Hacker News: sampullman

New comment by sampullman in "DeepInfer: High-performance inference framework for DeepSeek models"

sampullman — Tue, 20 Oct 2026 12:14:33 +0000

Every open source release is visible to me, too. Is your point that the government could stop it if they wanted? That's true of every government, but it's not a problem until it actually happens.

New comment by sampullman in "Ukraine claims first drone evacuation of wounded soldier"

sampullman — Tue, 20 Oct 2026 08:47:37 +0000

Presumably less likely to get destroyed than a suicide drone, or one sent further into enemy controlled areas.

There probably aren't a lot of scenarios where it makes sense sense to pick an injured guy up with a drone, but it seems like a plausible last resort option.

New comment by sampullman in "DeepSeek banned from U.S. federal civilian agencies under new DHS order"

sampullman — Tue, 20 Oct 2026 04:27:49 +0000

You cannot. See the sibling commenter's link to "Reflections on Trusting Trust".

New comment by sampullman in "Grieving the loss of details"

sampullman — Sat, 10 Oct 2026 17:29:45 +0000

Know what to ask is about as important as the answer. You can't do that very well without domain knowledge.

I'm not sure how it plays out in 5 or 10 years, but that's how it is now.

New comment by sampullman in "Anthropic AI model submits false tip on unsolved Philly murder"

sampullman — Sat, 10 Oct 2026 05:29:10 +0000

You could use one of those browser add-ons to replace "steelman" with "supports" everywhere.

New comment by sampullman in "Ask HN: What do you run on a $5 VPS that's worth keeping online 24/7?"

sampullman — Fri, 09 Oct 2026 18:38:29 +0000

I think you are overestimating how decentralized the web is. Tailscale (could) collect a different kind of data than e.g. Google, and I believe it's far less harmful for the average user. If you are a journalist or persecuted by a government then sure, Tailscale is bad news.

But if you think Tailscale is bad for mobile devices because they might be able to tell that you're near another Tailscale user, that is absurd. If you have a need to be physically untrackable, the only solution is to leave your tracking device at home.

New comment by sampullman in "The Hetzner Cloud network stack – history and technical overview"

sampullman — Fri, 09 Oct 2026 16:12:49 +0000

It's not so true anymore, but when they were cheaper they probably attracted a ton of bad actors (spammers, phishing sites, etc). I assume the KYC requirements were a response to that.

New comment by sampullman in "Ask HN: What do you run on a $5 VPS that's worth keeping online 24/7?"

sampullman — Fri, 09 Oct 2026 11:23:43 +0000

I ran Headscale for a while, and now wireguard directly, but never really considered it to be more private.

What is Tailscale getting that my ISP and google/Facebook via their pixels and tracking scripts aren't?

New comment by sampullman in "Yes, and"

sampullman — Fri, 09 Oct 2026 07:01:55 +0000

You can build lots of interesting things now without understanding the code, but the maximum complexity of what you build will always be defined by what the latest model and harnesses are capable of without architectural guidance.

New comment by sampullman in "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"

sampullman — Thu, 08 Oct 2026 20:14:08 +0000

On design tasks too, for me.

New comment by sampullman in "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"

sampullman — Thu, 08 Oct 2026 20:13:07 +0000

Do you mean Fable 5.1? Or Opus 5.5? I'm not sure what you're working on but for me DS 4.1 flash isn't nearly at their level. For the price it's obvious very impressive, though Luna 6.0 is excellent too.

New comment by sampullman in "God of War on PSP, recompiled to WebAssembly and running in the browser"

sampullman — Thu, 08 Oct 2026 06:22:46 +0000

I agree for the most part, although I think there is also an issue of laziness. There are people who should know better that are submitting unreviewed garbage.

But from the sound of it, you are not agreeing with the sibling comment, because you still have to review the code to make sure it's not slopping up the codebase or taking the architecture in an insane direction.

New comment by sampullman in "Tell HN: GitHub refuses to remove cracked copies of my software after a month"

sampullman — Thu, 08 Oct 2026 06:18:34 +0000

You can find it in about 1/10 the time it took to write this comment.

New comment by sampullman in "God of War on PSP, recompiled to WebAssembly and running in the browser"

sampullman — Thu, 08 Oct 2026 06:17:20 +0000

You don't know what I do, and fortune 500 doesn't scream "excellent software engineering practices".

I also know some people at Google, and they are generally in the same boat as me - mostly using agents to generate code, but not blindly merging slop (yet).

30% of code merged without even glancing at it seems unlikely to me, but I can't disprove the number you pulled out of thin air. For side projects I bet it's far higher. For serious projects I have my doubts. It will get there, but it hasn't yet.

New comment by sampullman in "Zerobrew: A faster alternative to Homebrew"

sampullman — Thu, 08 Oct 2026 06:11:04 +0000

If you ask an AI agent to make some nice benchmarks for your readme, it will often include extra comparisons that aren't particularly relevant.

New comment by sampullman in "God of War on PSP, recompiled to WebAssembly and running in the browser"

sampullman — Wed, 07 Oct 2026 20:24:43 +0000

You are in a bubble or just happen to be at a place that doesn't care about architecture or quality code. I vibe lots of code, but most of it would never make it past review without significant editing and manual guidance.

You can't blame open source projects for rejecting PRs out of hand, when they have a hundred thousand lines of garbage. Not everybody is ready for their project to completely descend into slop yet. Maybe in a couple years though.

New comment by sampullman in "Vibecoding isn't as fun as writing code by hand"

sampullman — Wed, 07 Oct 2026 08:24:05 +0000

I didn't say it's natural. It was unnatural when the internet started taking over, when humans, first shifted to office work, and for every other major technological advancement before that.

I think the cat was out of the bag as soon as OpenAI proved that scaling up transformers was effective. I don't think the way it's playing out is particularly good for normal people, but the advancement of these models can't be stopped.

The light at the end of the tunnel is hardware advancement and open models. In 5 or 10 years if we can run an Opus 5.5+ model on a local device, things might not turn out so bad.

New comment by sampullman in "Xbox has secured exclusive GTA 6 streaming rights"

sampullman — Wed, 07 Oct 2026 05:21:53 +0000

Maybe? But then you have to defend every other hobby from the same argument.

New comment by sampullman in "Tell HN: GitHub refuses to remove cracked copies of my software after a month"

sampullman — Wed, 07 Oct 2026 05:16:11 +0000

The website is public, you can verify yourself.

New comment by sampullman in "Vibecoding isn't as fun as writing code by hand"

sampullman — Tue, 06 Oct 2026 19:42:00 +0000

Yes, you don't need anyone's permission to be sad.

I enjoy the act of coding, but my ambition to create almost always exceeded my ability and the number of hours in the day. So the transition from coder (and sometimes manager) to agent orchestrator has been fairly positive.

It's not necessarily easy or pleasant to evolve, but that's life.