acchow

Opus:

No life ran more bleak and desolate than Yu Xin's —

yet in his twilight years, his verses stirred the rivers and the passes.

I don’t think you need a 10-30b model for most smartphone use cases.

But I meant to counter gp’s claim that “I can run a 27b model on an iPhone” is kind of pointless and disingenuous. Yes I’m sure someone will come up with a way to run a “27b model” at 0.1 bit quantization on an Apple Watch pretty soon misses the whole point of saying a model is “27b” in capability.

Achieving a parameter count is not the point. And is almost meaningless

We're too easily conflating parameter count with capability. That Bonsai 27B you're running is at 2-bit quantization. Is it really better than the best 10-18B models?

It makes it impossible to click on any links in the footer of the page

That's a UI bug. Real infinite scroll does not have a footer.

No, Apple Watch won't have the compute or the ram needed in 5-10 years. That will require 8-9 doublings in RAM and even more in compute.

and nothing would stop you playing it

Not true for PS5 games. Sony can push a firmware update to disable games, even if you own the disc.

There are only 170 ground stations (about 100 in the US).

So for someone living in rural america, it's really "famous last 300 miles".

Agreed. The graphs clearly show that opus 4.8 performs strictly better at the same cost per task

That $6700 is a $5000 upgrade over a base model Macbook Pro.

$5000 in US Treasuries (currently at 4.89%) yields $244.5/yr. That's more than enough to cover the annual Claude Pro subscription ($200/yr) which includes Claude Code with lots of Sonnet usage (far better than Qwen 3.6)

We know intuitively that a ball atop a 20ft ladder has twice the potential energy of a ball atop a 10ft ladder.

This assumes that energy increases linearly with distance (given constant force over that distance

Looks like in your 2nd equation you've already assumed kinetic energy is quadratic with speed

T = ½∑ⱼ mⱼ v⃗ⱼ²

MSFT is also down 3.5% today, so there's just lots of volatility in big tech.

Calgary is quite sunny at 2400 hrs/year.

But not nearly as much as Vegas (3800) or LA (3250) or SF (2950).

pareto frontier does not mean cheapest.

I had a similar issue with the Nikon Z5 full frame camera. It can be charged via the built-in USB-C socket only if the battery still has some charge.

If the battery is fully dead, you have to remove it and charge it using the separate battery charger.

That means you can't travel with only a USB-C cable.

Wikipedia say 3.6 billion Chrome users.

I get that there is undesirable real estate in Japan that gets sold "for a song".

But then how does it quickly get resold at 40x?

Most people are convinced LLMs can do this.

Cal AI, which claims to generate a nutritional breakdown based off a photo, has $30 million in annual recurring revenue.

senate and congress

Senate and congress are both elected. Their re-election is effectively jury nullification.

The people do not care about the crimes.

That post doesn't say anything about training for SVG generation

I think something like 96GB RTX PRO 6000 Blackwells would be the minimum to run a model of this size with performance in the range of subscription models.

GLM 5.1 has 754B parameters tho. And you still need RAM for context too. You'll want much more than 96GB ram.

Which in the series specifically?

H1b is tied to employment, not to the employer. You can change employers on the same H1.

It’s not great. But this is similar to how health insurance is tied to employment, not to the employer. Both citizens and H1 employees experience the same abuse here

asserting that LLMs will never generate 'truly novel' ideas or problem solutions

I don't think I've had one of these my entire life. Truly novel ideas are exceptionally rare:

- Darwin's origin of the species - Godel's Incompleteness - Buddhist detachment

Can't think of many.

Every MCP server injects its full tool schemas into context on every turn

I consider this a bug. I'm sure the chat clients will fix this soon enough.

Something like: on each turn, a subagent searches available MCP tools for anything relevant. Usually, nothing helpful will be found and the regular chat continues without any MCP context added.

But MCP uses Oauth. That is not a "worse version" of API keys. It is better.

The classic "API key" flow requires you to go to the resource site, generate a key, copy it, then paste it where you want it to go.

Oauth automates this. It's like "give me an API key" on demand.

Interesting pricing differential. Seems in your country, that IdeaPad is significantly cheaper than the price in the US. But for your Macbook Neo, it's the other way around.

Wonder why that is.

I had the opposite issue. Trackpoints stated hurting my hand because it requires significantly more force than the Mac's touchpad.

We probably shouldn't deprecate AM for emergency broadcasts given you can listen to AM radio with grass https://www.youtube.com/watch?v=b9UO9tn4MpI

Claude code always give me rate limits

50+ tps with qwen3.5 35b a4b on a 4090

But qwen3.5 35b is worse than even Claude Haiku 4.5. You could switch your Claude Code to use Haiku and never hit rate limits. Also gets similar 50tps.