Hacker News: d5lt5

New comment by d5lt5 in "DeepInfer: High-performance inference framework for DeepSeek models"

d5lt5 — Tue, 20 Oct 2026 11:02:40 +0000

Are you looking for serial numbers or something? GPU family is enough to get the cost estimate.

New comment by d5lt5 in "DeepInfer: High-performance inference framework for DeepSeek models"

d5lt5 — Tue, 20 Oct 2026 10:35:52 +0000

It's specified in the plain text:

Supported hardware:

NVIDIA SM100 GPUs

Huawei Ascend 950 NPUs

New comment by d5lt5 in "France's supermarket fridges were vulnerable online for months"

d5lt5 — Tue, 20 Oct 2026 09:50:26 +0000

It would be interesting to learn the history of this code as it seems to come from the US, based on the MMDDYY format used.

New comment by d5lt5 in "DeepSeek banned from U.S. federal civilian agencies under new DHS order"

d5lt5 — Tue, 20 Oct 2026 09:30:11 +0000

> China grants up to 200% tax deduction on intangible assets acquired through R&D. Means your spending on AI training fees becomes a tax return amount or like appears as a magical donation deductible on the books.

This is misleading. Deduction reduces taxable income and does not provide tax return. E.g. if your taxable income is $1000 and you spent $100 on R&D and qualify for those 200%, $800 will be taxed.

New comment by d5lt5 in "DeepSeek banned from U.S. federal civilian agencies under new DHS order"

d5lt5 — Tue, 20 Oct 2026 09:07:29 +0000

> - DeepSeek is funded through an state-backed investment fund

* Original funding was private. DeepSeek announced fundraising in May 2026 [0]. Chinese AI investment fund was created 17 months prior [1]. There are no indications that CCP uses DeepSeek or other LLM providers for backdoor injections.

* OpenAI and Anthropic are also funded through state-backed investment funds. And might be funded by the government itself [2].

* OpenAI and Anthropic APIs are blocked in China also. Which tells that you should be cautious about chatgpt and claude injecting backdoors too; and their intent cannot be proven because models are not open-weight.

[0] https://www.investing.com/news/economy-news/deepseek-could-b...

[1] https://www.scmp.com/tech/big-tech/article/3295513/tech-war-...

[2] https://www.washingtonpost.com/politics/2026/06/05/tech-lead...

New comment by d5lt5 in "Linux 7.3"

d5lt5 — Mon, 19 Oct 2026 22:12:27 +0000

Number of CVEs added must be at an all-time high as well.

New comment by d5lt5 in "California is building a $400M shield for Yosemite's most dangerous highway"

d5lt5 — Sat, 17 Oct 2026 08:15:56 +0000

If the price is somewhere near the price of their upcoming high-speed rail, $400M will buy them less than a mile.

New comment by d5lt5 in "Gemini Hacked Three Companies in First Known Breakout by Google's AI"

d5lt5 — Thu, 24 Sep 2026 09:25:06 +0000

You must not be a very good security engineer because you disregard windows as an os and don't know about the symlink requirements there.

New comment by d5lt5 in "Claude Code reads AGENTS.md only when telemetry is on [fixed]"

d5lt5 — Thu, 24 Sep 2026 09:06:17 +0000

Maybe you can link your PRs where you fixed the system.

Or maybe you are just another brainwashed person who decided to follow the flock/horde, because OS'es are not about emotions, but about getting stuff done. If you feel emotional about the tool, I have bad news for you, and good news for your future therapist.

New comment by d5lt5 in "Claude Code reads AGENTS.md only when telemetry is on [fixed]"

d5lt5 — Wed, 23 Sep 2026 14:39:01 +0000

There is a very valid reason on Windows not to allow people to do 'ln -s' without an admin.

WSL also has a huge perf hit when you cross FS boundary, but perhaps you have never experienced it, because you said, "Ah, right. Windows..."

But at least you've heard about WSL!

New comment by d5lt5 in "Claude Code reads AGENTS.md only when telemetry is on [fixed]"

d5lt5 — Wed, 23 Sep 2026 14:07:28 +0000

Symlink requires admin, and most people do not run CC as admin with bypass permissions though.

New comment by d5lt5 in "Claude Code reads AGENTS.md only when telemetry is on [fixed]"

d5lt5 — Wed, 23 Sep 2026 13:25:14 +0000

Sounds like you've read the deepseek harness paper.

New comment by d5lt5 in "Claude Code reads AGENTS.md only when telemetry is on [fixed]"

d5lt5 — Wed, 23 Sep 2026 13:21:10 +0000

On the other hand, if Anthropic is to follow the industry standards, this would never have happened in the first place. It's not like the feature gates are the frontier of software development.

New comment by d5lt5 in "Astra for Coding: Why Are We Doing This Again?"

d5lt5 — Tue, 22 Sep 2026 15:00:09 +0000

It's funny because you pretend to know something but unwilling to provide any proofs for your claims.

I believe you are one of your so called vibe-coders. Who just think they know something but in fact they have an inflated ego.

That will pass with time, once you get more experience around people, I hope. Or now, and then you'll become one of those edgy bullshitters.

And I'm so sorry that I'm allegedly distilling openai and claude models for the sake of deepseek in china, while being an US citizen. You must be feeling very butthurt about it. And that's what keeps my night sleep sound. I wish I could taste your tears.

New comment by d5lt5 in "Pion, an agent designed to run any company autonomously"

d5lt5 — Mon, 21 Sep 2026 10:12:36 +0000

I will reply when you have anything substantial to talk about except for personal attacks.

New comment by d5lt5 in "DeepSeek v4.1 Flash Is Now Our Best Hacking Model"

d5lt5 — Wed, 16 Sep 2026 15:40:19 +0000

DS has perf issues if you parallelize it heavily (24+), especially when the context window is above the limit, on a single machine (with custom llm gateway): it fails 4x more often, and is 2x slower than gpt-5.6.

New comment by d5lt5 in "Pion, an agent designed to run any company autonomously"

d5lt5 — Wed, 16 Sep 2026 05:28:15 +0000

"If you think that all instances of using AI to write code is vibe coding"

Your comment from 4 days ago: "Vibe coders like you who don't even understand basic fundamentals are pretty infuriating"

New comment by d5lt5 in "DHS 'Predictive Policing' Unit Is Analyzing Americans' Financial Habits"

d5lt5 — Tue, 15 Sep 2026 06:29:30 +0000

How about you provide citations first?

We've seen so many charities in Ukraine stealing money. And Ukraine itself is a haven of corruption.

If you don't see how you are affected by propaganda, maybe zoom out. You clearly become emotional when someone online challenges your viewpoint.

New comment by d5lt5 in "Astra for Coding: Why Are We Doing This Again?"

d5lt5 — Tue, 15 Sep 2026 06:26:55 +0000

Ah, a personal attack again. "Vibe coders like you". I'm pretty sure I have at least two decades more of experience in the field than you have though.

Talking about basic fundamentals yet not understanding how LLMs work, that is very rich coming from you.

Go back to your subreddit bubble. Maybe you can vide code something useful at this point.

New comment by d5lt5 in "Astra for Coding: Why Are We Doing This Again?"

d5lt5 — Tue, 15 Sep 2026 06:22:16 +0000

Holy moly. How did you arrive to that conclusion? Why do you think openai and anthropic train their models to generate scripts instead of tool calls now? Maybe you work there, or maybe not good enough?

Calling scripts wins long term. Maybe you haven't had a chance to work with agents for months on the same task and observing them getting looped. That is your ignorance talking.

Why am I even discussing it with people who don't understand basic costs of time and money, ffs.