riku_iki

> Recently, I got banned from Reddit for sharing my local news summarization website (www.cafelutza.ro)

how exactly this happened? What did you do?

how did you handle HA/replication story with sqlite?

I see you are trolling, bye

It means high availability.

DB with cross zone replications and HA for starter.

those who don't need cloud services are feel free to use other options.

cloud providers indeed could abuse vendor lock, but LLMs are not that easily vendor lockable.

gcp was net negative until last year.

Big part of why clouds are expensive is not necessary hardware, but all software infra and complexity of all services.

reliable network and partitioning are two different things.

If your network is down, your clients can't connect DB too, and nothing works. State of network partitioning is harder to achieve, its more like if connection between two datacenters is down where DB replicas live, which is more rare because of all global connection redundancy.

so, what do you use instead?

2 database migrations could be a huge effort.

> Postgres is not a CP database, and even with synchronous replication, it can lose writes during network partitions.

Argument is that partitioning almost never happens, because of network reliably rerouting traffic.

No compression.

I do it.

The big problem for me from running DB on Btrfs is that when I delete large dirs or files (100GB+), it locks disk system, and Db basically stop responding on any queries.

I am very surprised that FS which is considered prod grade having this issue..

> is new datasets + post-training shaping the model's behavior (instruction + preference tuning). There is no moat besides that.

sure, but acquiring/generating/creating/curating so much high quality data is still significant moat.

it depends on complexity of your infra.

Running full scale kubernets, with multiple databases and services and expected 99.99% uptime likely can't be handled by one person.

> We haven't felt the need to set up auto fail-over to the hot spare; that would take some extra effort (and is included with AWS equivalents?) but nothing I'd be scared of.

this part is actually scariest, since there are like 10 different 3rd party solutions of unknown stability and maintanability.

Sure, as well as other powers are actively aggressive during last N thousands years, that's how humans operate, who don't they extinct.

Its not approved yet I think.

Sure, both US and EU run multiple investigations.

While in US, potus can impose tariffs at whim, until scotus decides otherwise, my understanding is that EU tariffs are results of such rulings.

> Selling dollars for $.50 does that. It sounds like they have a business model issue to me.

its not. The idea is that majority subscribers don't hit limit, so they sell them dollar for 2. But there is minority which hit limit, and they effectively selling them dollar for 50c, but aggregated numbers could be positive.

tons of posts on reddit that they also significantly dropped quality

The question is to what extent. Both US/EU and WTO have anti-dumping rules.

> Never? For the sheer amount of moonshot bets he's doing, his track record would make any VC jealous. Zip2, PayPal, Tesla, SpaceX, Grok/xAI.

not never, but xAI is not clear if successful, and others were bootstrapped 10 years ago, and not clear if they need mask at this point.

> X.ai is controversial mainly because of Musk. But if you can look beyond that, it does actually have a bit of non trivial IP. Grok is not bad as a LLM.

the problem is that this market becomes commoditized, there are tons of not bad open weight LLMs available, and it is not clear if Grok IP is that not trivial. They even totally can run others LLMs under the hood, it is well known that they used Claude output at the beginning.

> How isn't China playing by the same rules?

one opinion is that tariffs on China was response of breaking rules by China (heavy subsidies on domestic EV and similar).

more evidence provided -> more chance concerns won't be ignored

per my previous research there is no opt out for gemini cli.

Google has uncertain privacy settings, there is no declaration they won't train their LLM on your personal/commercial code.

Leaderboard looks very outdated..