HNHacker News
TopNewBestAskShowJobs

mvkel

3,742 karma · joined March 1, 2011

mykel.org
submissionscomments
mvkel··on Original GrapheneOS responses to WIRED fact checker
If you have zero consideration for other people, sure.

"I can't believe you wrote this terrible code. You clearly don’t understand how concurrency works. Do it again."

Technically right, but when you run out of people who actually want to work with you, you'll be writing the code yourself.

mvkel··on ChatGPT Images 2.0
I wonder if this confirms version 1 of some kind of "world model."

It has an unprecedented ability to generate the real thing (for example, a working barcode for a real book)

mvkel··on Original GrapheneOS responses to WIRED fact checker
Being "right" shouldn't excuse bad behavior, especially if you depend entirely on a community to survive, which we all do.
mvkel··on John Ternus to become Apple CEO
That's like saying going to Washington DC at all is "politics."

It'd be bad business stewardship if he didn't; bad for shareholders, ultimately.

It looks silly optically, but it's plain to see that Apple has trump nailed psychologically. And it worked! They knew what was needed to get an administration to support the company in meaningful ways.

May we do better on the next election.

mvkel··on John Ternus to become Apple CEO
Right. We're not still running Windows Vista, or RedHat. Time for a rethink.
mvkel··on John Ternus to become Apple CEO
Cook is known to be monk-like, so the relative quiet of this announcement is no surprise. Hopefully Ternus takes some risks and revisits some things from scratch (the OS layer)[0] rather than continuing down the path of more service add-ons that Cook seemed to be excitedly geared up for. Personally, it's worth noting that Ternus did -not- directly oversee the Vision Pro, which is encouraging.

[0] As Steve Jobs said in 2005: "OS X is the most advanced operating system on the planet and it has set Apple up for the next 20 years."

How incredibly prophetic that 21 years later, MacOS is suddenly showing its age.

mvkel··on Vercel April 2026 security incident
That's only looking at half of the equation.

That lack of diversity also makes patches more universal, and the surface area more limited.

mvkel··on NASA Shuts Off Instrument on Voyager 1 to Keep Spacecraft Operating
Just have to nuke it into orbit!
mvkel··on Opus 4.7 to 4.6 Inflation is ~45%
The cope is real with this model. Needing an instruction manual to learn how to prompt it "properly" is a glaring regression.

The whole magic of (pre-nerfed) 4.6 was how it magically seemed to understand what I wanted, regardless of how perfectly I articulated it.

Now, Anth says that needing to explicitly define instructions are as a "feature"?!

mvkel··on Claude Opus 4.7 Model Card
I'm talking about gross margins, not revenue.

It's well known that GPT-4 is much more expensive to operate than the GPT-5 family.

Of course they won't drop the prices; it's pure profit if they make models more efficient.

mvkel··on Claude Opus 4.7 Model Card
It seems to be a rule that older models are more expensive than newer ones. The low end models have higher $CPT and worse output. I wonder if the move is to just have one model and quantize if you hit compute constraints
mvkel··on Elevated errors on Claude.ai, API, Claude Code
OpenAI sees an opportunity and is happy to set money on fire to have an edge over Anth. No issues
mvkel··on Elevated errors on Claude.ai, API, Claude Code
I use superpowers with both and have found the plan generation in codex to be a bit more thorough, so it's not the native planning mode necessarily
mvkel··on Elevated errors on Claude.ai, API, Claude Code
Better at planning, worse at execution. Ultimately, creates a working product.
mvkel··on Daily Claude outage is upon us. Waiting for Claude Status to update
500s as far as the eye can see
mvkel··on If you started a company two years ago, many assumptions are no longer true
Interestingly, this was always true, it's just made more obvious when it takes less time to bring a product to a potential customer.

Launching a product was never the finish line; it was always the start line. But technical founders could trick themselves into thinking that building a product was building a business.

The same "Lean Startup" rules apply. Build something and get it in front of real people who will pay you for the thing. If they won't, back to the drawing board.

The only real "shift" I've seen is that most startups don't actually need VC at all. That's a great thing.

mvkel··on Pro Max 5x quota exhausted in 1.5 hours despite moderate usage
Then my original question stands: why did this become an issue seemingly overnight if nothing changed?
mvkel··on Pro Max 5x quota exhausted in 1.5 hours despite moderate usage
Why did it suddenly become an issue, despite prompt caching behavior being unchanged?
mvkel··on Pro Max 5x quota exhausted in 1.5 hours despite moderate usage
Why did this become an issue seemingly overnight when 1M context has been available for a while, and I assume prompt caching behavior hasn't changed?

EDIT: prompt caching behavior -did- change! 1hr -> 5min on March 6th. I'm not sure how starting a fresh session fixes it, as it's just rebuilding everything. Why even make this available?

It feels like the rules changed and the attitude from Anth is "aw I'm sorry you didn't know that you're supposed to do that." The whole point of CC is to let it run unattended; why would you build around the behavior of watching it like a hawk to prevent the cache from expiring?

mvkel··on Artemis II safely splashes down
Landing on Mars is the artifact of all of the innovation required to get to Mars. We benefit from the innovations, not the landing per se.

Memory foam, smart phone cameras, tech miniaturization in general, GPS, baby formula, cordless tools... just a tiny sliver of things we use daily that are directly attributable to the pursuit of space travel.

It is far from useless

mvkel··on Native Instant Space Switching on macOS
One of the examples I gave is springboard behavior which is very much OS-level
mvkel··on Artemis II safely splashes down
I am trying hard to keep a positive attitude about this mission but I keep feeling like it's vanity marketing for America, more than science, or pushing the frontier. "Hey everyone, remember when we got to the moon FIRST? Good times." Ultimately, we did all of this a half century ago. The lasting impression is a reminder of how underfunded the space program has been all these decades. Why go to the moon again? The answer in the 60s was: because it's there. And that was enough. But now? Is it -really- a training ground for Mars?
mvkel··on The Zettelkasten method in Obsidian
Agreed in theory, but ultimately it's just bikeshedding.
mvkel··on Molotov cocktail is hurled at home of Sam Altman
Do you know him personally?
mvkel··on The Zettelkasten method in Obsidian
Systems like these made sense in the pre-AI era, where things needed to be organized at the outset to be useful later.

With AI, there's nothing stopping you from dumping a huge pile of information into a single folder and telling an AI what you want to make with it that day.

mvkel··on Native Instant Space Switching on macOS
The -only- time I experience it infuriatingly is in iOS natively.

For example: on the Home Screen, open a folder, then tap outside it to bring back the Home Screen. While the animation is happening, try opening an app. It won't.

Open the camera app. Swipe between modes and press the shutter button. It won't work until the animation completes.

An especially egregious one: open messages. Tap the + sign, and tap the letter P while it's animating. I had to tap the letter 10 times before it finally started showing Ps.

The OS is riddled with them.

mvkel··on How NASA built Artemis II’s fault-tolerant computer
If you look at code as art, where its value is a measure of the effort it takes to make, sure.
mvkel··on Native Instant Space Switching on macOS
This is true in iOS, too. Taps are ignored until any animation completes. Must be deep in the code!
mvkel··on Iran demands Bitcoin fees for ships passing Hormuz during ceasefire
This is like Legal 101. It's not that complicated.
mvkel··on System Card: Claude Mythos Preview [pdf]
This is Anth's typical marketing playbook, a hat tip to their so-called "safetyist" roots, a differentiator against OpenAI's more permissive access[0]. Coke vs. Pepsi.

"We made a model that's so dangerous we couldn't possibly release it to the public! The only responsible thing is so simply limit its release to a subset of the population that coincidentally happens to align with our token ethos."

The reality is they just don't have the compute for gen pop scale.

They did this exact strategy going back several model versions.

[0] ironically, OpenAI has some pretty insane capabilities that they haven't given the public access to (just ask Spielberg). The difference is they don't make a huge marketing push to tell everyone about it.

← PreviousPage 5 of 34Next →