jammo

This is inherent in shared services though, (VMS).

Just so you know the cheapest per kg beef entry is wrong.

Have you found any customers who are too price sensitive for you? Presumably at some point it is cheaper to go and rent bare metal.

I don't really understand why Nvidia are investing in datacentres, bubble behaviour. If they were so certain in the long term value they'd provide it IAAS out of owned DCs

Really big impact! I'm not sure how fly or render work but if your compute instances are in $city make sure your planetscale instance is too. You shouldn't be far off 'in region aws' latency at the point.

Imagine what they could do if they focused a little bit more on making TPUs easier to use in real world applications. They are fantastic value but you feel like you're doing so much busy work to use them.

We all need to move away from these big cloud providers. Two medium size smaller providers is enough.

-Cloudflare for R2 (object storage) and CDN (Fastly+backblaze also available). -Two VPS/Server providers with a decent reputation and mid-size (using a comparison site like https://serversearcher.com or look directly into people like Hetzner or latitude) -PlanetScale or Neon for database if you don't co-locate it, though better to use someone like digital ocean, vultr or latitude who offer databases too)

Yes, but there are options for dedicated server providers who offer dual PSU and ECC ram etc. It's more expensive though for e.g a 24 Core Epyc with 384GB RAM dual 10G netowork is like $500/month (though there's smaller servers on serversearcher.com for other examples)

Equinix Metal is now EOL, so worth bearing that in mind..

I have a 2020 intel 10nm quad core MBP and my god even the M2 is so much faster. They are doing absolutely incredible work to be getting >10% improvement every single year without fail starting from that point.

I can see why they add the fee, but they would both garner so much goodwill by giving free accounts if the app you publish is open source. I don't think it would be that hard to automate by requiring a GitHub link.

It's particularly a problem with multiple access points, if it's just one and you need 'ok' coverage you're good.