mercutio2

oMLX makes prefill effectively instantaneous on a Mac.

Storing an LRU KV Cache of all your conversations both in memory, and on (plenty fast enough) SSD, especially including the fixed agent context every conversation starts with, means we go from "painfully slow" to "faster than using Claude" most of the time. It's kind of shocking this much perf was lying on the ground waiting to be picked up.

Open models are still dumber than leading closed models, especially for editing existing code. But I use it as essentially free "analyze this code, look for problem <x|y|z>" which Claude is happy to do for an enormous amount of consumed tokens.

But speed is no longer a problem. It's pretty awesome over here in unified memory Mac land :)

“Scared” to “take risks”?

This is a bizarre way of saying “if they ship it and it has reliability problems, they know they’re skating on thin ice”.

Apple’s brand has taken a beating (I’m as aghast with the latest macOS as the next nerd), but people love that when Apple ships a product, it generally works and the hardware doesn’t break.

Butterfly keyboards are a terrible stain on the hardware team’s reputation. “Scared” is the wrong word for how these things work.

I was working at Apple and wondering the same thing ;)

Turns out people like them. Not so much the HN crowd, but c’est la vie.

I wonder if the reason that most people don’t agree with me about Antigravity is because you were used to VS Code?

For me, Antigravity is possibly the worst GUI experience I’ve had since Clippy.

It’s completely filled with arcane buttons, prompts that are effectively modal appear in at least 3 different places, it’s constantly… doing stuff, without any reference to where I should focus my attention.

I appreciate Google giving away absurdly generous quantities of tokens for FREE just to get me to use the thing, but I can’t bring myself to, because when I get into the flow of a feature with an LLM, I’m suddenly stuck and can’t figure it out.

It’s like peak Google UI for me.

I used to feel this way, then a week or two ago, after an automatic update, it started hanging. All. The. Time for me. Launching the app now frequently takes 30 seconds before it shows me the “load a git repo” screen.

Speed was its main advantage, before. It has become nearly unusable.

The price of extraordinarily rapid iteration, I suppose.

I'm surprised that there were downvotes. This is an excellent answer, and better interfaces between linguistic definitions of color and physicists' than saying "Rayleigh scattering impacts blue more than red"!

I guess I’ve never tried the pro models, because I’ve used gemini-cli free every day for the last three months or so.

It does eventually finish its quota, but then I just switch to a different Google account (which, amusingly, is what Gemini told me to do).

Happy to consume Google’s free tokens! The free model is a distant third for coding, but it’s fine for leaf node work in a larger project.

It hadn’t occurred to me this was a billing bug.

That would be heartening, if I wasn’t consuming tokens 10x as fast as expected, and they just had attribution bugs.

Do you have references to this being documented as the actual issue, or is this just speculation?

I want to support Anthropic, but with the Codex desktop app *so much better* than Anthropic’s combined with the old “5 back and forths with Opus and your quota is gone”, it’s hard to see going back

Yeah, the generosity of Anthropic is vastly less than OpenAI. Which is, itself, much less than Gemini (I've never paid Google a dime, I get hours of use out of gemini-cli every day). I run out of my weekly quota in 2-3 days, 5-hour quota in ~1 hour. And this is 1-2 tasks at a time, using Sonnet (Opus gets like 3 queries before I've used my quota).

Right now OpenAI is giving away fairly generous free credits to get people to try the macOS Codex client. And... it's quite good! Especially for free.

I've cancelled my Anthropic subscription...

Wealth taxes are very, very different from higher income taxes.

People are mad about buy-borrow-die, so they’re proposing extraordinary new measures.

Personally, I’d just make capital gains taxes apply at the “borrow” stage to actually fix the problem. That would have a host of compliance issues but they’d be localized in the finance industry which already has an army of people figuring out compliance.

Can you tell me more about your agent harness? If it’s open source, I’d love to take it for a spin.

I would happily use local models if I could get them to perform, but they’re super slow if I bump their context window high, and I haven’t seen good orchestrators that keep context limited enough.

What toolchain are you going to use with the local model? I agree that’s a Strong model, but it’s so slow for be with large contexts I’ve stopped using it for coding.

This is one of the very few non-money-laundering use cases for crypto.

I would support a “5 cents per unsolicited email” email system, in a similar way. If you make it a mildly enjoyable $5/hour task to read the first sentence or two of your spam folder, the overall internet would be better.

Maybe it's because I loved the books, but I loathed the Netflix adaptation. Possibly the worst sci-fi adaptation I've ever seen.

The casting was OK, but they mangled the plot and motivations of every character nearly beyond recognition!

The Apple TV (the device) has a “stuff this user watches” app (called Apple TV) which has a tiny subset of its features dedicated to AppleTV+ (the service).

Netflix refuses to participate in “stuff this user watches”, it would be trivial to do, but Neflix jealously guards its viewership numbers and I expect this is the main reason they don’t do it. That and… they’d rather you just browse Netflix and not watch other services.

The “stuff this user watches” app is very useful! I like it a lot, when I’m not watching Netflix stuff! It works with every service except Netflix!

But the moment the family shifts over to watching some Netflix show, it forces us out of the habit of using the TV app, and then we go back to the annoying “spend 90 seconds trying to find what we were watching on Hulu” experience, which is worse in every way.

Why are you assuming active heat transfer? Passive is the way to go.

Yes, so?

Everyone keeps talking past each other on this, it seems.

“Generating power in space is easy, but ejecting heat is hard!”

Yes.

“That means you’d need huge radiators!”

Yes.

OK, we’re back to “how expensive/reliable is your giant radiator with a data center attached?”

We don’t know yet, but with low launch costs, it isn’t obviously crazy.

Yeah. A City on Mars made me want to throw the book at the window so many times. Building and tearing down straw-men right and left. Almost every legitimate note of caution suffered from the nirvana fallacy.

There’s a weird thing in discussions about space. Lots of people just don’t like space, it makes them think they’re being blasted with science fiction.

So much criticism of space seems to fall into a few categories:

  1. They think there were ever any serious engineers who thought STS was a good idea, (rather than congressional-pork, which is what it always was), and thus assume actual space technologists are basically always wrong about the possibility of ever creating anything new and reliable
  2. They think cost/kg to LEO is somehow a physical law, and can never be improved on
  3. If they accept that SpaceX might actually have better technology that allows new things, they still refuse to wrap their heads around 2-3 orders of magnitude cost reductions due to improved technology, they update, but mentally on the order of “it will be 50% cheaper, no big deal”
  4. They just hate Elon Musk. On this one, I’m at least sympathetic

Space based data centers are probably not going to happen in the next decade, but most criticism (including this article) just reads as head-in-the-sand criticism, not serious analysis. I’m still waiting for more serious cost-benefit analysis assuming realistic Starship mass budgets.

If I worked for SpaceX, I imagine I’d focus more on just getting more Starlink mass in orbit for at least 3-4 years, but after that, we might have spare capacity we might want to spend on orbital power loads like this.

You mean, not very much? Everything about space-based anything is dependent in the short to medium term on Starship making mass to LEO cost about as much as air freight.

Starship, at least as a rapidly reusable second stage, may fail, rockets are hard. But you aren’t really engaging with people’s dreams if you start from “we don’t have access to the technologies that appear to represent a one to two order of magnitude cost shift”.

We’re the same age, and I had exactly the same reaction.

AT&T and the baby bells were widely loathed (man I hated Ameritech…), so the idea they would extend their tentacles in this way was the main thing I reacted to. The technology seemed straightforwardly likely with Dennard scaling in full swing.

I thought it would be banks that owned the customer relationship, not telcos or Apple (or non-existent Google), but the tech was just… assume miniaturization’s plateau isn’t coming for a few decades.

Still pretty iconic/memorable, though!

Wow! I just spent a good chunk of time last week setting up headscale and split horizon SSL behind my network, and I expected I was going to just expose a Wireguard UDP port, but discovered no, it’s DERP or nothing. DERP has been OK, but I think just exposing a UDP port on my local network is better.

If we’re really confident in the security of that UDP client, that is. I feel very comfortable exposing a Wireguard bastion, time will tell how secure whatever protocol tailscale is serving, here, will be.

LEO is the last place you should worry about space debris.

Space is just unfathomably large. If you aren’t in the same orbital plane, you’re just not going to have a problem. And if you did, Kessler syndrome in LEO is a non problem.

Could be an issue for specific orbital planes in stable orbits, but even there, it’s overblown.

This is pretty classic Apple: hide useful information, expose it to power users who discover it.

To be fair, on macOS, it’s a longstanding tradition that if you want to see more complicated stuff (in menus, in apps) hold down option.

But yes, it’s not discoverable.

I guess I wasn’t clear in my question.

What are you getting beyond what normal docker-compose gets you was the question I should have asked.

There were like a dozen answers all referring to things I think you can just use docker compose for.

Did I miss something?

Hold down option to highlight what accounts a contact is in, BTW.

It’s rare that I get feedback like this about a feature I owned for a decade.

1) The vast majority of users have only one contact-sync account, so it’s not an issue for them, merge works fine

2) For users that have multiple contact-sync accounts, they almost never want a feature to silently choose one account’s contact and delete the other account’s contact. So linking is really what these users want if the contacts live in different accounts.

It’s interesting feedback that a combined “link or merge” command would be what you’d expect. That’s a reasonable request; in my day we generally steered clear of combining destructive operations (merging) with non-destructive (linking).

I was more focused on the fact that the macOS implementation of “look for duplicates” is pretty broken; there’s a decent iOS implementation we never got around to migrating to macOS.

What does having a DevContainer get you?

I’m all for documenting every bit of my setup, but beyond that…

Water vapor in the stratosphere has a very high radiative forcing. Offset somewhat by particulates in the upper atmosphere.

Cirrus clouds and contrails have a distinct, and large, additional forcing.

30 years corresponds to the biggest boom in equities the world has ever seen. There is no guarantee it will continue, a 100% stock portfolio is very risky to live off of.

4% is the generally accepted safe withdrawal rate (inflation adjusted) from a portfolio for 30 years holding a mix of bond funds and equities. If you’re trying to live off of it for 50-60 years, 3-3.5% is the recommendation.