← Feed

Posts

• Post #4498478

  Oh noez... Microslop wants to bring more Copilot into Windows again. https://www.windowslatest.com/2026/08/05/after-25-years-at-microsoft-i-can-tell-you-why-windows-was-neglected/ What is interesting here is, that the argument is that Windows doesn't produce any revenue. I thought that there is a store. Not that I ever used it... but I hear it exists. #microslop #copilot #devestment #redmond

• Post #4493770

  Publishing a little whitepaper + paper on my research on bringing game theory into algorithmic portfolios: https://hanse-hedge.fund/survival-first-benchmarking-trading-algorithms-under-a-prop-trading-survival-mandate/

• Post #4248248

  Extending a Windows disk is possible online. Easily with diskpart. Ok, you may have to delete the recovery disk. But it's doable. No reboot. On Linux I rarely succeed with this.

• Post #4085535

  It's possible to load OpenAI models (latest generation) into Claude Code. https://github.com/kaitranntt/ccs Easily. OpenAI gives you a usage reset, a 100 bucks account, and Codex Desktop is better structured now. They have Excel and PowerPoint Add-Ins. Just not Word for now. Models (Fable vs. GPT 5.6 Sol) are pretty decent on scale. Claude Code has better capabilities for Jupyter Notebook or Workflows than Codex. You can get the best out of both worlds.

• Post #4069720

  https://youtu.be/3JjkvLAvS9w?si=Nv7YaJGiil10vfSP The problem isn't the negative cash-flow. The problem is the ROI proposition. Alphabet, Microsoft, Meta, and Amazon are not spending hundreds of billions because they have a clear roadmap to absorb all white-collar labor; they are spending because not building the infrastructure risks total irrelevance if a competitor solves key reasoning capabilities first. a) Won't happen with current models. b) Oracle will fold first, and then bell...

• Post #4069527

  Per accident today I found out about WOPI (Web Application Open Platform Interface). And why online document collaboration (Collabora / LibreOffice or OnlyOffice / also MS Office Web) is such a heavy piece of software. Even for today's standards. It's all about locking document and ensuring statefulness. For a sheet or for a document. So I setup my old Filerun server (Filerun is a nifty NextCloud alternative for small-scale adoption) and added my OnlyOffice (Docker) instance. WOPI did...

• Post #4067626

  At this point already, models like Fable earn more per hour than a developer. Let that sink in. #ai #costexplosion #idiocracy

• Post #4043791

  Puh finished my research paper (Quant finance). Brain drained. Asked Fable to write the conclusion paragraph. It was all wrong and irritating. AI is a wonderful tool, for repetitive tasks. But its understanding of new things is zero. For Quants this means: nice, but no.

• Post #4028000

  HuggingFace has some issues with sandboxing an OpenAI model. Model was able to execute some code, that got it extra rights and access. Everywhere. Everywhere? Turns out not everywhere: "while our investigation is not officially over yet, we were able to assert with certainty that no data from Because Security were compromised in this attack, and as a result, no safeguard action is needed on your end." That is intelligence. Know your limits :) #openai #huggingface #escapeartist #bu...

• Post #3963369

  TMV aka TeamViewer was breached. In Germany, not only do you need to disclose to Data Privacy Adminstrations (DPAs). Either federal or locally, depending on size and proportion. You also must inform the BaFin, the SEC equivalent. The BaFin is aware of cyber incidents and will apply rigorous assessments under MAR if you fail. TeamViewer is not a bank, but gets held to a similar standard (here). As a takeaway: even larger German companies remain unaware of their duties to report cyberattacks. No...

• Post #3961000

  On the one hand I am happy to see open-weight model demand from developers. On the other hand I know that China will not give away free AI, and that this move is strategic. The Chinese have read the Peter Thiel playbook and know how to seize growth opportunities in the tech sector. True open-weight models with an open training pipeline and a future in Europe will have to be from Europe. There is no way out. #ai #china #strategy

• Post #3838593

  The potential for mass surveillance grows with new AI capabilities. They add nee Ai cameras now, at the subway stations here in Hamburg. It’s only a matter of time until the data gets abused.

• Post #3690656

  German company bankrupt after cyber attack. https://www.zego-tvz.com/wichtige-information/ 37 years in business. Takeaway: * German SMBs are not aware or capable to handle their IT * underinvestment is a red flag, profitability isn't all you need * recovery is hard There are more and more stories like this. But yes, it's always a mixed bag. There could also be financial problems, that presisted beforehand. #infosec #germany #ransomware #bancrupty

• Post #3688728

  I got my personal account verified via Anthropic CVP. That removes some security guardrails. That being said, sadly this is account wide. If I were to do this for the company account, all staff would be able to perform pentests etc. Imho CVP should be per account. #anthropic #fable #mythos #ai #infosec

• Post #3671792

  The vendor of my office chair was in Ukraine. They literally no longer exist. :( Maybe I can 3d print a spare part

• Post #3665824

  In Germany ai see no shortage of Cybersecurity / Information Security professionals. We were able to hire within 6 week (Hamburg area). Professional networks are vibrant, conferences are full. The usual Ghost job postings remain, but as far as I see it that’s irrelevant. It would be best to correct the narrative out there sooner than later.

• Post #3656762

  Reuters is wrong about a Chinese export control of general purpose AI (GP AI). They are right about the concern, that the EU relies too much on technology imports considering future "frontier AI" models (domain-specific AI included). https://ipc.court.gov.cn/zh-cn/news/view-5766.html A proposed tiered export-control list for open-source AI: filing-only for generic tech, substantive export control for strategic/frontier tech, no open-sourcing for national-security-critical tech 数字法治｜...

• Post #3646506

  Passed my CISSP-ISSAP (preliminary, process involves experience check by ISC2 iirc). This is a CISSP + architecture + governance + management topics. Decent. Would have wished for SANS to have something like this, but it is what it is. I didn't take any training. Imho IT trainings need an AI reform before they are worth their money again. I also hear many private schools invest heavily in AI for teaching purposes. * NotebookLM AI quizzes work well * official books * Boson ExSim for CI...

• Post #3643109

  Last week I had an MRI (x-ray with magnets). I got the images and all data. I have an interesting body. That being said: use the standing desk. AI means less typing. This can save the ( German ) health system millions. So many young men have back problems. It’s a massive problem. #ai #ergonomics #awareness

• Post #3540739

  Makes my day. #crypto #based #finance

• Post #3496181

  https://www.supremecourt.gov/opinions/25pdf/25-332_qn12.pdf The US Supreme Court revoked the independence of the FTC. Media: Panic, we must exit US cloud. Blabla... journalists with MacBooks drinking Coca Cola, and listening to YouTube. Reality: Most contracts, like with AWS EMEA etc., have Standard Contract Claudes (SCCs) anyway. Every DPO can make a Transfer Impact Assessment (TIA). Yes, in some cases you should not use US services. But that isn't a new problem. Keep cool. There are...

• Post #3485845

  Do not invest in AI ETFs. 1. AI isn't exclusive to them. Any high school kid can get AI tools. 2. The edge represents the most likely output of a language model fed with literature An AI model predicts text from the corpus. It's not innovating. It's predicting. The most likely answer, for everyone. 3. LLMs cannot do math They are text crunchers. If that's enough for an ETF, it's slop. 4. LLMs do not have any fundamental capability to assess 10-Qs, 10-Ks, Earnings call...

• Post #3485646

  It's logical that SpaceX moves into MSCI world. It's also logical to balance the portfolio due to this. US tech isn't a safe harbour. #msciworld #spacex #riskmanagement

• Post #3485605

  The playbook isn't new. Open-source software was fought by Microsoft with the same arguments for ages. "Open-source AI is dangerous" Good playbook by Amodei. 1. take a fake hit (Mythos), instrument the US regulator to label AI as a threat 2. recover with diplomacy 3. use the same force and channel it against Chinese AI. Because, let's be honest, there will never be a EU AI. EU cannot concentrate capital or innovate. EU is too busy. https://xcancel.com/shaunralston/status/2...