Hacker News: mholm

New comment by mholm in "The AI Great Leap Forward"

mholm — Wed, 08 Apr 2026 20:50:15 +0000

Decent sentiment and analogy, but writing this with AI with hackneyed examples undercuts the point

New comment by mholm in "Project Glasswing: Securing critical software for the AI era"

mholm — Wed, 08 Apr 2026 16:16:29 +0000

This product is explicitly not being released for usage

New comment by mholm in "Claude Code's source code has been leaked via a map file in their NPM registry"

mholm — Tue, 31 Mar 2026 17:23:42 +0000

The only real secret sauce is the training methods and datasets used for refining harness usage. Claude Code is a lot better than gemini-cli/open-code/etc because Claude is specifically trained on how to run in that environment. It's been rlhf'd to use the provided tools correctly, and know the framework in which it operates, instead of relying solely on context.

New comment by mholm in "Judge blocks Pentagon effort to 'punish' Anthropic with supply chain risk label"

mholm — Fri, 27 Mar 2026 16:39:45 +0000

My org gave us 5 months to transition our personal usage away from Anthropic products. Presumably that's paused now, though they haven't made an announcement. I work with medicare/medicaid, and the supply chain risk was directly cited in the initial decision.

New comment by mholm in "Newly purchased Vizio TVs now require Walmart accounts to use smart features"

mholm — Thu, 26 Mar 2026 14:50:11 +0000

That's been the standard for a decade. $200 for a 4k 55" TV is not realistic without a secondary profit source

New comment by mholm in "Hide macOS Tahoe's Menu Icons"

mholm — Sun, 22 Mar 2026 02:30:36 +0000

I don't even have any special handling set and haven't had any Tahoe prompts beyond maybe the first one. I often forget that I'm not on the latest anymore.

New comment by mholm in "Our commitment to Windows quality"

mholm — Fri, 20 Mar 2026 21:24:12 +0000

Not OP, but I've got a friend of a friend in the Windows org that backs this up. Most engineers are teamed up by manufacturers. HP team, Lenovo team, etc. These are the primary drivers of feature development. If it won't sell grandma another $500 HP laptop, they're not interested.

New comment by mholm in "Time to Dump Windows?"

mholm — Fri, 20 Mar 2026 20:40:28 +0000

I think this is the year for me too. I'm in a similar situation as the author: remoting into a windows computer for games (Moonlight + Sunshine). But everything I play is Proton Gold/Platinum, which means I can move pretty seamlessly.

The biggest issues with the platform had previously been one-off errors 'Audio switcher doesn't work' or the annoyances of desktop environments. I don't want to browse stack overflow during my free time. LLMs solve this entirely. Claude solves these problems trivially, and helps me take advantage of the platform without bogging me down on its ux failures.

New comment by mholm in "Spotify playing ads for paid subscribers"

mholm — Wed, 18 Mar 2026 19:14:23 +0000

Doesn't even feel like they've committed code since December 2023

New comment by mholm in "Spotify playing ads for paid subscribers"

mholm — Wed, 18 Mar 2026 19:13:32 +0000

Agree with this; I want to cancel my spotify subscription so badly. I already have Apple One, which includes Apple Music. But the recommendations and UX are so incredibly bad that I still pay for Spotify.

I was doing my yearly attempt at switching over to Apple Music and the 'similar music' radio had somehow saw fit to include Kendrick Lamar with my indie synth. Swapped back to spotify and immediately loved some of its similar suggestions.

New comment by mholm in "Home Assistant waters my plants"

mholm — Mon, 16 Mar 2026 14:55:14 +0000

HA in particular creates a lot of log churn. It's not a 100% certainty, but after running for 4 years I finally had to copy the SD image to a new one because it had become unwritable.

New comment by mholm in "Debian decides not to decide on AI-generated contributions"

mholm — Tue, 10 Mar 2026 18:43:20 +0000

This very much depends on the domain you work in. Small projects in well tread domains are incredible for AI. SaaS projects can essentially be one-shot. But large projects, projects with specific standards or idioms, projects with particular versions of languages, performance concerns, hardware concerns, all things the Debian project has to deal with, aren't 'solved' in the same way.

New comment by mholm in "Uber is letting women avoid male drivers and riders in the US"

mholm — Mon, 09 Mar 2026 18:28:09 +0000

Unfortunately necessary. Essentially every girl I know has had at least one bad experience with a creepy uber driver. These are people that are entering their address and often their workplace into the app. It's a big reason why a lot of my friends are picking Waymos instead.

New comment by mholm in "MacBook Air with M5"

mholm — Tue, 03 Mar 2026 19:51:52 +0000

Fairly short, I'm a Go developer generally working with terraform and microservices. I'd expect some throttling if you're doing 3+ minute compiles, I think. But I think the problem is overblown by the tech video reviewer population that regularly does extremely intensive workloads.

New comment by mholm in "MacBook Air with M5"

mholm — Tue, 03 Mar 2026 19:17:54 +0000

> thermal throttling under sustained heavy load

This gets mentioned a lot, but I do quite a bit of dev work on my M4 MBA and have never even felt it get warm. Sustained heavy loads are extremely rare with how quick this thing is.

New comment by mholm in "Mac mini will be made at a new facility in Houston"

mholm — Tue, 24 Feb 2026 22:28:29 +0000

Most openclaw users are not running the models locally.

New comment by mholm in "I’m joining OpenAI"

mholm — Mon, 16 Feb 2026 19:21:14 +0000

The entire marginal cost to serve AI models is paid for by the API costs of all providers by nearly every estimation. The cost not currently recouped is entirely in the training and net-new infrastructure that they're building.

New comment by mholm in "Kimi Released Kimi K2.5, Open-Source Visual SOTA-Agentic Model"

mholm — Wed, 28 Jan 2026 00:15:44 +0000

Jeff Geerling and a few others also got access to similarly specced mac clusters. They replicated this performance.

The tooling involved has improved significantly over the past year.

New comment by mholm in "Kimi Released Kimi K2.5, Open-Source Visual SOTA-Agentic Model"

mholm — Tue, 27 Jan 2026 15:54:16 +0000

The reason Macs get recommended is the unified memory, which is usable as VRAM for the GPU. People are similarly using the AMD Strix Halo for AI which also has a similar memory architecture. Time to first token for something like '1+1=' would be seconds, and then you'd be getting ~20 tokens per second, which is absolutely plenty fast for regular use. Token/s slows down at the higher end of context, but it's absolutely still practical for a lot of usecases. Though I agree that agentic coding, especially over large projects, would likely get too slow to be practical.

New comment by mholm in "High-speed train collision in Spain kills at least 39"

mholm — Mon, 19 Jan 2026 16:08:31 +0000

Animals become a fine red mist when presented with these sorts of forces. The train feels a bump, but will not crash. I'm unsure at what size a rock will cause issues, but I would expect in most cases they would be kicked away by the train without issue, if a person can move them.