One thing you haven't noted is radio.

Some local radio DJs frequently play songs I enjoy that have under 1K plays on youtube. No algo or platform is surfacing those. Local radio gets me both local and international music. A friend of mine prefers critically acclaimed stuff, so he streams radio shows from NTS and the like.

it's disappointingly easy

Warcraft 3 had some of the greatest mods that kept the player base alive for decades

I've never been a fan of Warcraft, but I did play the custom Warcraft 3 TD maps at lan parties. Some were just so epic.

>When I'm building out a new feature, I can churn through millions of tokens in Claude code.

+

>Not sure what model provider profit per token is, but let's say it's .001 cents.

So you'd be willing to pay thousands for a new feature, right?

I did not realize Eno could not read sheet music. I always thought he used graphical expressions in his presentations as an artistic choice.

imo, this paragraph covers the essence of a good chunk of the article: https://en.wikipedia.org/wiki/Discovery_of_penicillin#Replic...

A bit of OT, but I have four iPhone 5/5s/SE (the SE is peak design and form factor, fight me) lying around that I use strictly as offline devices for things like saving data from my heart rate monitor, controlling my action camera, doing voice/field recordings through the 3.5mm connector – stuff I'd prefer never to leave my device (or data that should be open to user control but requires an invasive app to work, I have very few apps on my daily driver).

These devices are are small, snappy and powerful enough in 2025.

I am going farther offtopic, too. I have removed the radio on Samsung Galaxy IV device† and they still work. I don't know if that is possible for iphones, just throwing it here.

† I don't have the skills to reverse the process, though :)

This is based on HSBC's model, which assumes some incredible numbers, such as: > user numbers on an S-curve that by 2030 reaches 3bn, “equivalent to 44 per cent of the world’s adult population” ex China.

Unfounded statements (outside of language tasks, fwiw), such as: >LLM subscriptions will become “as ubiquitous and useful as Microsoft 365”, HSBC says.

As well as this bold claim about OAI's potential to double the conversion rate: >It models that by 2030, 10 per cent of OpenAI users will be paying customers, versus an estimated 5 per cent currently.

Does not include a major player in its market share analysis at all: >Google is excluded entirely

And, still, it suggests that: > OpenAI is expected to still be subsidising its users well into next decade

Fascinating.

Still, it's only 40bn per year divided by 3b so equal to around 15$/person/year

Isn't that super cheap? Just think of the revolutionary impact it would have on education, health, work etc.

I don't understand how anyone can call it a bubble.

I think the question is whether people are willing to pay for an LLM when there are equivalent or good-enough free competitors available.

One could argue that LLMs will change the world, but that doesn't guarantee that LLM companies will capture any of that value.

The additional rub is that the paying power users are arguably costing these companies more money than the free users.

> I don't understand how anyone can call it a bubble.

Perhaps because in this scenario, even after (only) an additional $40bn a year for the next 5 years, OpenAI will still be losing money.

I wonder if the intern uses ChatGPT.

I've been distro hopping recently and missing the Web Apps application in Linux Mint. Somehow, Tangram slipped my radar until now. Seems to have never been mentioned on HN too.

more: https://github.com/sonnyp/Tangram

SEO slop is what the LLMs were trained on. GIGO