ve heard it say that if you consider yourself a peaceful person but are not actually capable of great violence, then are you actually peaceful or simply harmless?

Why not both peaceful and harmless (for the moment)? This is a function of time. I don't think it's possible to make that statement without supplying the time parameter.

Also, I don't think peaceful people are not capable of great violence. It's about choice. I'm capable of violence, but I choose to not use that capability.

← PreviousPage 2 of 8Next →