spprashant

I don't know how much they are spending to be fair.

I am basing my observation on the noises they are making. They did put out a model called Nova but they are not drumming it up at all. The model page makes no claims of benchmarks or performance. There are no signs of them poaching talent. Their CEO has not been in the press singing praises about AI unlike every big tech CEO.

Maybe they have a skunk-works team on it but something tells me they are waiting for the paint to dry.

I personally prefer being explicit in this case. Seeing the word INNER triggers the right way to visualize the query in my head.

I get the feeling Amazon wants to be the shovel seller for the AI rush than be a frontier model lab.

There is no moat in being a frontier model developer. A week, month, or a year later there will be a open source alternative which is about 95% as good for most tasks people care about.

Agreed on all points.

I don't view these apps as net negative for a country like India which is helped immensely by digitization.

My comment was just pointing out that governments have a way to get you install the app if they really need to.

They ll make it mandatory to access critical services at a later point. Tax payments, utility enrollments stuff like that.

That is how they ramped up enrollment in Aadhaar UID.

Its also just financially lucrative? The right tends to have more politically incorrect things to say and its no surprise click-farms from Asia would want to capitalize on that shock value.

For a long time, I did not get twitter either. But it seems to be the only popular platform where the academics and intellectual class want to hangout. Economists, researchers, policy wonks prefer posting on twitter over any other social platform.

The whole enterprise of the data center buildout reminds me of the chapter from Hitchhikers where they build a huge supercomputer for answers but never ponder what the question is.

Like someone else said your question is kind of broad. But I d recommend reading TigerBeetle's documentation to understand how financial transaction processing may differ from what they call general purpose databases like PostgreSQL.

https://docs.tigerbeetle.com/single-page/#concepts

A lot of outages off late seem to be related to automated config management.

Companies seem to place a lot of trust is configs being pushed automatically without human review into running systems. Considering how important these configs are, shouldn't they perhaps first be deployed to a staging/isolated network for a monitoring window before pushing to production systems?

Not trying to pontificate here, these systems are more complicated than anything I have maintained. Just trying to think of best practices perhaps everyone can adopt.

I have nowhere near the experience managing such complex systems, but I can empathize with this. In a high-pressure situations the most obvious things get missed. If someone is convinced System X is at fault, your mind can make leaps to justify every other degraded system is a downstream effect of that. Cause and effect can get switched.

Sometimes you have smart people in the room who dig deeper and fish it out, but you cannot always rely on that.

Initial impressions are currently worth a lot. In the long run I think the moat will dissolve, but currently its a race to lock-in users to your model and make switching costs high.

I think they could at least offer some sort of reprieve for people waiting in line. Their status is tied to employer whims. If someone has lived in the country for 5 years and in line for citizenship perhaps give them some protection in case their employment gets taken away. Some grace period, perhaps access to healthcare.

The US is doing something right if so many people are ready to wait in limbo for decades of the one life they get on this planet.

For people on employment visas - they are one economic downturn away from everything being undone. They ll get 60/90 days to leave the life and relationships they have spent years building.

I agree that YAML has its issues. But saying it's bad because it has too many features seems misguided. Do those advanced features get in the way of using a simple YAML file? At least not in my experience.

That being said TOML would be my choice.

This was a great breakdown. Loved the different aspects they captured beyond memory safety, including the skill improvements within the team as they grew comfortable.

I think thats part of the issue I have with it constantly.

Let's say I am solving a problem. I suggest strategy Alpha, a few prompts later I realize this is not going to work. So I suggest strategy Bravo, but for whatever reason it will hold on to ideas from A and the output is a mix of the two. Even if I say forget about Alpha we don't want anything to do that, there will be certain pieces which only makes sense with Alpha, in the Bravo solution. I usually just start with a new chat at that point and hope the model is not relying on previous chat context.

This is a hard problem to solve because its hard to communicate our internal compartmentalization to a remote model.

That d be his last name

Okay thanks for the links. I have supposedly not been reading the news.

I think we are getting to point where we are trying to figure how important is human experience to intelligence.

Things we do like sleep, meditate, have fun, listen to music etc. do they add to our intelligence? Do they help us have a consistent world model that we build on everyday?

Will we be able to replicate this is in a artificial neural net which is extremely smart in spurts but does not "enjoy" the world it operates in?

I am curious to see if jobs are actually moving offshore. Is there any data suggesting this?

I was under the impression Big Tech is cutting jobs across the board, all regions.

This is embarrassing.

The thing I find counter intuitive about AWS and hyper-scalers in general is, they make so much sense when you are starting out a new project. A few VMs, some gigs of data storage, you are off to the races in a day or two.

As soon as you start talking about any kind of serious data storage and data transfer the costs start piling up like crazy.

Like in my mind, the cost curve should flatten out over time. But that just doesn't seem to be the reality.

For upgrades which have enough risks as it is, I would keep the number of variables low. Once upgraded and stable, you can replicate to a secondary instance with io_method switched and test on it before switching over.

No issues. Trump can forgive almost anything for the right price.

JD Vance is a great debator, and might appeal to some urban Republic voters. But I don't see him having and away on the rural voters like Trump does.

That was a bit of ramble. Not something I d expect from a CEO of AWS who probably handles press all the time.

It reminds me of that viral clip from a beauty pageant where the contestant went on a geographical ramble while the question was about US education.

Yeah ON CONFLICT and MERGE are just syntactic sugar. At a fundamental level there are only Insert, update, delete and select operations on a SQL row.

Not all of these are "anti-patterns", your query clause not matching your index is a problem of not understanding how indexes work.

Some of these have nothing to do with SQL the language itself, and more to do with database schema design. If you have to do a DISTINCT, it means your primary key design is likely not right. If you are layering too many views, something is broken in the base table design, requiring the creation of all these views.

A good database model goes a long way to avoiding all this.

The thing we really strive for with Postgres is to keep the UPDATE traffic as low as possible. Because of MVCC, table bloat and the subsequent vacuum jobs will kill your IO even further. This means designing the applications and data model in a way that most write traffic is INSERT, with occasional UPDATEs which cannot be avoided. If you know you are going to have a UPDATE heavy table, be sure to set the fill_factor on the table ahead of time to optimize for it.

Also, in my experience "Faster SSD Storage" point applies to both read and write heavy workloads.