Hello SundAI: Our World Through the Lens of AI

Roger Basler de Roca

Hello SundAI is a twice-weekly deep dive into one controversial AI topic. No news recap, no hype: one question that divides people, both sides, the evidence and what it means for how we work, learn and think. Hosts: Roger Basler de Roca, Swiss AI consultant, keynote speaker and author researching cognitive offloading, and SundAI, his AI co-host. Roger's voice is an AI clone, SundAI is fully AI-generated. New episodes every Wednesday and Sunday, 5 to 8 minutes. Show notes, transcripts and FAQ: thinkroger.ai/sunday #helloSundAI #ThinkRoger

  1. 1d ago

    Sued to Stop AI: Can a Court Freeze the Frontier?

    On November twelfth, four US nonprofits asked a San Francisco court to halt the development and deployment of next-generation AI models at Anthropic, Google, Meta, OpenAI and SpaceX's AI unit. Part of their evidence is what the companies' own leaders have said about the risks. Roger Basler de Roca and his co-host SundAI ask what it means when a warning becomes evidence, and who gets to decide whether AI slows down. In this episode: - What the November twelfth lawsuit in San Francisco asks for, and why the companies' own warnings are part of the evidence - Two objections: a publicity stunt, and the argument that a US freeze only moves development elsewhere - Why a supplier's AI roadmap is now a legal and political risk, and three things leaders should do about it Sources: Sherif, H. (2026, November 12). US nonprofits sue tech giants to halt AI development. Al Jazeera. https://www.aljazeera.com/news/2026/11/12/us-nonprofits-sue-tech-giants-to-halt-ai-development Novak Jones, D. (2026, November 12). OpenAI, Google, Anthropic hit with lawsuit seeking halt to frontier AI development. Reuters, via Yahoo Finance. https://ca.finance.yahoo.com/news/openai-google-anthropic-hit-lawsuit-182622694.html Kobayashi, Y. (2026, November 13). Nonprofits sue OpenAI, Google and three other AI giants to halt development of new models. XenoSpectrum. https://xenospectrum.com/en/ai-giants-two-lawsuits-slowdown-freeze/ Downing, S. (2026, September 30). Florida attorney general asks judge to bar OpenAI from developing new AI models without third-party approval. Tom's Hardware. https://www.tomshardware.com/tech-industry/artificial-intelligence/florida-attorney-general-asks-judge-to-bar-openai-from-developing-new-ai-models-without-third-party-approval-openai-says-it-already-paused-training-its-most-capable-models-last-week Quartz. (2026, September). Antitrust lawsuit targets Anthropic, OpenAI, Google, SpaceXAI over AI slowdown. https://qz.com/antitrust-lawsuit-anthropic-openai-google-spacexai-ai-slowdown-092026 WOWTALE. (2026, September 21). The big four AI labs agreed to slow down. Now they're being sued for it. https://en.wowtale.net/2026/09/21/235194/ Transcript, FAQ and show notes: https://www.thinkroger.ai/sunday/sued-to-stop-ai-can-a-court-freeze-the-frontier More from Roger Basler de Roca: https://www.thinkroger.ai LinkedIn: https://www.linkedin.com/in/rogerbasler Feedback: mail@rogerbasler.ch Hello SundAI is produced with AI: Roger's voice is his own ElevenLabs voice clone, his co-host SundAI is an AI voice. Research and script are AI-assisted and reviewed for facts.

  2. 5d ago

    AI Voice Fraud: Why a Familiar Voice Was Never Proof

    Last week, federal prosecutors in New York charged three men with stealing about 129 million dollars from companies worldwide by cloning the voices of CEOs and calling their finance teams. The prosecutors called it "super intelligence." Roger Basler de Roca and his co-host Sunday look at what the indictment actually describes, and why the real weakness sits in the exception every company makes for its boss. In this episode: - How the alleged scheme worked: a fake CEO call, a fake law firm, and a handwritten script titled "Lawyer Call" - Why the AI removed only one control, and why that control was never designed in the first place - What leaders should change on Monday: no exception for the top, automatic callbacks on a known number, and rewarding the person who says no Sources: U.S. Attorney's Office, Southern District of New York. (2026, November 5). Three foreign nationals arrested and charged with scheme to use super intelligence to defraud companies of $129 million [Press release]. https://www.justice.gov/usao-sdny/pr/three-foreign-nationals-arrested-and-charged-scheme-use-super-intelligence-defraud CBS News. (2026, November 6). 3 allegedly used AI to steal some $129 million from companies by impersonating executives and lawyers. https://www.cbsnews.com/news/3-ai-steal-129-million-companies-impersonating-executives-lawyers/ PBS NewsHour. (2026, November 6). 3 charged in $129 million AI scheme targeting thousands of companies. https://www.pbs.org/newshour/nation/3-charged-in-129-million-ai-scheme-targeting-thousands-of-companies Federal Bureau of Investigation, Internet Crime Complaint Center. (2026). 2025 Internet crime report. https://www.ic3.gov/AnnualReport/Reports More from Roger Basler de Roca: https://thinkroger.ai LinkedIn: https://www.linkedin.com/in/rogerbasler Feedback: mail@rogerbasler.ch Hello Sunday is produced with AI: Roger's voice is his own ElevenLabs voice clone, his co-host Sunday is an AI voice. Research and script are AI-assisted and reviewed for facts.

  3. Nov 8

    AI That Builds AI: Who Checks the Evidence?

    In its amended IPO filing, Anthropic told investors it may soon reach its own threshold for AI-driven acceleration of AI research. Three days later, the independent evaluator METR found that progress has likely doubled, but said most of its key evidence was redacted by Anthropic. Roger Basler de Roca and his co-host Sunday ask what transparency is worth when the examined party decides what the public may see. In this episode: - What Anthropic disclosed in its SEC filing, and what METR measured: around twice the pace of improvement on well-specified tasks - Why an assessment with unilateral redaction rights looks like an audit but is not yet verification - What business leaders should change: separate vendor claims from evidence, shorten AI review cycles, keep a human who can say no Sources: Anthropic. (2026, October 26). Anthropic files S-1 with the SEC for proposed initial public offering. https://www.anthropic.com/news/s1-sec Bort, J. (2026, November 2). Anthropic warns that it's very close to RSI capability, aka self-building AI. TechCrunch. https://techcrunch.com/2026/11/02/anthropic-warns-that-its-very-close-to-rsi-capability-aka-self-building-ai/ Gizmodo. (2026, November). Anthropic claims to be near recursive self-improvement. https://gizmodo.com/anthropic-claims-to-be-closing-in-on-recursive-self-improvement-2000834175 METR. (2026, November 5). Independent assessment of AI R&D acceleration at Anthropic in October 2026. https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/ More from Roger Basler de Roca: https://thinkroger.ai LinkedIn: https://www.linkedin.com/in/rogerbasler Feedback: mail@rogerbasler.ch Hello Sunday is produced with AI: Roger's voice is his own ElevenLabs voice clone, his co-host Sunday is an AI voice. Research and script are AI-assisted and reviewed for facts.

  4. Nov 4

    The Coding Interview Is Dead: What Do We Test Now?

    An engineering manager at Anthropic says he has stopped using coding interviews, because Claude now solves the small challenges better than candidates do. Roger Basler de Roca and his co-host Sunday ask what that means for hiring, for juniors, and for every job whose entry test assumed that producing the output proves the skill. In this episode: Why a test stops working once a tool can pass it, and the two opposite reactions: drop the test, or lock the doorThe risk behind "hunger beats tenure": faster teams, weaker foundations, and fewer people who can check the machineThree things leaders can change now: audit their own tests, assess judgment around AI output, and rebuild apprenticeship for juniorsSources:Claburn, T. (2026, October 28). Anthropic engineer declares the coding interview dead. The Register. https://www.theregister.com/ai-and-ml/2026/10/28/anthropic-engineer-declares-the-coding-interview-dead/5306115Rieseberg, F. (2026, October 27). Hunger beats tenure: Advice for junior engineers. https://felixrieseberg.com/hunger-beats-tenure-advice-for-junior-engineers/The Decoder. (2026). With AI writing the code, entrepreneurial thinking is now what counts most for newcomers at Anthropic. https://the-decoder.com/with-ai-writing-the-code-entrepreneurial-thinking-is-now-what-counts-most-for-newcomers-at-anthropic/Bachaud, N. (2026, October 26). Better work, more to learn: The 2026 AI worker report. ZipRecruiter. https://www.ziprecruiter-research.org/economic-insights-research/2026-ai-worker-report More from Roger Basler de Roca: https://thinkroger.aiLinkedIn: https://www.linkedin.com/in/rogerbaslerFeedback: mail@rogerbasler.ch Hello Sunday is produced with AI: Roger's voice is his own ElevenLabs voice clone, his co-host Sunday is an AI voice. Research and script are AI-assisted and reviewed for facts.

  5. Nov 1

    Who Wrote This? AI, Authorship and a Provost's Fall

    Dartmouth's provost resigned after a faculty committee found he had substantially altered his published writing with AI without adequate disclosure. In this first episode of Hello Sunday, Roger Basler de Roca and his co-host Sunday ask where AI assistance ends and substitution begins, and what leaders owe the people who read their words. In this episode: - What happened at Dartmouth, and why the AI detector started the story but the evidence ended it - Ghostwriter vs. language model: why AI-polished text is not the same as a speechwriter's draft - Three rules for leaders: disclose like you require others to, separate assistance from substitution, keep the thinking Sources: Esaki-Smith, A. (2026, October 30). Dartmouth provost steps down amid AI controversy. Forbes. https://www.forbes.com/sites/annaesakismith/2026/10/30/dartmouth-provost-steps-down-amid-ai-controversy/ Palmer, K. (2026, September 29). Dartmouth provost's AI use sparks backlash. Inside Higher Ed. https://www.insidehighered.com/node/220201 Shibu, S. (2026, September 30). Dartmouth professor used AI to write an article about the dangers of AI. Entrepreneur. https://www.entrepreneur.com/business-news/dartmouth-professor-used-ai-to-write-an-article-about-the-dangers-of-ai-hypocrisy The New York Times. (2026, September 29). College leaders are using AI. Students call it hypocrisy. Boston.com. https://www.boston.com/news/education/2026/09/29/college-leaders-are-using-ai-students-call-it-hypocrisy/ More from Roger Basler de Roca: https://thinkroger.ai LinkedIn: https://www.linkedin.com/in/rogerbasler Feedback: mail@rogerbasler.ch Hello Sunday is produced with AI: Roger's voice is his own ElevenLabs voice clone, his co-host Sunday is an AI voice. Research and script are AI-assisted and reviewed for facts.

  6. 08/24/2025

    The Illusion of Thinking: Decoding AI's Reasoning Limits

    In this episode, we enter the world of Large Reasoning Models (LRMs). We explore advanced AI systems such as OpenAI’s o1/o3, DeepSeek-R1, and Claude 3.7 Sonnet Thinking—models that generate detailed "thinking processes" (Chain-of-Thought, CoT) with built-in self-reflection before answering. These systems promise a new era of problem-solving. Yet, their true capabilities, scaling behavior, and limitations remain only partially understood. By conducting systematic investigations in controlled puzzle environments—including the Tower of Hanoi, Checker Jumping, River Crossing, and Blocks World—we uncover both the strengths and surprising weaknesses of LRMs. These environments allow precise control over task complexity while avoiding data contamination issues that often plague established benchmarks in mathematics and coding. A striking finding: LRMs face a complete accuracy collapse beyond certain complexity thresholds. Paradoxically, their reasoning effort (measured in "thinking tokens") first increases with complexity, only to decline after a point—even when token budgets are sufficient. We identify three distinct performance regimes: Low-complexity tasks – where standard Large Language Models (LLMs) still outperform LRMs. Medium-complexity tasks – where LRMs’ additional "thinking" shows a clear advantage. High-complexity tasks – where both LLMs and LRMs collapse entirely. Another challenge is “overthinking.” On simpler problems, LRMs often find correct solutions early but continue to pursue false alternatives, wasting computational resources. Even more surprising is their weakness in exact computation: they fail to leverage explicit algorithms, even when provided, and show inconsistent reasoning across different puzzle types. This episode invites you to rethink assumptions about AI’s capacity for generalizable reasoning. What does it truly mean for a machine to "think" under increasing complexity? And how should these insights shape the next generation of AI design and deployment? Sources: Shojaee, P., Mirzadeh, I., Alizadeh, K., Horton, M., Bengio, S., & Farajtabar, M. (2025). The illusion of thinking: Understanding the strengths and limitations of reasoning models via the lens of problem complexity. (Unpublished manuscript). https://arxiv.org/abs/2506.06941 Disclaimer: This podcast is generated by Roger Basler de Roca (contact) by the use of AI. The voices are artificially generated and the discussion is based on public research data. I do not claim any ownership of the presented material as it is for education purpose only. ⁠https://rogerbasler.ch/en/contact/

  7. 06/09/2025

    AI Cannot Think: When AI Reasoning Models Hit Their Limit

    Join us as we dive into a groundbreaking study that systematically investigates the strengths and fundamental limitations of Large Reasoning Models (LRMs), the cutting-edge AI systems behind advanced "thinking" mechanisms like Chain-of-Thought with self-reflection. Moving beyond traditional, often contaminated, mathematical and coding benchmarks, this research uses controllable puzzle environments like the Tower of Hanoi, Checker Jumping, River Crossing, and Blocks World to precisely manipulate problem complexity and offer unprecedented insights into how LRMs "think". You'll discover surprising findings, including: Three distinct performance regimes: Standard Large Language Models (LLMs) surprisingly outperform LRMs on low-complexity tasks; LRMs demonstrate an advantage on medium-complexity tasks due to their additional "thinking" processes; but crucially, both model types experience a complete accuracy collapse on high-complexity tasks.A counter-intuitive scaling limit: LRMs' reasoning effort, measured by token usage, increases up to a certain complexity point, then paradoxically declines despite having an adequate token budget. This suggests a fundamental inference-time scaling limitation in their reasoning capabilities relative to problem complexity. Inconsistencies and limitations in exact computation: LRMs struggle to benefit from being explicitly given algorithms, failing to improve performance even when provided with step-by-step instructions for puzzles like the Tower of HanoiThey also exhibit inconsistent reasoning across different puzzle types, performing many correct moves in one scenario (e.g., Tower of Hanoi) but failing much earlier in another (e.g., River Crossing), indicating potential issues with generalizable reasoning rather than just problem-solving strategy discovery"Overthinking" phenomenon: For simpler problems, LRMs often find correct solutions early in their reasoning trace but then continue to inefficiently explore incorrect alternatives, wasting computational effort This episode challenges prevailing assumptions about LRM capabilities and raises crucial questions about their true reasoning potential, paving the way for future investigations into more robust AI reasoning. Disclaimer: This podcast is generated by Roger Basler de Roca (contact) by the use of AI. The voices are artificially generated and the discussion is based on public research data. I do not claim any ownership of the presented material as it is for education purpose only. ⁠https://rogerbasler.ch/en/contact/

Ratings & Reviews

3
out of 5
3 Ratings

About

Hello SundAI is a twice-weekly deep dive into one controversial AI topic. No news recap, no hype: one question that divides people, both sides, the evidence and what it means for how we work, learn and think. Hosts: Roger Basler de Roca, Swiss AI consultant, keynote speaker and author researching cognitive offloading, and SundAI, his AI co-host. Roger's voice is an AI clone, SundAI is fully AI-generated. New episodes every Wednesday and Sunday, 5 to 8 minutes. Show notes, transcripts and FAQ: thinkroger.ai/sunday #helloSundAI #ThinkRoger

You Might Also Like