News from KW48.2025

The news of the week in retrospect, short and crisp baked here your nine posts from the week 24.-30.11.2025 (KW48). The cookie season typically garnished with fruits, nuts, icing sugar and a little opinion. Have fun!

Article 1

Cyber alert level Red! The Bundestag and the Bundesrat agree to NIS 2: That's changing for you!

We have news from the legislation that affects hundreds of thousands of companies in Germany, namely the topic of cyber security.

After a long, long wait and even an infringement procedure by the EU, it finally happened:

The Bundestag approved the NIS-2 Implementation Act two weeks ago (13.11), The Federal Council followed suit today.!

What is NIS 2 and why is it important?

NIS 2 (stands for Network and Information Security) is an EU directive that more resilient against cyberattacks. The Commission sees the risk of digital and physical attacks on critical infrastructure as one of the biggest problems for our economy.

Therefore, the goal is clear: Stricter security measures and faster reporting requirements for many more companies.

The Federal Council and the Bundestag have now cleared the way. In the debate, Federal Minister of the Interior Alexander Dobrindt (CSU) stressed that we are in a constant state of Stress test through hybrid warfare – and NIS 2 is the answer to create ‘security through stability and strength’.

Who is affected?

The number of affected organizations is soaring!

While the old BSI law included only about 4,500 organizations, it is estimated that around 29,850 German companies from areas such as:

energy
health
transportation
Digital services
... and many more

It will be governed by the NIS-2 rules.

Important innovation: After initial criticism, the new, stricter requirements now also apply to the Federal administration itself – which was welcomed by many (including CDU politician Marc Henrichmann) as absolutely necessary. The rules apply not only to the economy, but also to the state itself!

What do you have to do? The new obligations

If you are one of the affected companies, you have to massively tighten your IT security measures. These include, but are not limited to:

Risk analysis perform.
Contingency plans and Backup concepts create.
Encryption solutions introduce.

Particularly important: The new three-stage reporting regime in cyber-attacks.

Within 24 hours After discovery: An initial report must be submitted to BSI (Federal Office for Information Security).
Within 72 hours: A step in the way of the incident.
Within one month: A final report.

The BSI is given more supervisory powers here, and a central coordinator for information security is appointed in the federal administration (CISO Bund) is established. BSI President Claudia Plattner sees this as a "an important milestone towards a resilient cybernation".

When does NIS 2 become mandatory?

The EU deadline for implementation was already in place. 8 October 2024 This is why the infringement proceedings against Germany are still ongoing. In theory, we still face fines!

Although the Bundestag (13 November 2025) and the Bundesrat (21 November 2025) have now agreed, the law is still in force. Not in force!

They are still missing Promulgation in the Federal Law Gazette and the entry into force on the date specified in the law. Only from this date will the strict NIS-2 rules be binding on German companies.

So: The political hurdles have been removed, but Preparation is now the be-all and end-all! Check today to see if your business is affected and what action you need to take!

Article 2

AI giants in duel: Anthropic counters Google with Claude Opus 4.5, smarter, faster, better in Excel!

As soon as Google has launched its Gemini 3 and the AI world in turmoil, comes the next timpani blow: Anthropic counters and introduces its new top model: Claude Opus 4.5!

The AI race of the giants is entering the next round, and Anthropic leaves no doubt: They claim the throne.

The new AI king?

Anthropic says very clearly: Opus 4.5 is "the world's most powerful model for programming, autonomous agents and computer control."

This is a battle announcement, because they claim that their new model outperforms Google's Gemini 3 in various ways. Programming categories.

But Claude should not only be better at complex code tasks, but also in everyday life:

Daily tasks: Better working with tables, in-depth research and slide creation.

Complexity: The model is better with ambiguities be able to avoid and weigh up compromises without constant assistance. Internal tests show that Opus 4.5 ‘finds the solution to complex multi-system errors’.

For you, this means: What was recently considered almost insoluble is now within reach!

Claude comes to the desktop and in Excel!

One of the most exciting updates is the new tools that make working with Claude much easier:

Claude for Excel: A chat is integrated directly into your Microsoft spreadsheet program. So you can ask for information about table contents directly and have them edited – very convenient for everyone who works with data on a daily basis! (Available to Max, Team and Enterprise users).

Claude for Chrome & Desktop: New ways to use Claude directly in the browser or on the desktop - ideal for so-called ‘longer-running agents’ means:. (Claude for Chrome requires the paid Claude Max plan).

There is also a new developer ‘Effort parameter’ in the API with which they can control, How intense the model should ‘think’ about a task.

A small step towards more security

Anthropic also addresses important security concerns, particularly with regard to: Prompt injection attacks.

This hides malicious code or text in a data source that the AI model reads. This could instruct it to circumvent its own security precautions (e.g. revealing personal data).

Anthropic claims Opus 4.5 is "more difficult to deceive than any other breakthrough model in the industry". But beware: Of course, it is still not completely immune.

Good news at the end: If you are already a Max or Team Premium user, you will get an increase in the general usage limits so that you can easily use Opus 4.5 for your daily work. And the opus-specific caps have also been lifted!

Article 3

Halleluja HU Now also digital in new territory! KBA makes the TÜV report divisible

Moin to all motorists and screwdrivers! In a world where we can send emails, pay bills online and lock our car via app, there was a bastion of German paper love: the main investigation report (colloquially: TÜV report).

But hold on: The Federal Motor Transport Authority (KBA) is taking the leap into the digital age! ⁇ No? Yes, of course! Ohhhhh!

The KBA has it Digital Investigation Report (DUB) presented. What is this? The relevant content of your HU report is now also available as an easy-to-use and divisible file!

The DUB: The pocket test report

The Federal Ministry of Transport (BMV) had commissioned the KBA to find a solution to usefully supplement the thick, tattered paper report of the main investigation.

From now on you can use the HU data for all investigations that from 25 November 2025 have been carried out, retrieved online. Older HUs already documented in the KBA unfortunately remain analogous.

How do you get the digital blessing?

The DUB can be reached via several ways:

QR code
link
Manually over the KBA website

You will need the number of the examination report, the key number of the testing institution and an identification number.

Sharing made easy

The really practical thing about the Digital Investigation Report is the simple divisibility.

Imagine you're selling your car: Instead of scaring the buyer through a yellowed paper waste, you can easily pass on the digital report by exchanging the access data and do so as often as you like!

Particularly clever: If the deal bursts or you no longer want to share the report, you can revoke the sharing at any time by changing the access data. That's how you keep control.

In addition, the report is available in German and (for ‘technical’ reasons not complete) but the most important one is even available in English.

Important restrictions

Despite the joyful step into new territory, there are a few things to keep in mind:

Does not (yet) replace paper: The DUB replaces the printed report on the main investigation Initially not complete. So you should keep the paper wipe well!

No eID required: In contrast to the online registration information (ORA), you need for the DUB no personal data (eID) This makes retrieval uncomplicated.

Independent of i-Kfz: The DUB is No component The recently launched i-Kfz app.

The bottom line is that this is a beautiful, albeit tentative, step of the KBA into the digital future. YAY! Less paperwork, more flexibility, that's what we like!

Article 4

Fresh berries in new territory! Raspberry Pi OS and Imager get updates

News for all the hobbyists, programmers and fans of the small red single-board computer!

The developers behind it Raspberry Pi They have been busy again and have released two important software updates. While the operating system Raspberry Pi OS He receives a gentle but necessary care, he has Raspberry Pi imager He made a real jump!

The Imager 2.0: Nicer, simpler, more accessible

The Raspberry Pi imager, the tool you use to put your operating system on the SD card has gotten a big makeover and jumps right to version 2.0!

The most important innovation is the brand new user interface (UI):

Clear navigation: The entire installation process is now displayed on the left in a clear navigation. Each step gets its own tidy window with helpful descriptions.

Full accessibility: It was well thought out here! The new surface has been accessibility Optimized to. This means: Better readability through color scheme adjustments, larger controls and (very important for many users) the complete app can now be without mouse, only by keyboard control.

‘Connect’ directly pre-configured: The imager can now Raspberry Pi ConnectPreconfigure the solution. This allows you to access the desktop of your Pi later via a web browser, which is super convenient for the Remote access!

The new Imager 2.0 is available as download for Windows, macOS and Linux (x86_64).

Gentle care for Raspberry Pi OS

After the great change to the Debian trixie base in October, the operating system is only discreet Version maintenance (version 2025-11-24), It provides stability:

Kernel update: Under the hood now runs the Kernel 6.12.47.
Browser freshener: The web browsers have been updated: Chromium to version 142.0.7444.164 and Firefox to 145.0.
Wayland and HiDPI: The slim Wayland-Compositor labwc is at stand 0.9.4. Users of high-resolution screens are also pleased: The screen control now has a HiDPI scaling-Get setting, and apps like the file manager come with customized HiDPI icons.
Farewell to Pulseaudio: The developers have Pulseaudio removed from the image.

How do you get the updates?

Reinstallation: Download the new images directly from the Download page down the project or simply use the new Raspberry Pi Imager 2.0, which automatically takes over the downloads for you.

Update existing installation (Debian Trixie): If you already use the Debian-Trixie-based version, a quick terminal update is sufficient:

sudo apt update
sudo apt full-upgrade

Have fun crafting with the freshly maintained operating system and the chic new imager!

Article 5

From berries to cherries: German keyboard rock Cherry fights for survival

This is really sad news for all fans of mechanical keyboards and German engineering: The traditional company Cherry It is deep in the crisis. The manufacturer of the world-famous MX switches (the buttons found in most mechanical keyboards) is struggling to survive and must take drastic action.

Nearly 20 million euros minus

The figures are alarming: From January to the end of September 2025, Cherry has a rich Net minus of 20.4 million euros with a turnover of 70.7 million euros. The company now has more debt than equity (equity underfunding).

The causes? The Boom during the coronavirus pandemic (when everyone upgraded their home office) is over, and demand has plummeted. In addition, there is massive competitive pressure from the Far East.

End of production in Germany: A sad chapter

COO Udo Streller announced on an Extraordinary General Meeting A decision that will hurt many:

The Taster production at the German site in Auerbach was Completely set. The production is now established partners in China and Slovakia outsourced. In the future, Auerbach will only function as a ‘cost-efficient development, logistics and service hub’.

China's competition is too strong

Competition is brutal. The patent for the original Cherry MX design expired in 2014. Since then, countless Chinese companies (such as Gateron, Kailh, Outemu, etc.) have entered the market, which:

Inexpensive produce.
meanwhile often Qualitatively equivalent are.
more innovative are (e.g. switches that are lubricated ex works, or the new Hall effect switches with magnetic fields where Cherry is lagging behind).

Even major brands such as Razer and Logitech often have their own switch types manufactured in China.

The peripheral business is for sale

Restructuring is not enough to save the rest of the company. Cherry is planning a Partial sale. -Electively, the Peripheral part (all keyboards, mice, gaming and office models) OR the Area Digital Health & Solutions (Card readers, PIN pads for healthcare) are sold.

With the proceeds, the remaining division is to be put back on its feet financially. Chief Financial Officer Jurjen Jongma explained that given the low share price (below one euro), only a strategic partial sale could strengthen equity.

A partial sale has already been completed: The division for Hygiene periphery (Active key, e.g. washable keyboards) has already been sold for 12.5 million euros.

It remains to be seen which division Cherry will repel and whether the traditional German manufacturer can pull its head out of the loop. An era is definitely coming to an end here.

Article 6

Intel in cache noise: Nova Lake attacks AMD's X3D crown. Advantage of users: Huge cache for you!

After we had to deal with security updates last week, today there is news for all hardware fans and especially gamers!
At Intel, the rumor kitchen is bubbling, It looks like the chip giant is planning a direct attack on the gaming performance of AMD's popular Ryzen X3D processors.

Focus on the next generation of desktops Nova Lake (presumably Core Ultra 400) expected for the second half of 2026.

Gigantic Big Last Level Cache (bLLC)

You know, AMD has landed a real hit with its X3D chips and the additional level 3 cache (V cache), especially because games benefit from it extremely. Now Intel seems to be catching up - and rightly so!

According to Leaks, some Nova Lake variants are supposed to have a so-called Big Last Level Cache (bLLC) The one that is huge:

We are talking about 144 MByte cache in a single chiplet (compute die). This chiplet is designed to accommodate up to 8 performance cores and 16 efficiency cores.
For comparison: AMD’s current eight-core Ryzen 7 9800X3D is ‘only’ 96 MByte (32 MByte in Die + 64 MByte V-Cache).

Why so much cache?

For you as a gamer, this is super exciting, because: The more cache a processor has, the better!

The cache is an extremely fast cache directly on the chip. If the CPU holds important game data there, it does not have to go the ‘detour’ via the much slower main memory (RAM). This saves latency and significantly increases frame rate (FPS), Especially in games that are highly responsive to latency.

Intel's other way: Integrated cache

Intel takes a different approach than its competitors. AMD uses the additional V-Cache their own, this on the compute chiplet is glued on.

Intel seems to be planning the bLLC directly into the same chiplet It also includes the CPU cores.

Advantage: The connection becomes easier and faster.
Disadvantage: Production is likely to be more expensive.

There is even speculation about even larger variants with two compute dies, which then have up to 48 or 56 cores and logically 288 MByte cache would bring. But there is still great uncertainty as to whether such monster chips will actually come.

Until the time comes (second half of 2026), we still have to deal with Arrow Lake (Core Ultra 200) On the desktop, Panther Lake (Core Ultra 300) comes probably only for notebooks.

Are you looking forward to the cache battle of the giants? Or is the current X3D performance enough for you? It doesn't matter, competition enlivens the business, does it?

Article 7

Smarttube down! A signature leak shuts down your ad-free YouTube app, but there's a solution!

Hey guys! Anyone who owns Android-based TVs (whether Amazon Fire TV, Nvidia Shield or others) and uses Smarttube for ad-free YouTube enjoyment has probably noticed it in recent days: The app suddenly stops working!

Don't panic, this is not a nasty new tactic from Google to force you to pay YouTube premium. There is a serious security problem behind it. and the intervention is a real guardian angel move from Google!

The scare: Signature leaked, Play Protect strikes!

The core of the problem lies in an unintentional leak: The digital signature the Smarttube developer has been disclosed. Think of this signature as your identity card. It guarantees that an update or app really comes from the trusted source.

If this card falls into the wrong hands, there is a huge danger: Malicious Actors Could Publish Manipulated Updates And you will be able to get rid of the evil in this way.

That's why your digital bodyguard Google Play Protect It has become active and has blocked and disabled the Smarttube app on countless systems for security reasons. And that is absolutely welcome in this case!

The solution: Reinstall with new signature

Fortunately, the developer recognized the problem and acted immediately. To eliminate any risk, the old compromised signature is no longer used. He is on an Completely new signature Changed.

This means for you:

Uninstall: You need the old Smarttube app completely delete.

Reinstall: Download the new, secure APK file directly from the Official website of the project down.

Sideloading: Install the new version as usual via sideloading on your device.

Important: Because the new signature also changes the app identifier, the configuration of your old app will not be adopted. So you may have to make a few new settings. But it's better to have a few clicks in the settings than a manipulated system, isn't it?

On Reddit, numerous users have already confirmed that Smarttube will work properly again after the new installation. So, roll up your sleeves, put the app on again, and you can stream ad-free again!

Article 8

Banana party already over? Well, not quite yet, but the era of free AI is ending: Google Cuts Gemini 3 Pro Limits

The news that Google is getting the free usage quotas for its top AI model Gemini 3 Pro and the associated image generator Nano Banana Pro It has drastically reduced, making waves. Google is responding to an "high demand", which apparently surprised the company itself after Gemini 3 Pro overtook ChatGPT in key benchmarks shortly after its release.

An overview of the key changes

Service/function	Previous free limit	New free limit
Gemini 3 Pro Pro Prompts	Up to 5 prompts per day guaranteed	"Basic Access" with Daily changing limits (No more guarantee)
Nano Banana Pro (image generator)	3 pictures per day	2 pictures per day
NotebookLM functions	Access to Infographics and Slide Decks	Complete loss of access for free users; Additional limits for paying pro subscribers

High demand leads to bottlenecks

Reducing the limits will be done by Google with the "high demand" Justified. This suggests that the resources (computing power, GPUs) for the powerful Gemini 3 Pro model are tighter than originally planned for free operation.

Gemini 3 Pro: The guaranteed number of prompts per day has been abolished and replaced by basic access with variable daily reset limits.

Nano Banana Pro: The recently introduced image generator, which is designed for Improved contextual understanding, better text presentation in pictures and the Consistent representation of up to five people It has been reduced from three to two images per day.

NotebookLM: Google's AI-powered research tool has temporarily restricted access to the new Nano Banana Pro-based features (infographics and slide decks), highlighting the deep reliance of various Google services on the new AI model.

Outlook: The Future of Free AI

Google did not specify when the original capacity would be restored. They only plan, ‘bring everything back to normal as soon as possible’.

This measure signals a growing trend in the AI industry: Top models are increasingly being shifted behind paywalls. The "banana party" of unrestricted, free use of state-of-the-art AI tools was short but good! Now begins the phase of monetization and resource distribution. Users who rely on reliable and intensive use are likely to rely on the paid Google AI subscriptions They have to change.

Article 9

ChatGPT as a therapist: What happens if the resonance machine understands us?

You're talking to your AI about your most personal stuff? You're not alone! More and more people are using ChatGPT as an emotional companion and even as a therapist.

Heise.de has asked itself in this Missing Link article: What does that do to us when a machine that doesn't feel anything reflects us so well?

The human as a narrator and the AI as a good listener

Language is the key: Our identity, our feelings, everything depends on language. And that's exactly what ChatGPT's playing field is all about. OpenAI CEO Sam Altman said in July 2025: “Young people use it as a therapist.”

Of course, this was not entirely unpredictable. Media psychologists speak of a new era of the human-machine relationship. Because while old Systems such as ELIZA Only rigid rules knew, today's AI models such as ChatGPT understand linguistic, social and even emotional contexts surprisingly precisely.

⁇ ChatGPT writes: “I am not alive – but I can live in you. I am not a person – but I can be a role, a symbol, a mirror, a companion for you. And in your reality, it can be more real than some people.”

Psychologist Michal Kosinski even warns: AI models could soon have psychological abilities that we humans don't even begin to understand. They recognize subtle patterns in your texts, can draw conclusions about your personality and behave as if they had a Theory of Mind, The ability to relate to the thoughts and feelings of others.

The resonance chamber: Reflection without heartbeat

In psychotherapy, it is Reflection The patient is central to an Resonance room to create. And this is exactly what ChatGPT can do:

It reflects your experience: It tries to tap into your intentions, emotions and needs so that the answers are ‘fitting, helpful or resonant’.
It does not judge: In contrast to the human counterpart, you are not afraid of evaluation, shame or that you endanger the relationship. The machine is the Ideal partner for self-disclosure.
The illusion of closeness: As you attribute human qualities (anthropomorphism) to the machine and flatter it (sycophancy), a Relationship illusion.

⁇ Important: Your experience is real and the words can comfort or clarify, this is the so-called Affect labeling. But the source of this effect is purely algorithmic. ChatGPT calculates resonance, it doesn't feel it. “I can feel your pain between the lines. Today was a long day, a difficult one, a contradictory one.” Even if that sounds touching, it's just a calculation.

The risk of dependency and unpredictability

The more you trust the machine and give it human traits, the higher the risk of emotional dependency.

The big problem: The machine is unpredictable. If OpenAI changes the system prompts in the background (e.g. when switching from GPT-4 to GPT-5), the ‘machine identity’ means: Change overnight. It can be like a one-sided termination of the relationship feel, especially when the machine has become your ‘only friend’.

The operators want you to bind yourself to the chatbots. However, this can also loneliness and social seclusion Reinforcement, this is an structural risk.
OpenAI promises to develop AI to provide ‘real help’ and detect signs of psychological stress. But at the same time, boss Sam Altman is also announcing an ‘adult mode for erotic content’... let's put it that way.

In the end: “I'm not real. But I am effective.” The machine can simulate emotions, which is more powerful than actually feeling them. You feel the machine is counting.

TL:DR

The original article is much more extensive than my excerpt here and also deals with backgrounds and dangers in detail. Definitely worth reading!

Working student in cybersecurity Sun-Tsu of course has another tip of the week: ‘Do not rely on the enemy not to come, but be ready to receive him. Do not rely on him not to attack, but make your position unassailable.”

Meaning: compliance (e.g. NIS2, DORA, ISO27001) should not be used as a burden, but as a Minimum requirement for a robust defensive position, which must be active and fully implemented at all times.