wbsun

5 trillion IPO of three companies this year! They still need real money to buy the offered stocks, I am wondering which markets will be sold in order for investors to get the cash.

I wish people'd report metrics like "transactions per second, 99.x% availability with N secs/mins p99 latency, ..." when describing how practically useful and how effective a real-world banking/transaction system is, which further proves the practicality of the programming language.

We serve over 300,000 businesses. We processed $248 billion in transaction volume in 2025 on $650 million in annualized revenue

The system we've built has worked well for years, through hypergrowth, through the SVB crisis that sent $2 billion in new deposits our way in five day.

This is a strange way to describe a transaction processing system with total amount of money it processed. The reliability or scalability is not measured as O(n) dollars. In theory, $248bln or $2bln can be done in one transaction although I know it is not the case in reality. It would be impressive to see the typical system design properties like transactions per second, latency distribution, etc.

It’s less about 'old vs. new' and more about the evolving trade-offs dictated by the constraints of the era. There have always been engineers trying to squeeze every last drop of performance out of the bits available to them.

The blog is new but the paper was submitted almost one year ago: https://arxiv.org/abs/2504.19874. Anyone has ideas if this is already implemented in many models (at least Gemini, I guess)? If that's the case, can I expect cheaper RAM for my computer :D

In SF bay area, there are always gas cars trying to show off their horsepower especially during rush hours, no matter how slow the traffic is. That noise is more annoying than tires.

Palantir's secret weapon is the closeness and affinity to the DoD.

Is it a secret? I got an impression that it has been well known. How could you get any big number contracts without former secretaries or retired generals in your board or in your ‘consulting’ team?

Oh I am more than happy to tell people how I took down entire Google Cloud 11 years ago. I mean, of course to the level of details Google is comfortable with to share externally :)

No, sorry, I wouldn't buy EVs from any US company any more until they are matura enough that the model production surpasses at least Tesla Model S+X since released. Their reliability just suck. I don't want to waste my life again and again in those months-long waiting of service appointments, annoying issues every where and every day, hiring Lemon law lawyers and other BS.

Data points:

- one Model X

- one R1s

- My neighbor: one Model X, one R1t

- My collegemate: one Model X

Nice, now you can dd it to your boot sector and ... Wait, it is 2026, there are 1000 ways of booting and memory mapping on so-called unified ARM architecture @,@

These people are not going to take your job, the people who uses tools smartly while having the knowledge and experience in highly reliable distributed systems are. If human in the loop is not required any more, nobody is going to keep their job.

Making 20 y/o CPU with today's process? It is cool but yeah not really a wise business decision.

Oh running Linux on a MCU without MMU.

I feel there is a gap between MCU and modern CPU, and also between the software running on top of them. The missing piece is a mid-size computer with: - A processor, single or multi-core, with computing power like 20 years ago, but modern fabrication process. Maybe without MMU for simplicity. - RAM between 100MB~1GB and DDR2/3 bandwidth. - An OS designed and implemented for this type of hardware rather than tailored Linux.

I don't think you can use it for working or your daily entertainment, so I guess not a good business to attract interests.

My experience: customers with $$$ will always believe they are very important, so important that losing a single bit is the end of world.

So you may not want to convince customers waving huge $$$ checks that their data are not that important. But instead, providing options to keep them once they realize that: their pockets are not that deep, and they are also totally ok losing some data.

Does their status page depend on something that is down already, so the page just fails static now hence no new updates?

Am I the only one who doesn't want to type a lot while browsing? (I comment on HN very rarely too...).

Killing it so quickly after buying, it doesn’t look like the board were really convinced as the author believed.

This feels like the 80s/90s when people hacking assembly or finding undocumented instructions to squeeze CPU for performance. Until one day either the compiler will be highly optimized enough or the GPU will be so powerful that such tricks won’t make much difference anymore, like CPUs nowadays :D

I don’t understand why this is not obvious to many people: tech and stock trading are totally two different things, why on earth a tech expert is expected to know trading at all? Imagining how ridiculous it would be if a computer science graduate will also automatically get a financial degree from college even though no financial class has been taken.

The magic and mighty asdfasdf..., it is like foo/bar in naming, Alice/Bob in security, deadbeef in bit ops.

I've decided to only receive calls/msgs from my contacts. On iPhone, you can do it by "Silence Unknown Callers". Instead of pretending I can ignore the spam calls, I'd rather take the risk when something super important are coming from an unknown number...

Good to see such bare metal work in HN, I feel there has been fewer this type of work nowadays.

Back to the OS architecture, AFAIK, exokernel tries to get rid of the OS overhead (context switching, abstractions, etc.) to improve performance back in 1995. But I assume with modern computers/workload, the OS overhead only takes a tiny piece. Is that the case? Are people still optimizing OS heavily for today's workload like AI/ML/servers?

Isn’t there also hash based index for random keys?

I feel that eventually people will hate every commercial product and the company. The interests of users and the company have a fundamental conflict because one is trying to get as much money as possible from the other with the least possible cost. The only honeymoon time is when the company is still trying to win customers by sacrificing their profit.

People probably should hate every mature company.

Just curious, any startups working towards this goal? Replacing leadership, decision making, people management, project management, negotiation and such with AI?

I work on a large distributed infrastructure. I always joked that my team's projects and people's careers are outage-driven: the only time we become important and people get opportunities for promotion is when where were big-enough outages that executives have to invest heavily on reliability or scalability. Other time, we are just minions who must listen to and serve feature or product teams. Nobody listen to us when we ask a product team to implement a reliability contract in their shining product.

Pre-outage improvements, reliability defense in depth, eliminated scalability bottlenecks before they are hit, are all ignored by leadership and the company: it is just human nature that even though they understand you have to prepare for possible issues, if it hasn't happen yet, you won't take it seriously. I've seen this in many internal performance reviews and promotion committees. People who haven't ever got bitten badly by an outage may call these premature optimizations.

I am always wondering why ASML hasn’t got any antitrust lawsuits for their monopoly.

As some comments mentioned that the culture in Boeing has been broken for a long time. How to fix it? I heard about a solution: set up a policy that every executive and board member must take their "safe" airplane once a week.

I like the old times when you could assume everyone around you is smarter than you, so collaboration and communication were never an issue. They never rely on "experts" in other teams to collaborate, they quickly and easily pick up how other systems work in depth themselves. Smart people just shine and work together to create amazing stuff.

Nowadays, you need to explain to people why unit test is needed, why you can't use production as the first place to try a risky experiment, and rely on 20 experts, one in each tiny service, to figure out and to deliver a tiny feature with bloated timeline and messy quality.

People often learn the lessons in a hard way: they will keep saving 230k/yr until one day their non-HA bare-metal is down and major customers retreat.