Hacker News: pshirshov

New comment by pshirshov in "Hollywood's spent 25 years chasing the wrong lessons from The Lord of the Rings"

pshirshov — Sun, 05 Jul 2026 23:06:24 +0000

We live in the times when Jackson's LoTR, which was not faithful to the source, is considered a gold standard of "humanity"?

New comment by pshirshov in "“Beyond the limit”: Satellites and mirrors in space pose threat to the night sky"

pshirshov — Sat, 04 Jul 2026 19:03:02 +0000

> we will build Sun orbital structures and swarms,

Another episode of arrogant fantasy in the ponyworld.

New comment by pshirshov in "Synthesis is harder than analysis"

pshirshov — Sat, 04 Jul 2026 17:22:55 +0000

> You will soon discover that, unlike in Cal 1, there is no algorithm for computing the integral of an arbitrary function

There is this thing though: https://en.wikipedia.org/wiki/Risch_algorithm , the best practical implementation was Axiom CAS

New comment by pshirshov in "Why Switzerland has 25 gbit internet and America doesn't"

pshirshov — Fri, 03 Jul 2026 09:08:58 +0000

Irish model is similar (shared ducts, backbone networks which are mandated to provide connectivity to ISPs, including competitors). But Irish market is shit. All is done over pppoe, hard to get fair dual-stack, assymetric shaping (5gbit down, 50mbit up), hidden 10tb/month limits, etc.

New comment by pshirshov in "I Don't Maintain My Homelab"

pshirshov — Wed, 01 Jul 2026 14:21:36 +0000

Good for you. That doesn't necessarily mean your experience is representative enough though.

Unifi stronk. Noone needs working ipv6 or 2+ gigabit pppoe throughput and many other things, like an ability to assign a name to an entry in embedded radius server.

New comment by pshirshov in "I don't maintain my homelab, it maintains itself"

pshirshov — Wed, 01 Jul 2026 11:02:41 +0000

> UniFi supports automatic and scheduled updates,

Yeah, right until the moment it bricks after an update.

New comment by pshirshov in "Founding a company in Germany: €9600, 152 days and I still can't send an invoice"

pshirshov — Sat, 27 Jun 2026 19:05:57 +0000

Not necessarily. You need a local director and to be physically present in the country like once per year for board meetings.

New comment by pshirshov in "AI learns the “dark art” of RFIC design"

pshirshov — Sat, 27 Jun 2026 18:46:21 +0000

Chips? I've tried to task Opus, Gemini and Codex with a simple PCB. All of them placed holes correctly but can't understand that the traces should not cross physically.

New comment by pshirshov in "Founding a company in Germany: €9600, 152 days and I still can't send an invoice"

pshirshov — Wed, 24 Jun 2026 14:18:15 +0000

Well, create a company in Ireland. 5 days, €300 and you can send your invoices.

New comment by pshirshov in "AMD silently removes memory encryption from consumer Ryzen CPUs"

pshirshov — Thu, 18 Jun 2026 12:09:37 +0000

Well, I've experimented with this feature on several platforms (both ECC and non-ECC) starting with TRX40, most of the times I've been just getting hard freezes at GPU driver initialization. If it boots - it usually hangs when a VFIO VM spins up.

New comment by pshirshov in "AMD silently removes memory encryption from consumer Ryzen CPUs"

pshirshov — Thu, 18 Jun 2026 12:06:31 +0000

Well, I live in Ireland but not sure what you refer to.

Something being illegal does not imply it doesn't happen though.

New comment by pshirshov in "AMD silently removes memory encryption from consumer Ryzen CPUs"

pshirshov — Thu, 18 Jun 2026 11:23:20 +0000

This feature was off by default in all the mobos I've seen.

It causes many stability issues, as to my experience.

The attack is sophisticated, Mr.Nobody, generally, should not worry about expensive cryogenic attacks - three letter guys would extract your key with a wrench.

I mean the change is bad - it undermines already damaged trust, but the "average Joe" is extremely unlikely to be affected directly.

There are many much cheaper ways to force you to give up your keys.

New comment by pshirshov in "AMD silently removes memory encryption from consumer Ryzen CPUs"

pshirshov — Thu, 18 Jun 2026 10:52:08 +0000

To be honest it never worked great - many issues (mostly freezes) with VFIO, NVidia drivers, amdgpu...

New comment by pshirshov in "AI coding at home without going broke"

pshirshov — Sat, 13 Jun 2026 18:43:00 +0000

> and the hardware you buy today may look like a bad bet in a year.

3090s and 7900s are going well so far.

Next year an Arc Pro B70 won't produce you less tokens than today.

They aren't fast but if you have flows where you can make money with them - they are a bargain in terms of price per Gb.

New comment by pshirshov in "Claude Fable is relentlessly proactive"

pshirshov — Sat, 13 Jun 2026 12:21:21 +0000

Yes, I know all the flaws. As I said, it's not an objective way to measure performance of a model - but it is intended to produce something that only humans could mesaure. The goal is for you to being able to play the game and judge - and fill the human checklist for yourself if you wish.

You didn't get why the automatic review scores are there - all of the reviewers, including Fable, happily assign highest scores to code which can't even run. In my opinion that is a sort of an empirical evidence that these models are very far from the "AGI" state.

Anyway, while I didn't explain the methodology and the purpose of this experiment, I have something material to discuss. The "awesome Fable" claims are not material at all.

Can you bring something clearly showcasing Fable's superiority?

New comment by pshirshov in "Claude Fable is relentlessly proactive"

pshirshov — Fri, 12 Jun 2026 19:19:02 +0000

But can you bring anything measurable in support to your words? I did.

New comment by pshirshov in "Claude Fable is relentlessly proactive"

pshirshov — Fri, 12 Jun 2026 17:09:44 +0000

Ok, explain me one thing: I have a benchmark - I feed identical prompt to multiple models. Codex produces a rough but working program. Fable produces the same - but with more bugs than Codex. Opus produces something similar to Codex but with a critical bug.

That describes all my tests with Fable.

Why should I be hyped about all that "legitimate power" if the model performs on par with two other SoTAs?

I mean, well, yes, it is impressive. It could quickly generate a lot of garbage which sorta does look like code. Two others can do the same. I don't see any groundbreaking improvement - but the price is much higher. Why the hype?

New comment by pshirshov in "Claude Fable is relentlessly proactive"

pshirshov — Fri, 12 Jun 2026 15:31:26 +0000

I have a feeling like such posts come from a parallel reality. In my anecdotal experience confirmed by my (still subjective) benchmark (https://pshirshov.github.io/llm-bench-pi-oneshot/) Fable is not _that_ impressive. I performs on par with gpt-5.5 and opus 4.8, sometimes better, sometimes worse, it's definitely more expensive and it likes to refuse answering questions about React saying it can't help with chemistry.

Is this fuss really grounded or it's some pre-IPO AGI hype?

New comment by pshirshov in "If Claude Fable stops helping you, you'll never know"

pshirshov — Wed, 10 Jun 2026 08:11:25 +0000

Fable refused to answer some questions about React citing limitations on chemistry and biology.

New comment by pshirshov in "Solar Energy Saves Europeans $135M a Day"

pshirshov — Tue, 09 Jun 2026 20:52:53 +0000

Over last several years the cost to install 1kW was, ehm, less than €200 if you do it the retail way - 2..4 500W panels from the warehouse next street and a microinverter.

Thanks, China.

In addition to my primary system, I have a toy installation with 6 500W panels and a micro. I paid around 500 euros for that last August and by this time this toy installation generated me 1281 kWh. That's around €400 in terms of retail energy prices. So, I'll break even this August.

So, everything generated after would be exactly the "money saved".