tcldr

Ban distillation of our outputs, but our distillation of the sum-total of civilisation's intellectual output – proprietary or otherwise – is fair use?

Either everyone licenses, or nobody does. And if you can't enforce licensing bans for everyone, the de-facto loser is those who you'd probably want to support the most, start-ups and universities, while your adversaries gain the upper hand.

The strongest argument for restricting distillation is arms control – but distillation is the way to defeat GPU embargoes. So, distillation goes on regardless. Only pre-training is seriously attenuated.

If we're honest, the models are compressions of everything society has ever written. A few large corporations can't own that, no more than they can claim copyright for a zip file of the public library.

The genie is out of the bottle, now. So open it up – inputs and outputs, forward-looking – for everyone.

The price increases for dedicated servers are reflected in the higher setup costs. For example, I believe AX102 setup was EUR 39.00 now EUR 269.00.

Price would be a bit more bearable if their reserved instance discounts were more generous

Agreed. If the CFC (controlled foreign corporation) rules still apply for founders in EU-member states, it will fail.

I’m hoping they can be creative and find a way to distribute revenues to member states in a way that works for everyone.

For employment taxes, one way could be to tax EU-inc employees as if self-employed in their personal tax domicile.

> Of all the challenges you face as a startup, the legal entity you choose is possibly the least consequential.

The amount of founders who choose to domicile their company in Estonia because the ticket rates and ease look attractive and who don't understand that this will still need to be administered in their local market as a CFC (controlled foreign corporation) would probably say differently.

> Just choose a jurisdiction where investors understand how the legals work (Delaware C-corp, UK Ltd is OK too) and there's a finite administrative burden and/or commoditized tooling in place to help you handle it.

That's exactly what EU-INC is trying to provide/solve afaict.

Totally agree. If I don’t understand the code as if I’d written it myself, then I haven’t reviewed it properly. And during that review I’m often trimming and moving things around to simplify and clarify as much as possible.

This helps both me and the next agent.

Using these tools has made me realise how much of the work we (or I) do is editing: simplifying the codebase to the clearest boundaries, focusing down the APIs of internal modules, actual testing (not just unit tests), managing emerging complexity with constant refactoring.

Currently, I think an LLM struggles with the subtlety and taste aspects of many of these tasks, but I’m not confident enough to say that this won’t change.

Fairly subjective, but personally I find all apps being within the browser quite constrictive. I'd much rather have my apps unencumbered by browser chrome and unintended keystrokes, persisting their window size/position and all kind of other affordances. Definitely not a fan of the 'browser as the OS' philosophy, as it feels a bit inception.

That said, I'm less and less bothered by an app that's Electron under the hood, but I think that's more to do with the quality bar for native apps slipping over the past few cycles (macOS) and forfeiting their advantage.

No.

In Robert Bringhurst's The Elements of Typographic style – pretty much a bible amongst typographers – he states:

We should “[u]se spaced en dashes – rather than close-set em dashes or spaced hyphens – to set off phrases.” Bringhurst then adds this devastating indictment:

The em dash is the nineteenth-century standard, still prescribed in many editorial style books, but the em dash is too long for use with the best text faces. Like the oversized space between sentences, it belongs to the padded and corseted aesthetic of Victorian typography.

With a warrant from a judge people should be compelled to provide access to their encrypted files or be in contempt of court with all that entails. Anything else is overreach.

Scaleway has a broad range of services, but their price competitiveness/reserved instance discount is a bit lacking.

Hetzner's great, and I'm a customer, but we're missing lots of stuff there – and some of the stuff that is there isn't reliable.

For example:

* Object storage: lots of horror stories out there regarding flakey performance so hard to justify sticking mission critical stuff there.

* Private networking: Again, too many anecdotes about loss of service. Lots of people just using the public IP6 interfaces to avoid their private networking entirely. And private networks are IP4 only.

* Kubernetes CSI: I've had issues with this where a PV gets in some locked state and I can't remove from console.

I'd love to see more competition here.

Maybe not possible with today's SOTA AI but I have no doubt it's within reach.

Just because we're unable to compensate many millions, perhaps billions of people, for using their work without a) permission, or b) remuneration, doesn't justify giving a blanket license to use it without some form of *serious* compensation that reflects the gravity of what is being created.

The current winner-takes-all approach to the outcome is wholly inappropriate. AI companies right now are riding atop the shoulders of giants. Data, mathematics and science that humanity has painstakingly assembled discovered, developed and shared over millennia. Now, we're saying the companies that tip the point of discovery over into a new era should be our new intellectual overlords?

Not cool.

It's clear that model creators and owners should receive some level of reward for their work, but to discount the intellectual labour of generations as worthless is clearly problematic. Especially given the implications for the workforce and society.

Ultimately we'll need to find a more equitable deal.

Until then, forgive me if I don't have much sympathy for a company that's had its latest model distilled.

Exactly. This is the argument that I find lacking from today's discourse: AI companies are already extracting generations worth of human intellectual data into their models. If they want to argue that this is 'fair use' then model distillation is, too. Can't have it both ways.

I think the audience who can appreciate handcrafted code will be vastly smaller than the audience who appreciates hand carved wood.

You're right. When I first watched it, I was under no doubt they lived and breathed that philosophy. It matched my perception of their output 100%. Watching it again now, I'm reminded of how I used to feel and how much things have changed.

It's crazy that marketing hasn't worked out that quality and reliability can be spun as a feature. In fact, I remember with OS X, that was the baseline word-of-mouth feature when the comparison was made with Windows at the time.

"It just works"

Hetzner is great value, but their networking has a few issues:

1) Networking is mostly limited to 1Gbps. Even private networking. You can request a 10Gbps NIC, but it has to be housed in the correct data center and adds a $48 monthly fee.

2) Private networking is IPv4 only so dual-stack private networking isn't possible. Also each public IPv6 address is /64. Would be nice to get a /56 to setup dual-stack IPv6.

3) Can't specify a subnet to assign a server to when using hcloud API/Terraform. You have to specify the required IP on the subnet explicitly.

4) As I understand it, the private network traffic isn't truly secure between tenants, so needs to be encrypted between nodes anyway.

Still, I'm betting they'll fix these issues as their offering grows.

It will be fascinating to see if the protectionist foreign policy that's been adopted by the US will lead to an improvement in the quality of services available elsewhere.

Previously, the friction of using a service with slightly rougher edges would have tipped the scales against it. Now, it seems we have a kind of patriotism emerging in our purchase decisions.

Ultimately, it should give us all more choice through strengthened competition.

Thanks, I’ve seen bits from their engineering blog that have been really insightful, so that’s what got me to wondering if there’s more literature buried away on YouTube or elsewhere.

Try Decibel Meter Ultra (https://www.decibelmeterultra.com/) if you have an iOS device. It will give you a graph over time and you can leave it running over night. Free version is limited to 2 min duration but has a 7-day free trial.

> and happens a lot more frequently on Airbus models to the point that Airbus modified the door latch.

So you’re saying Airbus resolved this issue? Why hasn’t Boeing?

Actually, it's used pretty heavily by SwiftUI. The DSL for SwiftUI relies pretty heavily on n-element tuples that were previously limited to something like 10 children. So you'd have something like:

  VStack {
    ViewA()
    ViewB()
    ...
    ViewJ()
    ViewK() // ERROR: A VStack is limited to 10 elements
  }

The only way this could be implemented previously was with a separate init for each possible element count. A lot of auto-generated code.

This eliminates that. So even if you're not implementing it yourself, you're getting the benefits.

Beyond reading old code samples, knowing obj-c provides diminishing returns these days.

All relevant obj-c libraries provide an auto-generated native swift interface, and you can even ‘import objc’ if you fancy doing some obj-c runtime stuff – such as swizzling – and still write in Swift.

I wouldn’t expect to see and/or use it for anything but the most esoteric of use cases these days.

Dear Google, we were only adding ‘site:reddit.com’ or ‘site:news.ycombinator.com’ to search queries so we could get to opinions that weren’t being manipulated by SEO fiddlers. What’s our alternative now?

I’m aware. But when you’re trying to set a tiny brand mark over a photo in the corner of some social media thumbnail, a screen’s fidelity is not the limiting factor, it’s the human eye.

Sure, a restricted subset of serifs and typically when you’re reading a run of text i.e body copy. But the typical neoclassical serifs used in high fashion (think the Vogue logo) with their hairline serifs will look awful scaled to the sizes needed on mobile – regardless of screen definition.

Q: Why do everyone’s logo fonts look the same?

A: Because they need to be legible on a mobile device.

It’s no coincidence this trend started in the 2010s with the arrival of the smartphone. Brands need a consistent look that work across mediums. With over 50% of e-commerce sales happening on mobile, and the dominance of social media in the marketing of, for example, high fashion, a brand mark must excel in these kinds of treatments. Perhaps we’ll find another design trick to facilitate legibility at smaller scales but until then, those marks that looked great in print, aren’t fit for purpose.

The 'Happy Meal' strategy. a.k.a Get 'em while they're young. What you learn on today, is what you'll mass produce on tomorrow.