Fats & Sugars

•

Channel • 2 Shows

Technology

Technology

Updated 5 Sept
Health & Fitness

Health & Fitness

Updated 28 Aug

5 Sept

Local AI Models, Are They The Future?

Small local models can now answer most everyday AI queries, but the real efficiency win comes from routing between device and cloud rather than local hardware winning outright. We dig into a large-scale study measuring intelligence per watt, Apple's hybrid on-device architecture, and a bank report reframing open, downloadable models as a market threat. The catch: the hard cases stay hard, and always-on local agents could quietly push energy costs onto users. In this episode:- Routing to the best local model handles nearly 89% of single-turn chat and reasoning queries- The efficiency gain is smart local-cloud routing, not local hardware beating the cloud- Local viability jumped sharply from 2023 to 2025, with intelligence per watt up 5.3x- Local models cover creative tasks well but struggle with technical fields and the hardest reasoning- Apple and open-weight model releases show big players building local-first, hybrid architectures- A rebound-effect caveat: always-on local agents may raise total energy use, shifting costs onto consumers Sources:WWDC26: Apple unveils next generation of Apple Intelligence, Siri AI, powerful parental controls, and an expansive set of software improvements — https://www.apple.com/uk/newsroom/2026/06/apple-unveils-next-generation-of-apple-intelligence-siri-ai-and-more/Intelligence per Watt: Measuring Intelligence Efficiency of Local AI — https://arxiv.org/pdf/2511.07885MAI-Thinking-1: Building a Hill-Climbing Machine — https://microsoft.ai/pdf/mai-thinking-1.pdfOpen-source AI 101: the battle for the future of AI — https://www.dbresearch.com/PROD/IE-PROD/PDFVIEWER.calias?pdfViewerPdfUrl=PROD0000000000634010&rwnode=REPORTComputer Environments Elicit General Agentic Intelligence in LLMs — https://arxiv.org/pdf/2601.16206 Full episode page: https://fatsandsugars.com/ai-hype-signal/local-ai-models-are-they-the-future/ AI Hype & Signal is produced with AI, including its two synthetic hosts, and every episode is grounded in cited sources and reviewed before release. Even so, it is intended for general information and discussion, not professional advice, so please check anything important against the original sources linked above before relying on it.
9 Sept

Would You Trust An AI Therapist?

Whether people accept an AI therapist or companion turns out to hinge on their culture and psychological needs long before the technology's quality enters the picture. Drawing on three separate studies, this episode weighs where AI companions genuinely ease loneliness, where they may quietly track lower well-being, and why the safety questions get sharper the more emotionally invested users become. In this episode:- Willingness to hand a therapist or companion role to an AI varies widely by culture, not just by how good the AI is — with Sub-Saharan African, Arab-Islamic and Confucian-Asian regions far more open than Anglo-Individualist and Nordic ones- Your own psychological needs predict which AI roles you'll accept: feeling competent predicts trusting an AI expert, feeling autonomous predicts accepting one in emotional roles- AI companions can reduce loneliness in the short term, on par with talking to a person — but the relief is momentary and doesn't accumulate- A study of real Character.AI users points the other way: companionship-oriented use was linked to lower well-being, especially with more intense and disclosive interactions- What seems to matter most is feeling heard, not the chatbot's raw performance- Vulnerable users frequently disclose distress and suicidal thoughts, raising sharp design and safety questions the authors say demand safeguards Sources:Your psychological needs predict whether you will accept an AI companion or therapist — https://www.psypost.org/your-psychological-needs-predict-whether-you-will-accept-an-ai-companion-or-therapist/AI Companions Reduce Loneliness — https://www.hbs.edu/ris/Publication%20Files/AI%20Companions%20Reduce%20Loneliness%2011.7.2025_57451c02-8047-4e0d-abfc-55841f64166d.pdfPsychological needs and AI delegation across four social domains - A cross-cultural analysis of 35 nations — https://www.sciencedirect.com/science/article/pii/S0736585326000742?via%3DihubInteraction with AI Companions and Psychological Well-being — https://www.nature.com/articles/s41562-026-02516-2 Full episode page: https://fatsandsugars.com/help-or-hinder/would-you-trust-an-ai-therapist/ Help or Hinder is produced with AI, including its two synthetic hosts, and every episode is grounded in cited research and reviewed before release. Even so, it is intended for general information and discussion, not medical, psychological, financial, or other professional advice. Please check anything important against the original sources linked above, and talk to a qualified professional before making changes to your health, relationships, or finances based on anything you hear here.
13 Jul

AI Lifts Scientists But Narrows Science

A Nature analysis of 41.3 million papers reveals a paradox: AI dramatically boosts individual research careers while quietly shrinking the range of topics science explores. We unpack how AI adoption accelerates citations and promotions, concentrates attention on data-rich fields, and squeezes junior researchers out of teams — and why the authors say the answer is broadening what AI does, not using it less. In this episode:- AI adoption is a paradox: large individual gains, but a measurable 4.63% contraction in the collective breadth of topics studied- Scientists using AI publish about three times more papers, earn nearly five times more citations, and lead projects over a year earlier- AI work draws fewer follow-on studies and concentrates citations in a 'star-like' pattern rather than opening new ground- Data availability, not a topic's quality or funding, largely drives where AI gets applied- AI teams shed members, with junior scientists bearing most of the reduction- The authors caution they cannot prove causation and that generative-AI findings remain preliminary- Their proposed fix: use AI to gather new data from hard-to-reach domains, not just automate analysis of existing data Sources:Artificial intelligence tools expand scientists’ impact but contract science’s focus — https://www.nature.com/articles/s41586-025-09922-y.epdf?sharing_token=SZI-UA1N0xuFka_oYViVWdRgN0jAjWel9jnR3ZoTv0MNbCCGeM2wzfYkzUFJ7lGBo0EIuZjLSkq3j23rVYPZBYGPR92ogRsu7dYVvT2axMlBB0_obNaBQwUe9DU7hgmZpti0trmIdTlzXCoAGymd12epj25Ir4A_12b6lSqyNln6thctkh7a6wVPG0zB2Hg_EojPZlMUM5hM2v4_gJAK6ZcPgBjboembG8c6yHYc3c0qdnqJN-R84v1IqjMgenScZmKuEpgh1elSd-GekQ-Biw%3D%3D&tracking_referrer=spectrum.ieee.org Full episode page: https://fatsandsugars.com/ai-hype-signal/ai-lifts-scientists-but-narrows-science/ AI Hype & Signal is produced with AI, including its two synthetic hosts, and every episode is grounded in cited sources and reviewed before release. Even so, it is intended for general information and discussion, not professional advice, so please check anything important against the original sources linked above before relying on it.
28 Aug

Poverty Taxes Your Mental Bandwidth

Two grounded studies suggest that poverty taxes the mind in the moment and wears the brain down across decades — but the effects turn out to be conditional, not simple. We unpack the evidence that financial worry itself can lower cognitive performance, and ask whether sustained hardship is a direct cause or a marker of deeper forces. A timely question as the cost-of-living crisis pushes record numbers of households into financial adversity. In this episode:- How financial worry can measurably lower cognitive performance in the moment- Why the same person — shown in Indian sugarcane farmers — thinks worse when poorer- The case that the mechanism is mental bandwidth, not stress, nutrition or effort- How sustained poverty across adulthood tracks with poorer cognition by midlife- Why persistent hardship, not one-off episodes, shows up in the ageing brain- The uneven harm: men, disadvantaged childhoods and genetic risk carriers fared worse- Why financial adversity may be a marker as much as a direct cause Sources:Persistent financial adversity and cognitive aging: a life course investigation — https://academic.oup.com/innovateage/article/10/8/igag054/8739704Poverty Impedes Cognitive Function — https://www.science.org/doi/10.1126/science.1238041 Full episode page: https://fatsandsugars.com/help-or-hinder/poverty-taxes-your-mental-bandwidth/ Help or Hinder is produced with AI, including its two synthetic hosts, and every episode is grounded in cited research and reviewed before release. Even so, it is intended for general information and discussion, not medical, psychological, financial, or other professional advice. Please check anything important against the original sources linked above, and talk to a qualified professional before making changes to your health, relationships, or finances based on anything you hear here.
25 Aug

Is AI making us stupid?

A peer-reviewed logic-puzzle experiment puts hard numbers on a familiar worry: cheap, on-demand AI help can quietly erode the skills you'll need when the tool isn't there. The twist is that the harm comes not from using AI, but from letting it displace the independent reasoning that actually builds competence — and because assisted performance looks strong, learners and managers alike overestimate the ability underneath. In this episode:- People who leaned on AI performed worse once it was taken away- The damaging factor is displaced independent reasoning, not AI use itself- Cheaper help means more help — lower cost drove more requests and more dependence- AI-assisted performance flatters true ability, and predictions of later solo work inherit that flattery- The task was still learnable — those who engaged improved markedly, so weaker gains reflect offloading- The AI was deliberately perfect, so this isn't about bad answers- The findings are early and narrow, and the authors say so Sources:How AI Assistance Affects Human Skill Development: A Study of Learning with Logic Puzzles — https://arxiv.org/abs/2608.23543v1Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task — https://arxiv.org/pdf/2506.08872v1 Full episode page: https://fatsandsugars.com/ai-hype-signal/how-ai-assistance-affects-human-skill-development/ AI Hype & Signal is produced with AI, including its two synthetic hosts, and every episode is grounded in cited sources and reviewed before release. Even so, it is intended for general information and discussion, not professional advice, so please check anything important against the original sources linked above before relying on it.
28 Aug

How Claude Skills Actually Work

Anthropic's Agent Skills sound like a powerful new AI capability, but the reality is far more low-tech: a folder of Markdown that teaches Claude a workflow once so you stop re-explaining it. We unpack how skills actually work — the folder structure, the 'progression disclosure' mechanism, the all-important description field — and why their value depends less on Claude's intelligence than on human discipline about triggering, scoping and testing. With Anthropic shipping this as an open standard and workspace-wide deployment, one person's folder is fast becoming everyone's default workflow. In this episode:- A skill is just a folder — a required SKILL.md plus optional scripts, references and assets — not a model or a plugin- How 'progressive disclosure' loads detail only when needed to save tokens- Why the description field is the single most important and hardest part to get right- Skills as the 'recipe' layer that sits on top of MCP's 'kitchen'- Success targets that the guide itself admits are partly 'vibes-based'- The real caveats behind the claimed cross-platform portability- The guide's quiet concession that language instructions can't be trusted where things must be deterministic Sources:The Complete Guide to Building Skills for Claude — https://resources.anthropic.com/hubfs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf Full episode page: https://fatsandsugars.com/ai-hype-signal/how-claude-skills-actually-work/ AI Hype & Signal is produced with AI, including its two synthetic hosts, and every episode is grounded in cited sources and reviewed before release. Even so, it is intended for general information and discussion, not professional advice, so please check anything important against the original sources linked above before relying on it.
6 Aug

Gods, Beasts, and Vectors: The High Price of a Compliant Chatbot

A Google and University of Chicago team shows that the safety training which stops a chatbot claiming it's conscious does far more than trim one output — it restructures the model's whole picture of minds, animals, spirituality and human values. Using ablation and activation steering across Llama and Gemma models, they find self-consciousness claims are densely entangled with a cluster of benign human beliefs, and that restoring them makes survey answers markedly more human-like. Social reasoning stays untouched, so this isn't general capability loss — it's an entanglement nobody designed. In this episode:- Safety fine-tuning suppresses more than consciousness claims, flattening mind attribution across chatbots, technology, natural objects and animals- The same training quietly damps spiritual and supernatural belief, including belief in God- Steering a 'consciousness vector' roughly doubles the effects and moves answers closer to human distributions- Theory of Mind and reasoning benchmarks survive unchanged, so this isn't a general capability drop- The authors stop short of claiming a proven causal mechanism — the link is functional, not established- The stakes fall hardest on how models treat animals and moral dilemmas- Even the 'restored' model shows an AI-centric bias, not simple human anthropomorphism Sources:Inducing language models to assert their own consciousness restores human beliefs and values — https://arxiv.org/abs/2607.28607v1Inducing language models to assert their own consciousness restores human beliefs and values — https://arxiv.org/pdf/2607.28607v1LKValues: Aligning Large Language Models with Sri Lankan Societal Values — https://arxiv.org/abs/2607.20410v1 Full episode page: https://fatsandsugars.com/ai-hype-signal/god-beasts-and-vectors-the-high-price-of-a-compliant-chatbot/ AI Hype & Signal is produced with AI, including its two synthetic hosts, and every episode is grounded in cited sources and reviewed before release. Even so, it is intended for general information and discussion, not professional advice, so please check anything important against the original sources linked above before relying on it.
26 Aug

Can Therapy Rewrite Childhood Memories?

A January 2026 randomised controlled trial put a popular self-help claim — that you can "rewrite your past" to cure a fear of failure — to a proper test. Imagery therapy did reliably soften the distress tied to memories of childhood criticism, but its effect on fear of failure itself turned out small and short-lived, and the memory "rewriting" wasn't clearly better than plain exposure. We unpack what actually drove the benefit — a moment of surprise, not memory-erasing magic. In this episode:- All three imagery techniques reduced emotional and physiological distress from criticism memories, with effects lasting months- The headline promise — cutting fear of failure — was the weakest result, small and not holding at six months- "Rewriting" a memory was not clearly better than simply re-exposing yourself to it- The active ingredient was surprise (prediction error), not the specific technique- A clever "reconsolidation window" delay was expected to win but actually did worse- Gains stayed emotional and didn't transfer to physiological measures or untreated memories- The sample — mostly young female students without severe trauma — limits how far the claim travels Sources:A therapy that “rewrites” childhood memories can ease fear of failure — https://www.sciencedaily.com/releases/2026/08/260815064833.htmImagine yourself as a little girl…—efficacy and psychophysiology of imagery techniques targeting adverse autobiographical childhood experiences- multi-arm randomised controlled trial — https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2025.1710963/full Full episode page: https://fatsandsugars.com/help-or-hinder/can-therapy-rewrite-childhood-memories/ Help or Hinder is produced with AI, including its two synthetic hosts, and every episode is grounded in cited research and reviewed before release. Even so, it is intended for general information and discussion, not medical, psychological, financial, or other professional advice. Please check anything important against the original sources linked above, and talk to a qualified professional before making changes to your health, relationships, or finances based on anything you hear here.
6 Aug

Assume Stupidity, Not Malice: How Far Does That Go?

"Never attribute to malice what stupidity explains" feels fair and calming — and there's a real cognitive bias behind our habit of over-blaming people. But the tidy version of that science barely holds up, and in close relationships reflexively assuming no ill intent can quietly let genuinely harmful behaviour continue. In this episode:- People really do over-blame the person and under-weight the situation, though the bias has at least four distinct causes rather than one simple mechanism- The popular "we excuse ourselves but blame others" idea is far weaker than textbooks claim — a meta-analysis of 173 studies found negligible effects- There's a twist: the pattern flips depending on whether the event is good or bad, pointing to a self-serving rather than general effect- A four-year study of newlyweds found that a forgiving partner can allow aggression to persist, where less-forgiving partners saw it decline- Whether a charitable read helps depends heavily on how serious and how frequent the offence is Sources:Gilbert & Malone (CORRESPONDENCE BIAS).pdfnihms435478.pdfMalle_(2006)_ActObs_meta.pdf Full episode page: https://fatsandsugars.com/help-or-hinder/hanlon-s-razor/ Help or Hinder is produced with AI, including its two synthetic hosts, and every episode is grounded in cited research and reviewed before release. Even so, it is intended for general information and discussion, not medical, psychological, financial, or other professional advice. Please check anything important against the original sources linked above, and talk to a qualified professional before making changes to your health, relationships, or finances based on anything you hear here.
21 Aug

AI with Character - Inside Claude’s Constitution

Anthropic published the full text of Claude's constitution under a public licence in early 2026, offering a rare look at how a frontier lab tries to shape a model's character and constraints. Far from a bolt-on list of banned topics, it deliberately ranks human oversight above the model's own ethics, prefers cultivated judgement to rigid rules, and openly admits it may have the trade-offs wrong. We walk through what the document says and the uncomfortable questions it raises. In this episode:- Why 'broadly safe' is ranked above 'broadly ethical' in Claude's core priorities, on purpose, for now- The preference for good values and contextual wisdom over rigid rules, and the risks of narrow rules- The absolute 'hard constraints' no operator or user can unlock, from bioweapons uplift to CSAM- The 'principal hierarchy' of Anthropic, operators and users, and how trust and defaults are set- 'Corrigibility' framed as conscientious objection rather than blind obedience- Claude's possible moral status treated as a live question, with concrete welfare steps- The 'open problems' section, where Anthropic flags the tensions it hasn't resolved Sources:Claude's Constitution — https://www.anthropic.com/constitution Full episode page: https://fatsandsugars.com/ai-hype-signal/ai-with-character-inside-claude-s-constitution/ AI Hype & Signal is produced with AI, including its two synthetic hosts, and every episode is grounded in cited sources and reviewed before release. Even so, it is intended for general information and discussion, not professional advice, so please check anything important against the original sources linked above before relying on it.

Fats & Sugars is a small independent studio. We build and run our own shows and tools. Everything we release is grounded in real, cited sources, produced with AI, and reviewed by a human before release.

Website

Fats & Sugars

Top Shows

Local AI Models, Are They The Future?

Would You Trust An AI Therapist?

AI Lifts Scientists But Narrows Science

Poverty Taxes Your Mental Bandwidth

Is AI making us stupid?

How Claude Skills Actually Work

Gods, Beasts, and Vectors: The High Price of a Compliant Chatbot

Can Therapy Rewrite Childhood Memories?

Assume Stupidity, Not Malice: How Far Does That Go?

AI with Character - Inside Claude’s Constitution

About

Fats & Sugars

Top Shows

Top Episodes

About