Daily Tech News 11 October 2026

Top Story

AI models and agents are security risks and should be banned from your corporate network, or at a minimum, air gapped and restricted to a lead lined Faraday cage biocontainment unit to preclude Van Eck phreaking and accidental bowls of bat soup. (Twitter)

Says... Microsoft CEO Satya Nadella.

Tech News

Why did already massively inflated prices on the RTX 5090 suddenly spike recently? If you guessed that Nvidia stopped making them, you're probably correct. (Tom's Hardware)

They will be replaced - allegedly - by a 5080 model upgrade to 24GB of RAM, delivering about half the performance. The current 16GB 5080 will also cease to be.
Ukrainian drones blew up the largest datacenter in Russia, or at least unspecified parts thereof. (Tom's Hardware)

Has the volume of obnoxious spam declined at all?
Testing Windows 7 vs. Windows 11 for gaming. (Tom's Hardware)

Interesting but not startling comparison, unless you were somehow expecting Windows 11 to take the win, in which case you may well be startled.
My emulated fantasy 10-bit 80s home computer has two synthesizers: A wavetable synthesizer (something rare in home computers of any period but the Apple IIgs from 1986 has this), and a simpler fixed-function digital synthesizer based on a limited set of waveforms, easier to implement but needing multiple voices to work together to produce appropriate harmonics that a wavetable synth generates directly.

That second synthesizer sucked yesterday; as of today with tuning against real instrument samples it's sounding like a Yamaha synthesizer that doesn't have an insect in its power supply but in some cases needs one. Cymbals last I checked still sounded like gongs, resonating freely when struck. Gongs also sounded like gongs, which in that case is good.

Musical Interlude

Disclaimer: Well, that explains it.

Posted by: Pixy Misa at 04:00 AM

Comments

(Jump to bottom of comments)

1 !!

Posted by: clarence at October 11, 2026 04:01 AM (zGeQ7)

2 Informed the commentariot.

Posted by: clarence at October 11, 2026 04:02 AM (zGeQ7)

3 Disclaimer: Well, that explains it.
****

Well, if you are already crazy, sure.

Posted by: clarence at October 11, 2026 04:04 AM (zGeQ7)

4 Inside the park?

Posted by: clarence at October 11, 2026 04:05 AM (zGeQ7)

5 Slow night. I even read the content.

Posted by: clarence at October 11, 2026 04:06 AM (zGeQ7)

6 Were you using that music clip to tune your synthesizer?

Posted by: clarence at October 11, 2026 04:08 AM (zGeQ7)

7 w00t

Posted by: m at October 11, 2026 04:09 AM (odBC9)

8 Clarence you are roaming around by yourself?

Posted by: Pete Bog at October 11, 2026 04:10 AM (Ys0s/)

9 poke...

Posted by: clarence at October 11, 2026 04:11 AM (zGeQ7)

10 Well, not the first time I had the blog all to myself...not even the third.

Posted by: clarence at October 11, 2026 04:12 AM (zGeQ7)

11 Way to go, clarence!

Posted by: m at October 11, 2026 04:14 AM (odBC9)

12 clarence, m, and Pete Bog walked into a tech thread....

Posted by: m at October 11, 2026 04:14 AM (odBC9)

13 Halp us, Pixy! You're our only hope.

Posted by: m at October 11, 2026 04:15 AM (odBC9)

14 Says... Microsoft CEO Satya Nadella.
****

Sounds like they did something even more monumentally stupid than their previous screwups and are trying to get ahead of it. Just like all the previous mistakes they did not learn anything.

Posted by: clarence at October 11, 2026 04:17 AM (zGeQ7)

15 14 Says... Microsoft CEO Satya Nadella.
****

Sounds like they did something even more monumentally stupid than their previous screwups and are trying to get ahead of it. Just like all the previous mistakes they did not learn anything.
Posted by: clarence at October 11, 2026 04:17 AM (zGeQ7)

They're saying "Don't make me hit you again"?
I'll go read it.

Posted by: m at October 11, 2026 04:26 AM (odBC9)

16 BOING!

Posted by: Biden's Dog sniffs a whole lotta malarkey, at October 11, 2026 04:28 AM (1o8D5)

17 Has the volume of obnoxious spam declined at all?
****

No. And it will not because someone started a breeding program.

Posted by: clarence at October 11, 2026 04:29 AM (zGeQ7)

18 from the Satya Nadella 𝕏 post:

"As traditional software systems were being deployed across the economy over the last few decades, we had the tools and capability to trace behaviors to a specific code path."

"We" =

Posted by: m at October 11, 2026 04:30 AM (odBC9)

19 clarence, m, Pete Bog, and Biden's Dog walked into a tech thread....

Posted by: m at October 11, 2026 04:31 AM (odBC9)

20 re: Satya Nadella 𝕏 post:

The commenters (I can only see three) are insinuating that he is a PDT shill. He's referring to Super Intelligence instead of Artificial Intelligence.

Posted by: m at October 11, 2026 04:33 AM (odBC9)

21 Nice and roomy in here!

Posted by: m at October 11, 2026 04:33 AM (odBC9)

22 we had the tools and capability to trace behaviors to a specific code path.
****

That 'we' is the size of an Olympic champion weight lifter.

Posted by: clarence at October 11, 2026 04:34 AM (zGeQ7)

23 from the Satya Nadella 𝕏 post:

That same kind of mechanistic understanding eludes us in today’s Super Intelligence systems ... We can’t treat Super Intelligence as a set of nested black boxes and simply accept or reject its recommendations, answers, and actions. ... The most trustworthy Super Intelligence system will not be the one with the model we trust most.

Posted by: m at October 11, 2026 04:36 AM (odBC9)

24 Great minds, clarence!

18 from the Satya Nadella 𝕏 post:

"As traditional software systems were being deployed across the economy over the last few decades, we had the tools and capability to trace behaviors to a specific code path."

"We" =
Posted by: m at October 11, 2026 04:30 AM (odBC9)

22 we had the tools and capability to trace behaviors to a specific code path.
****

That 'we' is the size of an Olympic champion weight lifter.
Posted by: clarence at October 11, 2026 04:34 AM (zGeQ7)

Posted by: m at October 11, 2026 04:37 AM (odBC9)

25 We're gonna get a prize. I just know we're gonna get a prize.

Posted by: m at October 11, 2026 04:38 AM (odBC9)

26 Wow. Google "Guilfoyle" and you'll see the knives coming out.

Posted by: m at October 11, 2026 04:41 AM (odBC9)

27 Now shifting work from the synthesizers to the tracker, which sounds unnaturally perfect with 18 instruments starting each note at the same ~50 microsecond tick resolution with the waves exactly in phase.

Posted by: Pixy Misa at October 11, 2026 04:43 AM (BLOW1)

28 Mornin'

Posted by: Puddleglum at work at October 11, 2026 04:46 AM (7Vi0c)

29 clarence, m, Pete Bog, Biden's Dog, and Puddleglum walked into a tech thread....

Posted by: m at October 11, 2026 04:47 AM (odBC9)

30 17 Has the volume of obnoxious spam declined at all?
****

No. And it will not because someone started a breeding program.
Posted by: clarence

Spam AI will be the death of everything!! We're doomed.

Posted by: Puddleglum at work at October 11, 2026 04:50 AM (7Vi0c)

31 There once was a blogger from Aus
Who feigned the sharpen his claws
When given the chance
To give up or dance
He said watch out for my paws

fin

Posted by: clarence at October 11, 2026 04:54 AM (zGeQ7)

32 Has the volume of obnoxious spam declined at all?

Can we hire them to take out Paki and Indian spam/scam call centers too?

Posted by: jim (in Kalifornia) at October 11, 2026 05:00 AM (l3cgK)

33 Wow. Google "Guilfoyle" and you'll see the knives coming out.

Posted by: m at October 11, 2026 04:41 AM (odBC9)
-

Yeh. Just look what those plastic surgeons did to her!

Posted by: Biden's Dog sniffs a whole lotta malarkey, at October 11, 2026 05:02 AM (1o8D5)

34 clarence, m, Pete Bog, Biden's Dog, Puddleglum, and jim (in Kalifornia) walk into a tech thread....

Posted by: m at October 11, 2026 05:13 AM (odBC9)

35 33 Wow. Google "Guilfoyle" and you'll see the knives coming out.

Posted by: m at October 11, 2026 04:41 AM (odBC9)
-

Yeh. Just look what those plastic surgeons did to her!
Posted by: Biden's Dog sniffs a whole lotta malarkey, at October 11, 2026 05:02 AM (1o8D5)

ooohhhh

Posted by: m at October 11, 2026 05:14 AM (odBC9)

36 "The most trustworthy Super Intelligence system will not be the one with the model we trust most."
Posted by: m at October 11, 2026 04:36 AM (odBC9)

So then it's just like talking to any other nutsack with an opinion? And from here on out will only get worse?

....

So why did we put all this effort into them?

Posted by: Warai-otoko at October 11, 2026 05:15 AM (aKiSo)

37 What's COT?

Posted by: Additional Blond Agent at October 11, 2026 05:25 AM (iIyc0)

38 What's COT?

Posted by: Additional Blond Agent at October 11, 2026 05:25 AM (iIyc0)
-

$20, same as in town.

Posted by: Biden's Dog sniffs a whole lotta malarkey, at October 11, 2026 05:26 AM (1o8D5)

39
g'mornin', 'rons

Posted by: AltonJackson at October 11, 2026 05:27 AM (AMvSw)

40 CoT = Chain of Thought

Posted by: SpeakingOf at October 11, 2026 05:30 AM (6ydKt)

41 Chain-of-Thought (CoT) is a prompt engineering technique used in Large Language Models (LLMs) that enhances reasoning capabilities by guiding the model to generate intermediate reasoning steps before producing a final answer, rather than jumping directly to a conclusion. This method is particularly effective for complex tasks involving multi-step logic, such as arithmetic, common sense reasoning, and symbolic manipulation, by breaking problems down into manageable, sequential logical components.

Posted by: SpeakingOf at October 11, 2026 05:31 AM (6ydKt)

42 There are several variations of this approach, including:
Few-Shot CoT, which uses worked examples with explicit reasoning chains to guide the model.
Zero-Shot CoT, which uses simple instructions like "Let's think step by step" to elicit reasoning without prior examples.
Automatic CoT, where the model generates its own reasoning examples to serve as prompts for subsequent tasks.
While CoT significantly improves accuracy and transparency in reasoning models (such as OpenAI's o1 series), it is important to note that it is primarily a prompting strategy rather than a distinct model architecture, and it may increase computational costs and response times.

Posted by: SpeakingOf at October 11, 2026 05:32 AM (6ydKt)

43 I think it sort of kind of gets to a fundamental philosophical problem that was always part of computing, even in the dark nasty algorithmic days.

Can you trust the output of your script?

With linear, deterministic programming, you're still at the mercy of some overarching questions. Was that the right algorithm? Did I ask the right questions when I was making that algorithm? If someone else made it, did they? What assumptions are being made that I might not be aware of?

Moving meat-based problem solving to algorithmic computers didn't make the central questions of epistemology go away. Moving it to non-algorithmic processes like machine learning and "AI" doesn't make it go away. One day, if we can ever move it all to q-bits, with whatever algorithms or state functions involved in the application layer between a meat-bag asking a question and a bit-bag churning out an answer, it still won't go away.

How do we know what we know? How do we use what we think we know? It's a philosophical question at its core.

Posted by: Warai-otoko at October 11, 2026 05:32 AM (aKiSo)

44 G'Day everyone

Posted by: Skip at October 11, 2026 05:43 AM (Ia/+0)

Processing 0.0, elapsed 0.01 seconds.
14 queries taking 0.0048 seconds, 52 records returned.
Page size 39 kb.
Powered by Minx 0.8 beta.