Souk's Notes - Unofficial Summaries
Dr. Jon KrohnHeterogenous Integration: What It Is and Why AI Chips Need It (with Kai Beckmann)
- Heterogeneous integration scales performance beyond shrinking transistors. Beckmann describes it as the step between front-end transistor making and back-end packaging: combining several dies into one system, stacking dies into memory stacks or placing memory beside a GPU, as in CoWoS-style structures, to shorten data paths. It requires far smaller structures, better heat removal and power optimisation, and front-end-like precision, which opens room for materials and metrology innovation.
- Merck uses AI to skip unnecessary materials experiments. Beckmann says its "materials intelligence" approach applies AI across millions of candidate options to predict how a material will behave in a customer's setup, rather than improving properties sequentially by experiment.
Polis Pandit$What I Don't Understand About Trump's Tariff Strategy
- This post includes paid content that is not available publicly. See the original source.
Crosstalk SolutionsThis UniFi 10Gbps USB Adapter Made My Network 10X FASTER!
- Ubiquiti's USB-C 10 GbE adapter delivered near line rate in testing. Plugged into a Mac Mini's USB4/Thunderbolt port and cabled to a UCG-Fiber, it measured 9,750 down and 9,776 up (the captions give no unit) against an OpenSpeedTest server on a Minisforum MS-01; MSRP is $199.
Dr. Jon Krohn876: Hugging Face’s smolagents: Agentic AI in Python Made Easy — with Jon Krohn (@JonKrohnLearns)
- smolagents keeps agent logic small enough to read and tweak. The host describes Hugging Face's new Python library as giving any LLM the ability to use tools, call APIs and reason over multiple steps. Its core logic fits in a few thousand lines with minimal abstraction over raw Python, and it succeeds the transformers.agents module, which it will eventually replace.
- Its agents act by writing Python code, run in a sandbox. Rather than expressing actions in JSON or text, smolagents' code agents generate Python that is then executed, which the host says gives them Python's full flexibility, with sandboxed execution as the safeguard. A conventional tool-calling agent is also offered.
- Tools are plain decorated functions, shareable on the Hub, with any model. Any Python function becomes a tool with an @tool decorator, type hints and a docstring, and tools or whole agents can be shared and loaded from the Hugging Face Hub, "like an app store". Models can be open ones via Transformers or the inference API, proprietary ones from OpenAI or Anthropic via LiteLLM, or local ones.
- It lowers the bar for agents in business, data science and research. The host lists research question-answering, as in Hugging Face's demo of a leopard running across a Paris landmark, one-line CLI trip planning, browser automation via its web agent, and on-the-fly pandas analysis. He argues firms can prototype agents in a few lines and keep data in-house with local models, and researchers can benchmark small open models against large proprietary ones. He states the episode is not sponsored by Hugging Face.
- Heavier frameworks beat it for scale, data and multi-agent work. By the host's comparison, LangChain surpasses smolagents in scalability, memory management and integrations, and LlamaIndex in large-data retrieval. Pydantic AI offers type-safe validated outputs, while Microsoft AutoGen and CrewAI are built for coordinated multi-agent systems, where smolagents focuses mainly on single agents.
Scott HanselmanTiny Experiments with Anne-Laure Le Cunff
synopsis
Le Cunff, a neuroscientist, founder of Ness Labs and author of Tiny Experiments, ends on the view that running small, time-boxed experiments on your own life removes the social comparison of a shared ladder. In an experiment, failure just means a hypothesis was wrong. Doing nothing is also a choice: accepting the defaults.
Linear versus experimental. Le Cunff separates looking forward to events from organising a life around a linear goal. She keeps linear goals for narrow outcomes like a mortgage. Everywhere else she would start from a question you are curious enough to test.
The tyranny of purpose. Hanselman raises the pressure his teenage sons feel to have a plan. Le Cunff cites a 700% rise over two decades in books mentioning "finding your purpose". She argues this pins self-worth on one bet, as with founders who struggle to recover when a startup fails, and that people who found their thing describe a messy route. Hanselman blames the scrolling algorithm. She agrees that feeds act as "a giant artificial leaderboard", admits she still scrolls, and recommends making something generative rather than abstaining.
What an experiment looks like. An experiment is an action plus a duration: her 20-minute daily walk for 20 days during an 8 a.m. to 8 p.m. book-launch schedule. At the end she reviews how it felt and keeps going, tweaks it or stops. Her three-month weekly YouTube experiment looked successful from outside, but she disliked making the videos and stopped. Experiments come before habits, and she says the 66-day habit rule oversimplifies the research.
Community and learning in public. Learning in public can mean one other person or a short-lived chat group. It is optional, but she likens the support it brings to a marathon runner's team.
Procrastination. Her head-heart-hand "triple check" treats procrastination as a signal. Applied live to Hanselman's TED outline, stuck at 60%, it points to heart and hand. The fixes she offers are intentional imperfection, a fun collaborator, and two key tips from a past TED speaker.
- Swap linear goals for experiments wherever life allows. Le Cunff, a neuroscientist, founder of Ness Labs and author of Tiny Experiments, says linear goals suit narrow, predefined outcomes such as saving for a mortgage. Elsewhere she would start from a question or hypothesis you are curious enough to test, rather than a fixed outcome you tie your happiness to.
- The obsession with finding one purpose backfires. In her research for the book she found that mentions of "finding your purpose" in books rose 700% over two decades. She argues it puts all of a person's self-worth in one basket, citing founders who suffer chronic anxiety and depression after a startup fails, and that people who have "found their thing" describe a messy process, not a plan. Hanselman blames social media, and she calls it "a giant artificial leaderboard". She doesn't tell people to stop scrolling, since she still does it herself. Instead she suggests replacing some of it with sharing something useful once a week.
- A tiny experiment is an action plus a duration, reviewed honestly. Her own example: during a book launch with recordings from 8 a.m. to 8 p.m., she took a 20-minute walk daily for 20 days, then judged by felt signals such as mood and sleep whether to continue, tweak or stop. A weekly YouTube video for about three months drew good subscribers and comments, but she dreaded making them and quit. She says this kind of trial should come before committing to a habit, and that the popular 66-day habit figure oversimplifies the research.
- Learning in public is a spectrum, starting with one person. Le Cunff says it can be one friend or a temporary WhatsApp group exploring a new tool for three weeks, not a large following. It is optional, but she says a support system, like a marathon runner's team, is how tiny experiments grow into ambitious ones.
- Procrastination is a signal to diagnose, not beat with willpower. Her "triple check" asks whether the block is in the head (should this be done at all?), the heart (is it enjoyable?) or the hand (are the tools and skills there?). Applied to Hanselman's TED talk outline, stuck at 60% for three months, she diagnoses heart and hand. Her remedies are intentional imperfection, working alongside a fun friend, and asking someone who has given a TED talk for the two things that matter.
Stock Tok$Look Where Others Are Not 👀
- This post includes paid content that is not available publicly. See the original source.
Dr. Jon KrohnSemiconductors 101: From Raw Silicon to Chip (with Kai Beckmann)
- A chip takes about 1,400 repeated steps from blank wafer. Beckmann explains that leading chipmakers build transistors and power and data lines on 300 mm silicon wafers, the size of a vinyl record, cycling through lithography, patterning, deposition, cleaning and planarisation until the wafer is cut into centimetre-scale dies.
- Chipmaking draws on about 80% of the usable periodic table. Beyond photoresists for lithography, Beckmann says the specialty materials for etching, deposition and planarisation span about 80% of the non-radioactive periodic table. His examples mostly concern logic chips such as CPUs and GPUs, which he calls the main driver of compute, as distinct from memory.
- His route to leading Merck's electronics arm ran through IT and HR. Beckmann studied computer science with a semiconductor focus and worked on chip design in the late 1980s, then spent years in Merck's corporate IT, ran a country sales organisation, and served as board member for HR during a period of transformation before being asked to head electronic materials.
AI Analyst LabGenAI Roadmap 2026: Essential Skills for ML Engineers & Data | Ep21
synopsis
The guest, a senior machine learning engineer building generative AI into Adobe Acrobat, traces document AI from OCR to LLMs on the Data Neighbor podcast. His argument is that the engineering work has shifted from training models to choosing where an LLM is worth paying for, grounding it with retrieval, and evaluating its output rigorously. He ends with a roadmap for newcomers: build current projects while learning the fundamentals, and focus above all on business impact, in a field he expects to be dominated by agents this year.
Why documents were hard. He explains that a PDF layers text, images, tables and graphics on a page. Digital PDFs are searchable, but scanned ones are images, and phone scans that arrive tilted, inverted or cropped blocked automation in finance, legal and contracts until enterprise OCR arrived. His early work in India used object-detection and segmentation models trained on open-source and internally annotated data to label titles, sections, paragraphs and nested tables. Vision-language models such as GPT-4 can now read a page, but he says their non-determinism rules them out for enterprise OCR. A host's example is a model guessing a name smudged by pencil, which the guest says it might fill in from other pages.
What Acrobat does now. By his account, Acrobat's AI Assistant answers questions across about ten uploaded documents with citations back to the source passage, to show answers are not hallucinated. Liquid Mode, released in 2023, reflows PDFs for each device, which a host likens to responsive web pages. Prompted carefully, for example asking for every key point chapter by chapter, an LLM can summarize a textbook or generate exam questions from it.
Choosing the model. He says LLM prices have fallen roughly 100–200x in two to three years, so even entity extraction now goes to a JSON-returning prompt instead of custom spaCy or NLTK training. Enterprises move slowly, though. At scale, he prefers distilled or fine-tuned BERT-style models and small language models for classification and structure. He routes queries by complexity and reserves large models for generation and summarization, grounded with RAG, which he describes as retrieving relevant chunks and passing them as context.
Production RAG. Previewing his O'Reilly AI Superstream talk on March 4, based on a year of reading nearly every RAG paper, he says the common failure is naive ingestion: one chunker with default settings for every document. The fixes he names are chunking chosen and tuned per document type, better parsing, and parallel asynchronous ingestion so that indexing billions of documents does not add latency.
Evaluation. He calls evaluation blank for many enterprises. Open-source metric frameworks cover part of it, so his team adds proprietary ones. Retrieval is scored on relevancy and ranking, and generation on embedding-based groundedness and faithfulness. Answer quality is scored by a jury of LLM judges from several providers, each given defined anchors on a 1–5 scale. Subject-matter experts write reference answers, and where their scores diverge from the judges', for example a 2 against a 3, prompts are revised. A flywheel computes live metrics and stores metrics, not user data.
Roadmap and outlook. He advises newcomers to build agents and other current projects without waiting to master the internals, balanced with in-depth courses. He names business impact as the underrated skill. He expects new job titles, risk for repetitive work, and growth in AI engineers through the Jevons paradox, with this year centred on agents.
- Scanned PDFs, not text, are document automation's core obstacle. The guest, a senior machine learning engineer working on generative AI in Adobe Acrobat, says tilted, inverted and cropped phone scans blocked enterprise finance, legal and contract automation for years. Layout was handled with OCR and object-detection models trained on open-source and internally annotated data. Vision-language models can now read pages, but he says enterprises cannot trust them for OCR because their output is non-deterministic.
- Don't spend LLM calls where a small model will do. He says LLM costs have fallen roughly 100–200x over two to three years, so tasks such as entity extraction are shifting from trained spaCy or NLTK models to a single prompt returning JSON. At millions of requests, though, he favors distilled or fine-tuned BERT-style models and small language models. He also routes questions to models by complexity, keeping large models for generation and for RAG-grounded answers.
- Production RAG breaks first at naive chunking and ingestion. Previewing his O'Reilly AI Superstream talk on March 4, he says student-style projects chunk every document the same way with default settings. His alternative is to choose and tune chunking per document type, and to parallelize ingestion asynchronously so that indexing billions of documents does not add user latency.
- Evaluate generation with metric suites, judge panels and experts. He combines retrieval metrics such as contextual relevancy and ranking with embedding-based groundedness and faithfulness, then uses LLM-as-judge panels drawn from several providers. Judges score on a defined 1–5 scale, because he says models handle 1–100 scales badly. Where a subject-matter expert scores lower than the judges, he treats the gap as the signal for revising prompts. An automatic flywheel computes metrics on live traffic and stores the metrics rather than user data.
- Newcomers should build projects alongside fundamentals, aimed at business impact. He estimates a few months to a year to master machine learning, deep learning and LLMs from scratch, and says the field moves faster than that. His advice is to build current generative-AI projects while studying in-depth courses, not "master RAG in five hours" videos. He names understanding business impact as the most underrated skill. His forecast is a year of agents, risk for repetitive jobs, and more AI engineers rather than fewer, which he attributes to the Jevons paradox.
Scott HanselmanEPISODE 13 - Scott & Mark Learn To… YAP
- Internal audiences allow a candour external talks can't. At Microsoft's internal TechReady conference, Hanselman gave a talk arguing "we're selling Copilot wrong" and found the openness so freeing that he is now saying more of it publicly. Mark describes giving an Azure strategy talk on the hyperscaler landscape there that he could never give externally because it is business strategy and opinion.
- Mark publicly called for treating C and C++ as deprecated. In a 2022 tweet he argued Rust was ready and the industry should shift to memory safety, after reviewing trends showing 70% of vulnerabilities were memory-safety problems. He says he posted it as a technical opinion without press clearance and got a Teams call the next morning asking whether he really believed it. The stance built the reputation that led to his recent keynote at Rust Nation UK.
- Microsoft tolerates employee posting, but discretion still applies. Hanselman says Microsoft is far more relaxed than Google or Apple, where friends were told to stop blogging or YouTube. Mark says he has told team members that social media is the wrong forum for some points. Hanselman's team rule, "don't dunk on other teams", comes from a tweet about Windows 10 printers that brought three VPs down on him; the bug got fixed, but he says he should have asked internally.
- Being inside lets them get Windows bugs fixed directly. Mark recounts a printer failure that took a driver reinstall to clear and says the Windows printing team took it up; they point to the universal printer driver as the fix for new printers. Hanselman says a regression in how USB device INF files were set broke Zune on Windows 11, which he found while preparing Zunes for Marvel's Guardians of the Galaxy 3, and that the fix has rolled out.
Dr. Jon KrohnCan AI Be As Energy-Efficient as the Human Brain? (with Kai Beckmann)
- Neuromorphic chips may pay off first in inference, within about five years. Beckmann argues neuromorphic architectures could matter even more for cost-efficient, scalable inference on everyday devices than for training, given that data transfer is a key optimisation target. He calls their use in general compute "a five years plus question rather than a 10 years question", but says it probably won't reach human-brain efficiency; the light-bulb-powered brain comparison is the host's.
- Data-centre AI already scales on several fronts at once. Beckmann says chips such as Nvidia's Hopper and Blackwell combine transistor shrinkage via EUV lithography and gate-all-around nanosheet designs with CoWoS-style integration that places memory beside compute. The same latest technologies are reaching smartphones, laptops, cars and production systems.
Stock Tok$📈 Timothy's StockTok Tracker 📉
- This post includes paid content that is not available publicly. See the original source.
Dr. Jon Krohn875: How Semiconductors Are Made (And Fuel the AI Boom) — with Kai Beckmann
synopsis
Kai Beckmann, CEO of the electronics business of Merck KGaA, Darmstadt, argues that the semiconductor industry has entered an "age of materials". AI chips now improve through materials, integration and inspection as much as through smaller transistors, and his company's strategy is to co-optimise across all of them, using AI to build the materials that make AI chips. He ends on a cautious note: AI demand lifts only the leading edge of a cyclical industry, and Europe must speed up approvals to keep innovating.
The company and the scaling map. The electronics arm has 8,000 of Merck's more than 62,000 employees. Prompted by the host's account of Microsoft's Majorana 1 quantum chip, Beckmann lays out three routes to performance. "More Moore" means shrinking transistors. "More than Moore" covers packaging and integration. Post-von Neumann architectures include neuromorphic designs that bring logic and memory together to remove the bottleneck between them, and quantum computing. Superconducting qubits need near-zero-kelvin cryogenics, which he says is not easily scalable, and quantum materials remain a market of a handful of devices, though Merck's Intermolecular facility works on them.
Neuromorphic and edge AI. Beckmann thinks neuromorphic chips may matter more for cost-efficient inference than for training, and puts their use in general compute at "five years plus" rather than 10, though not at the brain-on-a-light-bulb efficiency the host raises. Today's data-centre chips, such as Nvidia's Hopper and Blackwell, already combine EUV-enabled gate-all-around transistors with CoWoS packaging that puts memory beside compute, and the same technologies are reaching phones, laptops and cars.
How chips are made. Chips take about 1,400 steps on a 300 mm wafer, the size of a vinyl record, repeating lithography, deposition, etching, cleaning and planarisation. The specialty materials involved span about 80% of the non-radioactive periodic table.
Integration, metrology and AI for AI. Beckmann argues complex yield problems are solved fastest by integrating chemistry, physics, electrical testing and chemical delivery systems, which can save years. The acquisition of UnitySC adds high-resolution inspection of 2.5D and 3D structures such as through-silicon vias, used in CoWoS, and its data stream feeds AI optimisation. Merck's "materials intelligence" uses AI across millions of candidates to predict performance in customers' setups instead of iterating by experiment. He also sees light-based data transfer, the focus of a new optronics unit, as a speed and energy gain.
Reinventing a 350-year-old firm. Founded in 1668, Merck pioneered liquid crystals from the late 1960s and held 80% of that materials market around 2005 to 2010. As displays plateaued it moved into semiconductors from 2014 through acquisitions including Sigma-Aldrich's high-tech business and Versum. Beckmann warns that assuming yesterday's success will last is the most fatal business mistake.
Cycles and regulation. Growth is driven predominantly by AI data centres, while smartphone and PC replacement lags as buyers keep COVID-era devices for a fourth or fifth year. Because fabs run full or idle, he says, the industry is inherently cyclical, so Merck stays humble at 20%-plus growth and manages costs in declines. On regulation, he wants data contributors protected when their data feeds AI, but says that needing about a year for a permit to make a new material in Darmstadt slows innovation.
Space. Darmstadt hosts the European Space Agency's operations centre. Beckmann sees microgravity pharmaceutical and materials research, materials for safer and cheaper missions, and space-generated data feeding industry platforms as areas for collaboration.
- Chip performance now scales along three dimensions, not just shrinkage. Beckmann, CEO of Merck KGaA's electronics business, distinguishes "more Moore" (smaller transistors via EUV and gate-all-around nanosheets), "more than Moore" (heterogeneous integration such as CoWoS, placing memory stacks beside GPUs as in Hopper and Blackwell) and post-von Neumann architectures. He expects neuromorphic computing in general compute in "five years plus" rather than 10, though not at human-brain efficiency. Superconducting quantum approaches need near-zero-kelvin cooling and remain a tiny market; Merck's Intermolecular facility in San Jose is exploring qubit materials with partners.
- Co-optimising materials, delivery and inspection is meant to compress years-long cycles. Beckmann says chipmakers' yield problems are solved fastest by integrating chemistry, physics, electrical testing and chemical delivery systems. That is the logic for acquiring UnitySC, whose high-resolution visual inspection checks 2.5D and 3D structures such as through-silicon vias. Merck calls its use of AI, including on metrology data, to cut unnecessary materials experiments "AI for AI".
- AI drives leading-edge chips while the rest of the industry stays in a cyclical slump. Beckmann says volume still depends on smartphone and PC replacement, which remains low as people keep COVID-era devices into a fourth or fifth year, and AI data-centre demand "doesn't bail everything else out". Because fabs run either at full capacity or idle, the industry is inherently cyclical; after semiconductor-materials growth above 20%, he says Merck stays humble, knowing shrinking years follow.
- Good AI regulation protects data contributors without slowing permits. Beckmann argues frameworks should ensure those who make data available get returns when it feeds LLMs, since that stability helps investment. But he says that needing about a year to get a permit to make a new material in Darmstadt hurts innovation, and that Europe is not leading on approval speed.
- Every innovation plateaus, so incumbents must switch gears deliberately. Beckmann recounts that Merck's liquid-crystal research from the late 1960s led to an 80% share of liquid-crystal materials at the flat-panel peak around 2005 to 2010. It then moved into semiconductors from 2014 through acquisitions. He calls assuming yesterday's success will last "probably the most fatal mistake you can do in business".
Stock Tok$Why "Long Term" Investing is STUPID Right Now
- This post includes paid content that is not available publicly. See the original source.
Dr. Jon KrohnAI Clones: Unlocking New Opportunities for Experts (with Natalie Monbiot)
- Experts can rent out vetted AI twins of themselves. Monbiot says she now advises a startup (spelled on air W-I-Z-L-Y) that admits experts selectively, lets them train a twin on knowledge they control, use it to generate leads, and monetize it in consulting calls once they trust it to represent them.
- A twin's value is augmented skill, not just bought time. Because the twin is cheaper, always available and speaks any language, Monbiot argues it can serve clients abroad at a price point they can afford, which she calls the "10x" of the idea.
- A CEO's twin can carry company culture across an organization. Monbiot describes building one for a healthcare CEO, first inside Slack, where employees can consult it on his values, have it interject, and get feedback on work, including staff at a company recently acquired; voice and other modalities are to follow. The host flags liability as an open question when people act on the twin's advice.
Mr. GlobalIf Doge Is Finding Fraud Why Hasn’t Anyone Been Arrested For Fraud?
- DOGE's fraud claims don't survive an obvious test: where are the arrests? He argues that if DOGE is genuinely finding billions in daily fraud, the easiest cases — supposed 120-year-olds collecting Social Security — should be trivial to prosecute since the payments trace to an address or bank account, yet no criminal investigation has followed.
- He contrasts DOGE's inaction on fraud with the speed of immigration enforcement. Thousands of immigrants are rounded up and deported daily, he says, while not a single person has been identified or charged over the fraud DOGE claims to uncover.
- He accuses Elon Musk of fabricating specific savings claims and of self-dealing. He cites a case where Musk claimed to have cut a contract that saved billions, when the spending had already occurred under the Trump administration, and says Musk is canceling government contracts held by others while steering new ones to his own companies.
Stock Tok$Stock Market - 🔻Bull Trap or Blast Off? 🚀
- This post includes paid content that is not available publicly. See the original source.
Mr. GlobalThe Good News Lie, The Strategy Republicans Have Been Beating Democrats With Forever
- A routine federal lease sale is being marketed as a major Trump win, and he calls that dishonest. The Department of the Interior's first-quarter 2025 lease sale raised about $39-40 million, a figure he says is "chump change" in oil and gas terms and typical of sales that often run into the billions.
- The sale's timeline shows it couldn't have been triggered by Trump's energy-emergency order. He points to the BLM's public process, including a project start date of September 2024, to argue lease sales are planned roughly a year ahead and long predate the administration's announcement.
- He frames this as evidence of a broader Republican messaging advantage. He argues Republicans habitually claim and amplify wins, deserved or not, while Democrats fail to counter or claim credit, calling this pattern — not policy substance — a key reason Republicans consistently out-message Democrats.
Dr. Jon KrohnAI vs. Human Intelligence: What Makes Us Unique? (with Natalie Monbiot)
- Each step from books to Google to LLMs removes cognitive work. The host argues that looking things up in a dictionary or encyclopedia exposed him to incidental information that built a web of connections and aided memory, while a search page surfaces ads and an LLM output does the comparing for you.
- Humans should double down on embodied experience to stay competitive with AI. Monbiot holds that AI is part of human cognitive evolution the way language is, but not all of us: being embodied and embedded in the world, noticing things and making real-world connections is the human advantage AI lacks. Her prescription is to use AI to solve big problems and free ourselves "to be more human".
Dr. Jon KrohnAI is Making Us Dumber (with Natalie Monbiot)
- Outsourcing whole cognitive processes to LLMs makes us dumber, unlike Google. Monbiot answers commenters who compare the worry to early fears about Google: search returns facts that feed into your own reasoning, whereas an LLM can tempt you to hand over the entire process without reviewing or challenging the result. She expects this to change how children are educated and how adults reskill to avoid that trajectory.
Scott HanselmanThe AI Revolution in Medicine with Dr. Peter Lee, President, Microsoft Research
synopsis
Peter Lee, president of Microsoft Research, ends up arguing that AI is heading toward systems that demonstrably work but that no human can fully comprehend, and that in science the only thing holding back the next leap is training data, which he hopes quantum computers will supply. The episode was made in association with the ACM ByteCast.
Indispensability and the flip. Lee frames the question as whether LLMs become another technology people cannot bear to be without, like the phone, which he says contains six major triumphs of computer science research. Hanselman offers his phone-run artificial pancreas as an example. Lee calls this "the flip": ultrasound took about a decade from its proposal in the late 1950s to standard of care, and refusing it now would look like malpractice. He concedes skills are lost along the way, but finds the intensity of the AI debate, which he likens to the one after the genome was mapped, reassuring. He admits ethics was not a serious thought early in his career.
The cost of being first. Lee puts GPT-2-level capability at $5–10 million of compute, GPT-3 level at about ten times that, and GPT-4 level near $1 billion. Followers reach the same point far more cheaply, which raises the question of whether being first still pays.
Beyond comprehension. Lee predicts that in much less than five years an AI-written proof in a language such as Lean will type-check as valid while defying human understanding, and he calls that his first mark of superintelligence.
Empathy and next-word prediction. Writing his book over Christmas 2022, Lee went through what he calls the stages of AI grief, from skepticism to joy. He cites a UC San Diego and Stanford study in which GPT-4's replies to patient emails matched doctors' replies for accuracy and were rated more empathetic 9 to 1. On whether next-word prediction can be intelligence, his example is filling in "and the murderer is" at the end of a mystery novel. Optimizing that prediction at scale, he argues, can accidentally discover circuitry that does deductive reasoning. He adds that transformers have very imperfect memory, unlike curated tools such as UpToDate.
AI for science and quantum. Lee says many scientific domains have reached GPT-2-level capability, and GPT-3 level is what makes products possible. He wants 2025's first practical quantum machines to run precise simulations that generate labelled data for proteins, materials and weather.
- Technology adoption in medicine "flips" expectations. Lee's example is ultrasound: proposed as a diagnostic tool in the late 1950s, feared, then standard of care about a decade later, to the point that refusing it now would look like malpractice. He concedes skills are lost along the way, but says the depth of debate among leading thinkers, which he compares to the one after the human genome was mapped, makes him optimistic.
- Reaching a capability first costs roughly ten times more per tier. Lee puts a GPT-2-level model at about $5–10 million of compute, GPT-3 level at $50–100 million, and GPT-4 level near $1 billion, in any domain. Once someone shows it can be done, followers get there "smarter and a hell of a lot cheaper", and he asks whether the premium on being first still holds.
- AI will soon build things that work but no human understands. Lee predicts that in "much less than 5 years" an AI will produce a mathematical proof in a proof language such as Lean that type-checks as valid but defies even the smartest humans' comprehension. He says that would be his first mark of AGI or superintelligence.
- Patients rated GPT-4's replies far more empathetic than doctors'. Lee cites a UC San Diego and Stanford paper, published a month after his book, in which GPT-4's answers to patient emails were equally accurate and judged more empathetic by a factor of 9 to 1 in a blind test. His reading is that a frazzled doctor writes two or three sentences while the AI writes paragraphs and recalls personal details, which he calls both interesting and disturbing.
- AI for science is stalled at GPT-2 level for lack of data. Lee says Microsoft Research's models for weather, proteins, molecules and materials are published in journals and released on Azure Foundry simultaneously, but there is "no internet of molecular dynamics simulations". He predicts the first practical scalable quantum machines in 2025, and wants them first used to generate large sets of perfectly labelled training data.
Mr. GlobalIMPORTANT! Save your money for next winters utility bills. Share!
- He predicts utility bills will be worse next winter than the $1,200-$1,400 bills seen this past winter. He cites EIA estimates of domestic natural gas consumption rising nearly 2% to 92 billion cubic feet and LNG exports rising nearly 20% to 14.2 billion cubic feet a day, leaving storage refill season starting roughly 600 billion cubic feet behind last year.
- Next winter's natural gas is already pricing above $5, up from about $4 this past winter. He attributes this to Bank of America Securities forecasting demand growth exceeding 4.1 billion cubic feet a day, plus drilling companies holding off on ramping production amid political uncertainty.
- He expects a hot summer to worsen the situation and dismisses an academic's reassurance. With temperatures already hitting 90°F in Oklahoma in March, he argues a hot summer limits inventory builds, and he pushes back on a USC Marshall energy transition director's confidence, warning prices could approach the $9-10 levels last seen before the fracking boom.
Scott HanselmanForking around (with the Linux Desktop) and Finding Out with Justin Garrison
synopsis
Justin Garrison, co-host of the Fork Around and Find Out podcast, makes the case that the Linux desktop has become dependable enough to recommend to non-technical people. He credits image-based distributions with atomic, reversible updates, and ends up recommending Bazzite or Bluefin, after a Chromebook, as appliances for ordinary users.
Families and drivers. Garrison describes the desktop landscape as mostly Ubuntu/Debian, then Fedora-based and other RPM distributions, then a vocal Arch minority. He says the families differ mainly in their opinions on file layout and configuration. Each distribution ships its own kernel flavour, and licensing shapes what it will ship: Ubuntu offered an "easy button" for proprietary Nvidia drivers, and Bluefin delivers Nvidia drivers baked in where Fedora will not. He credits Valve's Steam Deck and Proton for making gaming on Linux first-class.
Bazzite and Bluefin. On his Steam Deck Garrison runs Bazzite, which he calls a better SteamOS, and he has run Kubernetes on it. His 13-year-old son uses a docked Steam Deck as his only computer for schoolwork. Garrison spent years switching between distributions before settling on Bluefin, which is based on Fedora Silverblue. Its A/B images make an upgrade a reboot, and rpm-ostree lets him pin and roll back OS versions. Command-line tools come from Brew, GUI apps from Flatpak, and there is an escape hatch for layering extra packages. He sees read-only systems as a server security practice spreading to the desktop, a mindset he credits to Chromebooks.
Linux at work. At Disney Animation every artist workstation ran Red Hat Linux and the company ran on Google's web apps, and Garrison says other studios work the same way. He found Windows breaks just as often. The move from X11 to Wayland was rough, he says, because Wayland refused broad compatibility, but its security boundaries are deliberate even when sandboxing stops him dragging a file into Slack. Hanselman, noting that he works at Microsoft, points out that Microsoft recently donated Mono to the Wine project.
Owning the stack. Garrison is de-Googling with Firefox and finds many web apps Chrome-only. He keeps a Plex library of ripped DVDs, local home automation, and solar with batteries, so the house keeps running through outages.
- Image-based, A/B-updated Linux desktops give Chromebook stability without the restrictions. Garrison runs Bluefin, built on Fedora Silverblue, where rpm-ostree works like "git for your file system": a full distribution upgrade is just a reboot, and he can bookmark an OS version and fall back to it. He installs command-line tools with Brew and GUI apps as Flatpaks, and sees the read-only system as a security practice moving from servers onto the desktop.
- Linux already runs the media industries' workstations. Garrison managed Linux at Disney Animation, where all the artist workstations ran Red Hat Linux and the company ran on Google's web apps rather than Microsoft Office, and he says DreamWorks and ILM run Linux too. Having also managed thousands of Windows machines at a university, he says Windows "gets broken just the same".
- Wayland's strictness hurt, but its sandboxing is deliberate. Garrison says the move from X11 was rough because Wayland refused broad compatibility and much software was never updated. In his view the isolation is there for a reason, though it still costs him, for instance when he cannot drag a file from an NFS share into a sandboxed Slack.
- Owning more of your stack is worth the friction. Garrison has switched to Firefox and is "de-Googling", and finds many web apps are Chrome-specific rather than web-standard. He rips DVDs into Plex and runs local home automation that keeps working when the internet is down. Hanselman calls pushing back against a browser monoculture noble rather than self-inflicted.
- Give non-technical users a Chromebook, or an appliance-style distro. Garrison's first answer for family is a Chromebook. For Linux he now recommends Bazzite or Bluefin, because updates arrive automatically on reboot, and he says Linux Mint, which he once promoted, lacks a good update story.
Mr. GlobalWhere are your balls?
- He argues press secretary Caroline Leavitt's misinformation attack on the media is hypocritical. Responding to a clip of Leavitt accusing Jeffrey Goldberg and the media of manufacturing hoaxes, he notes she was given a platform through a role he says her husband's money helped secure.
- He says Democrats let Republican claims go unchallenged. He argues Republicans are handed open microphones to accuse Democrats and the media of misinformation without pushback, citing sex-change-in-schools claims as an example he says go undebunked.
- He faults Democratic leadership, naming Jasmine Crockett, for failing to counter-punch. Despite what he calls a daily supply of contradictory or false Republican statements, he says Democrats lack the "balls," quoting Scarface, to challenge or shame them the way Republicans do to Democrats.
Dr. Jon Krohn874: How AI is Transforming Baseball (with Lessons For All of Us) — with Jon Krohn (@JonKrohnLearns)
- Baseball analytics has gone from Moneyball experiment to AI arms race. The host says every MLB team now employs data scientists, and that Statcast, introduced in 2015 with high-speed cameras and radar in every ballpark, generates seven terabytes of data per game, far more than human analysts can sift.
- Machine-learning scouting finds players that traditional stats undervalue. Models ingest college and minor-league stats, biomechanics and video, and weight metrics such as exit velocity, launch angle and spin rate; the host's example is a prospect with a low batting average who consistently hits the ball hard. Teams also use the analysis to tailor drills, such as for a hitter who struggles with high fastballs.
- In-game decisions now lean on simulated matchups and positioning models. Managers consult win-probability tables and pitcher-versus-batter predictions on tablets in the dugout, and data-driven defensive shifts proved effective enough that MLB introduced rules limiting extreme shifts. The host stresses the tools augment managers rather than replace them.
- AI is reaching fans and officiating as well. Broadcasts show live win-probability graphs and catch probability on replays, and the host reports that during 2025 spring training MLB let players challenge human umpires' ball-strike calls to an automated system, which drew mostly positive reviews.
- Any traditional industry can be transformed by pairing domain expertise with AI. The host's lesson is that if America's oldest major sport can reinvent itself, finance, healthcare, manufacturing and retail can too; early adopters gain an edge until others catch up. He forecasts AI won't replace product managers, physicians or supply-chain planners for the foreseeable future, but will arm them with better information.
Mr. GlobalIs Trump Falsifying GDP Numbers? Please Answer If You Know.
- He questions whether a new "gold-adjusted" GDP metric was added to soften recession forecasts. He says the Atlanta Fed's GDP forecast for Q1 stood at negative 2.5-2.8% weeks earlier and has since improved to negative 1.8% under the original method, around the time the Trump administration said the GDP calculation needed changing.
- A newly added "gold adjusted GDP" line reframes the outlook from contraction to growth. He reports this alternative model, which adjusts for gold imports and exports, shows 0.2% growth versus negative 1.8% under the standard calculation, and he says he has never previously seen such a metric used.
- He is not certain who added the metric or whether it is legitimate, and asks viewers with expertise to weigh in. He frames the open question as whether the administration is adjusting GDP methodology to avoid the appearance of an impending recession, without asserting an answer.
Mr. GlobalI Warned Them In October. Here We Are.
- He says his October 2024 warning that a Trump presidency would cost oil and gas jobs has been borne out. He reports over 30,000 job losses in the industry by late March, attributing this to companies cutting staff even before any drop in oil prices materialized.
- He argues policy uncertainty under Trump, not the Biden administration's policies, is what's paralyzing industry investment. He contrasts this with the Biden era, where, despite industry disagreement with the policy, companies "knew what he was going to do" and could plan billion-dollar investments accordingly.
- He reports rig counts falling and job losses spreading across the energy sector while gasoline prices stay elevated. He cites gas prices around $3.16, higher than at Trump's inauguration, and predicts continued job losses even as energy demand grows faster than in decades.
Mr. GlobalHow We Created The Trade Deficit With Canada.
- He argues the US-Canada trade deficit is largely a byproduct of oil, and of deliberate US policy. He cites 2024 figures of about $350 billion in US goods sold to Canada versus $412 billion in Canadian goods sold to the US, and says the gap is driven mainly by oil purchases, not manufactured goods.
- He says the US intentionally built up Canadian oil imports as a national security move. He reports Canada's share of US oil imports rose from about 15% twenty years ago to today's dominant position, replacing Middle Eastern and OPEC supply, a shift he says the US helped engineer by investing in Canadian pipelines.
- He predicts the deficit with Canada will shrink while deficits with Middle Eastern producers grow again. He frames this as preferable, since he says money sent to Canada supports a security-aligned partner while money that would otherwise go to Middle Eastern countries "funds terrorism."
Technology and FriendsRichard Campbell on the Current State of Energy Production [EPISODE 848]
synopsis
Campbell argues that low-carbon energy is further along than the pessimistic messaging suggests but well behind its goals, that no single source will win, and that nuclear's passive-shutdown designs offer low-carbon baseload the US risks losing the skills to build.
Progress and price. He puts the world about a third of the way to non-carbon energy and the likely warming at two to two and a half degrees, down from the Kyoto-era four. Wind and solar are now the cheapest power by levelized cost, at roughly $25 and $23–24 per megawatt-hour, but they are not baseload.
The grid and storage. Small, uncontrollable sources broke a strictly top-down industry: rooftop solar around Sydney overloads the grid at midday and can be disconnected. Grids cope with 10–20% adoption; beyond that needs storage, which at grid scale he says means pumped hydro.
Regional mix. Every area he has studied gets 1–5% from burning agricultural waste; mountainous, rainy British Columbia has cheap hydro, and flat farmland suits tall wind turbines.
Commitments and financing. Asked by Giard, who works for Microsoft, about its carbon pledges, he warns of greenwashing and says data centres need round-the-clock baseload. Combined-cycle gas, at about $40, is the quick option; higher interest rates hurt plants that must be fully built before earning, unlike solar and wind farms that grow in blocks.
Nuclear. Giard dates the loss of faith to Three Mile Island; Campbell says the weapons-proliferation movement mattered more, since light water reactors bred plutonium. Passive-shutdown designs would have spared Fukushima units 1–3, he says. Vogtle's two reactors in 2024 were the first new US ones in 40 years, with none planned after them.
- Decarbonisation has progressed further than its advocates' gloomy messaging admits. Campbell cites an argument that the world is about a third of the way to non-carbon-emitting energy; the Kyoto-era projection of a four-degree rise is "off the table", though he says the 1.5-degree target will be missed and the outcome is somewhere between two and two and a half degrees. Straight-line projections look worse than real curves, which twist and accelerate.
- Wind and solar are now cheapest, but not baseload power. Measured by levelized cost of energy, he puts wind at about $25–26 per megawatt-hour and solar at $23–24, against roughly $40 for combined-cycle natural gas. Because they generate only when wind or sun allow, rooftop solar in New South Wales now destabilises the grid at midday, and Sydney's power company can disconnect a home's panels; he says the real change of the past 20–30 years is grid command and control moving from top-down to two-way.
- Adoption is hitting a wall without storage, best done by pumped hydro. He says grids cope with 10–20% rooftop adoption but going higher needs more storage and transfer; residential solar is expensive per unit of power, with a 5–20 year payback depending on electricity prices. At grid scale he names pumped hydro as the cost-effective store, citing an Australian university study that found many hills with water at the bottom where a reservoir could be dug at the top.
- Goals are well behind schedule, and costlier money made baseload harder. Data centres draw tens of megawatts around the clock and need baseload; he points to Microsoft's Quincy sites beside two hydroelectric plants. Gas, nuclear and coal plants must be fully built before earning anything, which recent inflation and interest rates punished, whereas a solar farm can connect its first 20 of 1,500 acres, or a wind farm its first 10 or so turbines, and grow from revenue. He warns corporate claims involve measurement games and "greenwashing".
- Nuclear stalled over weapons proliferation more than over accidents. Campbell says light water reactors won because they bred plutonium — about 200 pounds from a 50-ton assembly over two years — and that the US has not reprocessed fuel since the Ford administration stopped it, while France does. Newer passive-shutdown designs cool themselves without pumps; by his account Fukushima units 1–3 failed because they lacked it and their diesel generators were in flooded basements. The two Vogtle reactors in Georgia, operational in 2024, were the first new US reactors in 40 years, but with no further plans he expects the skills to decay; he suggests the US might need about 20 built over 10–15 years.
Dr. Jon KrohnMastering AI Before It Masters Us (with Natalie Monbiot)
- Step-by-step reasoning models could scale to many Nobel-level minds. The host, prompted by OpenAI's o1 release, argues synthetic data could broaden reasoning beyond math, chemistry and code, and forecasts that within "a matter of years" a data center could hold a million Nobel-prize-level machines, the "powerful AI" of Anthropic CEO Dario Amodei's essay.
- Relying on LLMs for cognitive work makes people dumber. Monbiot cites studies showing users' cognitive capabilities decline when LLMs do their thinking, and warns of a "great intelligence divide" in which a few build ever more powerful AI while everyday users degrade. Her focus is on how AI can augment human agency rather than on alignment.
- AI's real payoff is freed capacity for ingenuity, not a modest speed-up. She calls the roughly 15% average performance gain for white-collar workers found by Ethan Mollick and others uncompelling, and prefers Andrew Ng's framing of the 10x engineer as exponentially more effective because AI frees time and mental capacity.
- Used well, AI expects more of people, not less. Monbiot cites the former CEO of Drift building an AI-agent-first customer-success startup whose first hire after engineers was a lawyer who also oversees HR. She argues people must think laterally and drop ego: engineers whose work is democratized, and non-technical people who can no longer claim "I'm not a coder".
- The "anyone can now code" narrative isn't fully true yet. Monbiot, a linguist by training, says she got halfway through a Python course, tried building, and had to go back and learn more. The host argues the best learning pairs a specific goal with just enough fundamentals, with the balance differing by person.
AI Analyst LabHow to Build a Data Analytics App in Minutes Using Replit (Cursor, Bolt, V0, Lovable)| Ep 20
synopsis
Episode 20 of the Data Neighbor podcast has Colin Matthews, a product manager and course instructor who founded Tech for Product, arguing that AI coding tools make functional prototypes cheap enough to test an idea on users before committing to build it. The episode ends on a live Replit demo. A one-line prompt produced a working Streamlit analysis app. A deliberately vague "make me money" prompt produced a roughly 3,000-line data-analysis product. Colin attributes both results to an easy, well-trodden case rather than to the tool being able to do anything.
Why prototype. Colin's thesis is that the usual sequence, Figma mocks and then a long build and then the discovery that nobody wants the product, can be cut short. Feedback on mocks is "meh", particularly for recommendation or ML features users need to try. His own side projects took at least 3 months each, and he now estimates about a week for a testable web app. The deeper benefit, he says, is reaching the user-acquisition problem early instead of solving "all the fun parts first".
Limits and jobs. AI is best at speeding up work you already know how to do, because you can see when it goes wrong. Non-technical builders lack a model of clients, servers, databases and auth, so errors compound, and vibe-coded security or payment integrations can look as though they work when they do not. He expects teams to shrink as output rises 2–3x, starting at startups, but not the end of engineers, since system design remains the hard part. A host describes asking Cursor for simple group-bys that would have been quicker to write by hand, and finding it harder to debug code it had written.
Choosing tools. Technical users should use Cursor with a JavaScript stack (React, Node, Postgres, Railway), since models have seen the most JavaScript. Others should use Replit, an all-in-one environment that provisions a database on request and runs Python. Bolt, Lovable and v0 depend on Supabase for back ends, and he finds that integration weaker, so he uses them for client-only prototypes with mocked data and mocked AI responses.
Working method. He keeps a PRD.md spec with components, styling and phased implementation, set as a global instruction in tools that support one, so the agent refers to the spec instead of a confused chat history. He keeps to a "happy path", reverting to the last working state rather than debugging forward. He asks the model to explain a problem without writing code before asking for a fix, because these tools, all on Sonnet 3.7, default to writing code. Being more specific, he says, fixes most failures, much as a vague request from a PM frustrates a data scientist.
The demo. Replit planned, coded, hit an import error, installed the missing libraries and screenshotted its own output to check it, committing to git at each checkpoint. It found the Iris dataset without being given it and built box plots, scatter plots, histograms, correlations, a pair plot, PCA and, on request, decision tree, random forest and gradient boosting views. The "billion-dollar app" prompt led Replit to ask clarifying questions, then build "Money Maker": upload, cleaning, feature engineering, charts, descriptive statistics and an HTML report, run on fake transaction data written by Claude. Its upgrade link and user guide did not work. Colin warns that long autonomous runs rarely succeed, that Next.js would have broken where Streamlit did not, and that Replit applies changes without the accept/reject review Cursor offers. Pricing as he described it is usage-based: $25 a month buys $25 of credit, agent messages cost 25 cents and assistant edits 5 cents.
Keeping up. He calls following the news of the moment mostly wasted effort, citing MCP's sudden popularity months after its release. Most people use none of these tools, so anything that works at all is a good start.
- Prototype with AI tools to test demand before building. Colin Matthews, a PM and founder of Tech for Product, argues that testing users on Figma mocks gives weak feedback, especially for ML features people need to "feel". His fastest side project once took 3 months to ship, and he estimates a web app now takes about a week. The bigger gain, he says, is that you hit the real problem of getting users sooner.
- AI speeds up work you already understand and compounds errors otherwise. Non-technical builders lack a mental model of client, server, database, hosting and authentication, so they cannot spot the AI's mistakes or describe a change correctly. Vibe coders also risk shipping broken auth, security or Stripe integrations that only look like they work. He expects smaller teams with 2–3x output, starting at startups and spreading to enterprises, but not the end of engineers, because system design is the real work and "coding is a medium for your thoughts".
- Choose the tool by your technical skill. If you can already ship, he recommends Cursor with a JavaScript stack (React, Node, Postgres, hosted on Railway), because models perform best on JavaScript. Otherwise he recommends Replit, which handles client, server, database and auth in one place and is the only tool he names that runs Python. Bolt, Lovable and v0 lean on Supabase for the back end, which he finds less consistent, so he uses them for client-only prototypes, with AI responses mocked as JSON. He runs a few simple production apps built this way and moves complex ones out to Cursor.
- Plan in a spec file, revert on errors, diagnose before fixing. A global prompt makes his tools start every project with a
PRD.mdspec listing components, styling and implementation phases, so the agent works from that file rather than a long chat history. When bugs pile up he reverts to the last working commit instead of debugging in chat. He asks the AI to "explain the issue, don't write any code" first, because the tools, all running on Sonnet 3.7 at the time, write code by default. Non-technical users should retest every feature after each change. - The Replit demo built working apps, on an unusually easy case. From a one-line prompt and no uploaded data, Replit built a Python/Streamlit analysis of the Iris dataset (box plots, scatter plots, histograms, correlations, a pair plot, PCA), then added decision tree, random forest and gradient boosting models. A deliberately vague "billion-dollar app" prompt produced "Money Maker", a data-cleaning, charting and report app of about 3,000 lines that worked on fake CSV data, although its upgrade button went to a random domain. Colin warns that long autonomous runs like this are rare, that the run worked because Streamlit is simple (Next.js "would have 100% broken"), and a host notes that thousands of Iris analyses already exist online. Pricing as he gave it: $25 a month for $25 of credit, 25 cents per agent message and 5 cents per assistant message, so he put this demo at roughly $1–2.
Mr. GlobalTrump Misses Great Opportunity
- He argues the Trump administration missed an easy chance to defuse the Signal group-chat controversy by simply admitting fault. He says presidential administrations rarely take responsibility for mistakes, and doing so here would have been the fastest way to end the story politically.
- He says the administration instead chose to fight the story for months by deflecting blame. He reports officials blamed Biden, the reporter, and the publication rather than the people responsible for the chat.
- He frames the refusal to admit error, not the chat's contents, as the larger problem. He compares the response to workplace conduct, arguing that scapegoating in place of accountability would get an ordinary employee fired.
Mr. GlobalStop Rewriting History
- He rebuts a claim that oil rose 20% when Biden signed the Keystone XL executive order. He says oil actually fell the day of and days following the signing, taking a couple of weeks to return to its prior level, contradicting the viral claim.
- He argues the true, simpler point — that gas prices were higher under Biden overall — gets buried under exaggerated claims. He says people undermine a valid argument by reaching for a "home run" instead of the accurate, still-damaging version of the story.
- He frames this as part of a broader pattern of people misrepresenting basic facts online. He offers to help correct the record but says it requires the person to first acknowledge being wrong.
Mr. GlobalWhere We Went Wrong
- He argues the country's core problem is the belief that the opposing party offers nothing worth keeping. He says this belief pushed both parties further apart until the extremes became self-fulfilling, damaging the economy, healthcare, and education along the way.
- He says each administration erasing the prior one's work sets the country back rather than forward. He compares this to why oil companies won't drill in ANWR, since a billion-dollar investment could simply be reversed by the next president, illustrating how policy whiplash discourages long-term commitments.
- He warns the endpoint of this dynamic is one side eventually holding power permanently. He predicts this outcome as the logical result of continued zero-sum politics, while noting listeners' likely disagreement as evidence, in his view, that the pattern is already entrenched.