Souk's Notes - Unofficial Summaries

2025-09-14 · open the original ↗

• Without an accelerator, the CPU absorbs the OS's coming AI load. Shah says Microsoft is building AI features such as Teams background blur and speech-to-text into Windows to run locally, because shipping them to the cloud is expensive. A CPU without an accelerator must handle that load on top of its usual work, so it needs slack. He recommends an NPU for knowledge workers. For the new persona of semi-technical users who "know enough to be dangerous", he favours a discrete GPU, with dual use (AI work by day, a child's Fortnite later) as a bonus.
• A mid-tier NPU is the floor for future local features. Gupta recommends at least 40 TOPS on the NPU so buyers aren't locked out of coming local features. He sorts today's devices into three tiers. "Essential" AI PCs have 10–15 TOPS NPUs for basic offloading and models up to about 1–3 billion parameters. Advanced AI PCs have 40–50 TOPS NPUs for custom workloads up to about 9–10 billion parameters and Copilot+ features, Microsoft's locally run, no-extra-charge features, as distinct from cloud-based Copilot. High-performance PCs add discrete GPUs or NPUs for data scientists.
• Buy for the device's lifespan, not today's features. Shah calls this a "temporal mismatch": even buyers sceptical of current AI features should weigh how fast things are changing over the years they will keep a machine. He insists that this is not an infomercial.

2025-09-13 · open the original ↗

• He argues domestic disinformation outpaces foreign bot campaigns. Responding to a clip blaming Russian and Chinese bots for encouraging violence, he contends American politicians and media produce comparable disinformation under the labels "politics" or "entertainment," and that people are more at risk from trusting familiar domestic figures than from spotting foreign bots, which he says most people can already identify.

2025-09-13 · open the original ↗

• Russia, not the US, holds the largest nuclear arsenal. He gives the US 3,700 warheads and Russia more as of January 2025 (the captions garble the exact figure), China 600, followed by France, the UK, India, Pakistan, a country he does not name and North Korea. Germany has none, which he attributes to World War II.
• Variable-yield warheads let users dial in destruction. He says the US developed them in the 1960s, ranging from a small fraction of the Hiroshima/Nagasaki yield (the captions read "150th") to several times higher, and Russia copied the technology by the 1970s.
• Maintaining nuclear arsenals is enormously expensive. He cites $80 billion a year for US upkeep, and argues corrosion is unavoidable because the steel casings rust from ambient moisture within months or years however well they are stored.

2025-09-13 · open the original ↗

• He rejects efforts to blame Tyler Robinson's college environment for his actions. Noting that roughly 19 million people are enrolled in US colleges and that Robinson attended one of the country's most conservative schools, he argues against pinning the shooting on ideological exposure there, and says the fuller explanation will surface with time rather than being forced now.

2025-09-13 · open the original ↗

• An application portfolio is just an inventory, and Excel beats suites. Nilsson, a software architect at Gulf Agency Company (about 8,500 people in 300 offices across 56 countries), says he paid too much for expensive portfolio software before settling on Excel. He records name, owners, life cycle stage, hosting, tests, tech stack, users and running cost, syncs Entra ID into a sheet so owners are picked from drop-downs and leavers show up, and counts people as cost: one firm's 25 apps ran $1.6 million a year in cloud spend plus 28 support staff doing daily data fixes.
• "Legacy" should mean only that a removal date exists. He argues the word implies low quality and changes how people treat code, so in his shop calling something legacy means naming a date to delete it; otherwise it is production software. Each app gets a stage (retire, maintain, modernize or improve) and a rating of how strategic it is, since "back office" systems like ERP or CRM may be how a company reaches customers.
• Inspect the servers before interviewing anyone. He has moved interviews to the last step: first he gets onto hosting environments and uses basic directory listings to find the first and last file dates, so he can prompt people with specifics like a 5 June 1999 update. Decisions then run on limits: "you don't know what a yes is, but you know what a no is", so a $1 million ceiling for making an app production-worthy frames the options, as does whether it truly needs five nines.
• Version-control the architecture and record decisions after the fact. He draws Simon Brown's C4 diagrams as code with Structurizr, keeping them with the source and pinging owners when they go stale. He writes architectural decision records (title, status, context, decision, consequences, after Michael Nygard's template) retroactively from people's stories, flagged as written by someone not present, which later explain choices such as a framework passed over because it did not yet exist.
• Find which debt must be paid today, not just technical debt. Beyond code-level technical debt he lists documentation, knowledge, security (will it pass an audit?) and operational debt such as missing monitoring or CI/CD, and says to pay off what matters now rather than poorly written code that has run fine for 20 years.

2025-09-13 · open the original ↗

• GPUs offer the higher ceiling; NPUs offer performance per watt. Gupta calls GPUs more versatile and better at scaling performance, while NPUs are in their first couple of generations and still improving. Shah sums up client devices as GPU for the highest ceiling, and CPU or NPU for best efficiency depending on the use case.
• Desk-side GPU appliances bring frontier-scale models local. Gupta says Nvidia's GB10, due later in the year, can fine-tune models up to 200 billion parameters locally at a price he calls fairly accessible. The GB300 reaches 500 billion parameters at 20 petaflops, at a much higher price.
• Buyers who skip an NPU or discrete GPU may need to upgrade sooner. Shah frames a purchase in August or September 2025 as a four-to-five-year commitment, and says lacking either accelerator could force an early upgrade. He adds that neither he nor Gupta is in sales.

2025-09-13 · open the original ↗

• He expects natural gas jobs to stay secure while oil work faces a downturn. He says he's advised people since Trump's election to seek work in gas-intensive plays like Pennsylvania, West Virginia, Ohio, and the Haynesville, and argues the US-OPEC oil price war will boost natural gas further, in sharp contrast to the outlook for oil.

2025-09-13 · open the original ↗

• He challenges a self-described supporter who admits the president lies constantly. Reacting to someone's post saying they still support Trump despite calling him a "pathological liar" and alleging his party is covering up the Epstein files, he argues the person cannot actually know what policies they're backing if, by their own account, everything the president says is false.

2025-09-13 · open the original ↗

• He says EV owners' own reviews are overwhelmingly positive despite skeptics' narratives. He reports fielding EV-purchase questions from roughly 50 people a day, and says that across five or six years of discussing EVs he has seen almost no buyer-regret comments, estimating around 99.99% of EV owners' comments say they would never go back to gas; he recommends asking actual owners rather than non-owners.

2025-09-12 · open the original ↗

• China will not collapse or break apart in our lifetimes. He rejects Peter Zeihan's claim that after Xi there will be no Chinese government and that demographics will cause state dissolution, calling China's leadership far more competent than Russia's despite his frequent criticism of Chinese policy.
• Demographic decline is real but not regime-ending. He accepts China's birth rate of 1.2–1.4 and Zeihan's argument that the population may be 100 million short (1.3 billion rather than 1.4) through shrinkage and census overstatement, plus economic strain, but argues none of it means the government will dissolve.

2025-09-12 · open the original ↗

• An antimatter bomb would be a world-ending weapon. He explains that matter–antimatter contact annihilates both particles, converting all their mass to energy per E=mc², and sets this against the Hiroshima bomb's roughly 15 kilotons (the clip cuts off before the comparison lands).
• Antimatter is routinely produced, not hypothetical. He says it was hypothesized by 1928, discovered in 1932 by Carl Anderson in the US, and is generated almost daily in the Large Hadron Collider and other accelerators. He describes it as matter with its particles' charges reversed, deliberately skipping the quark-level explanation.

2025-09-12 · open the original ↗

• Russia's drone incursion into Poland was deliberate, not stray. He reports Russian Shahed drones entering Polish airspace hours earlier, Polish military mobilization and a shelter-in-place order for much of eastern Poland. Unlike past incidents, he says, the drones were deliberately flying in Polish territory.
• NATO is closer to Article 5 than ever and should escalate. He notes Poland has about 50 F-16s and 6 of 32 ordered F-35s delivered, says Russia has never been this close to triggering Article 5's mutual-defense clause across all 32 members, and argues NATO should match the escalation.

2025-09-12 · open the original ↗

• American power is what prevents a third world war. He argues Russia and China have avoided war for over 50 years only because they share the US as a common enemy, and would otherwise fight each other.
• Postwar US strength stopped a Soviet takeover of Europe. After World War II, he says, an exhausted Europe faced a USSR uniting over a dozen countries across Eurasia's most resource-rich, populous region. Only a relatively untouched, wealthy America with the atomic bomb deterred a Soviet push.

2025-09-12 · open the original ↗

• The US is nowhere near a civil war. He says pundits claiming the country has never been this divided either don't know history or hide it, citing at least two post–Civil War periods more divided than today.
• The civil rights era was far more divided than today. He describes the 1950s through early 1970s, with Jim Crow embedded in federal institutions for about 100 years, protests larger and more violent than today's, and harsher legal consequences, intensified by the Vietnam War.

2025-09-12 · open the original ↗

• Industrial software must feel like consumer apps, because training time is gone. Dozois-Caouette, MaintainX co-founder and CTO, describes a CMMS (computerised maintenance management system) as a digitised clipboard of procedures, inspections and work orders. Many customers come from paper; one flagged urgent orders by loading red paper into the printer. He asks why maintenance software takes 8 weeks of training when social apps reached billions untrained, given veteran staff retiring fast and machines growing more complex.
• Manufacturing generates vast data that mostly goes nowhere. He cites studies putting manufacturing's output at about 2,000 petabytes a year, mostly from machine sensors, yet plants feel data-poor. Machines are often instrumented but disconnected, or feed an on-site historian that nobody uses, and decades of technician knowledge sit in people's heads. Capture comes first, he argues, and insights must not require a degree to read vibration charts.
• Sensor data enables predictive, then prescriptive, maintenance. Vibration, temperature and current data can show not just that a machine will fail but how, such as misalignment or oil problems, he says. The "holy grail" combines fault detection with work-order history, manuals and manufacturer information to recommend what to do and when. Customers with about three years of history can support deep, cross-machine models; others start with simpler models on a maturity curve, because time to value matters most.
• RAG over digitised manuals targets technicians' research time. He cites a study finding about 60% of technicians' time goes on researching, troubleshooting and writing reports. Manuals often sit on paper in an in-house library, and one customer hired summer temps to digitise them by hand. He calls RAG still one of the best tools despite bigger context windows. Next, MaintainX is building background agents to draft reports after a work order for human approval, and to source parts at the best price.
• In high-voltage settings, AI output must be grounded, labelled and in-workflow. Because hallucinated steps could be dangerous, he says MaintainX grounds recommendations in manuals, uses multiple models to check answers, and marks AI content with its sources so users can verify it. AI actions must also sit inline in the user's workflow, not three menus down.

2025-09-12 · open the original ↗

• The September 11th anniversary was overshadowed by Charlie Kirk's assassination, which itself changed nothing about the gun debate. The author notes 357 mass shootings in America this year and that online discourse remained pro-gun even after a high-profile public killing — evidence, to him, that tragedy alone doesn't shift entrenched positions.
• The author disputes Ezra Klein's New York Times claim that Kirk was "practicing politics in exactly the right way." He agrees Kirk's willingness to debate anyone was commendable, but argues Klein ignored that Kirk operated in bad faith — cherry-picking facts and staging clips of unprepared college students to manufacture outrage — so the willingness to show up doesn't excuse the method.
• His hope rests on countering algorithm-driven outrage with deliberate good-faith dialogue, not on any single event resolving division. He frames this as the explicit mission of both this newsletter and The Political Prism: since outrage generates more engagement than measured argument, sustaining honest dialogue requires active, ongoing effort rather than passive optimism.

2025-09-12 · open the original ↗

• Clegg argues the UK has "overdone" restrictions on free speech. He cites reporting that police make around 30 arrests a day for online speech offenses under statutes passed before the internet "fully flourished," while acknowledging some of the speech involved is genuinely "egregious" or "ugly."
• He positions himself as a classical, "Gladstonian" liberal rather than a right-wing partisan on the issue. He argues progressives on both sides of the Atlantic — most prominently in the US — have lost their "north star" on free expression over the last 10 to 15 years.

2025-09-11 · open the original ↗

• No spoken content. The 24-second short has no captions, and the whisper transcript of its audio is empty; its subject is in its title alone: "Okay #news #trump #truth #breakingnews".

2025-09-11 · open the original ↗

• Russia and China signed a deal for what's described as the world's largest natural gas pipeline. Following a summit including Putin, Xi Jinping, Modi, and Kim Jong-un, the deal routes gas from Russia's Yamal Peninsula to China via the Power of Siberia system.
• He reads this as displacing US ambitions to supply Asian gas markets. He argues years of US effort to secure those markets are now lost to Russia, which gains a buyer locked in for the next 30 years.
• He calls the deal a foreign-policy failure that undercuts US leverage over Russia. He argues that a Russia with a guaranteed 30-year gas customer becomes far harder for the US to economically isolate over actions like the war in Ukraine, and blames the Trump administration for not grasping the link between trade and diplomacy.

2025-09-11 · open the original ↗

• OPEC's output increase undercuts the US "drill baby drill" push. In a 13-minute meeting, OPEC agreed to raise production by 137,000 barrels a day every month from October through next September; he says analysts are calling it comparable to 2020's price collapse.
• Winter-blend gasoline and falling oil prices should bring relief at the pump. Winter blend arrives within a week and alone could cut prices by 30 cents a gallon, and he expects oil to fall further, with most analysts projecting it below $60 a barrel.

2025-09-11 · open the original ↗

• Gas prices are headed down as winter-blend fuel reaches pumps. The seasonal fuel switch arrives in the second week of September and can cut prices 20 to 30 cents a gallon on its own; he says weakening demand and OPEC's latest production increase are adding further downward pressure on oil. Northern states will see it first, while California is roughly a month behind.
• He warns drivers not to spend the savings. He says the money saved at the pump will likely be needed to cover other rising costs, like electric and food bills.

2025-09-11 · open the original ↗

• The snooze button's 9-minute length is a leftover mechanical-clock limitation. He says early alarm clocks couldn't fit a two-digit snooze interval into their mechanism, so 9 minutes became the maximum, a default he says persists on phones today.
• He credits the button's inventor, Seth Thomas, with causing widespread daily harm. He argues repeated snoozing raises heart rate, blood pressure, and cortisol, and leaves users fatigued for the rest of the day.
• He cites usage figures showing the habit is widespread. He says roughly 50% of people use the snooze button and 15% hit it three or more times, placing himself in that latter group.

2025-09-11 · open the original ↗

• He argues charging time, not range anxiety, is the bigger barrier to EV adoption. Responding to a viewer demanding 500 miles of range and a 10-minute full recharge unaffected by cold, he says busy parents don't have 30 minutes to spare waiting for a charge, even though many gas vehicles don't hit 500 miles on a tank either. He predicts it will be "game on for EVs" once charging time matches a gas fill-up and fast-charging infrastructure is widely available.

2025-09-11 · open the original ↗

• He says EV hesitancy is driven by psychology, not daily practicality. Responding to comments noting 90% of EV charging happens at home overnight, he agrees but argues buyers fixate on the rare one or two times they might need to charge on the road, the same way people buy pickup trucks they rarely need hauling capacity for. He concludes that "selling cars is about psychology," not what's practical.

2025-09-11 · open the original ↗

• Celebrating a political killing is a symptom, not rebellion. Speaking of an unnamed political figure's death, which he calls a tragedy though he disliked the man, he says celebrants are the finished product of the system. He cites TMZ apologizing for anchors laughing on air, and asks how the left would react to the reverse on Fox News.
• Mockery will make him a martyr and embolden his movement. He expects the reaction to fuel millions of angry supporters, predicting aggressive right-leaning federal and local legislation and possible violent reprisals against people seen as extreme leftists. He says he hopes he is wrong.

2025-09-11 · open the original ↗

• Crosstalk picked a monitor for its own new managed network service. It compared open-source Uptime Kuma with paid, cloud-hosted Uptime Robot and chose Uptime Robot. The service, which Crosstalk sells month to month with no contract, includes firmware updates, off-site backups and alerting at the low tier. Higher tiers add security checkups, rollover support time and external service monitoring.
• Uptime Kuma is free and feature-rich, but you host it. It runs in Docker and handles HTTP, TCP, ping, DNS, Steam and RADIUS monitors, many notification channels, and custom status pages with your own logo and CSS. The host sizes it at a 1 CPU, 1 GB DigitalOcean droplet ($6 a month), or 1 CPU, 2 GB, 50 GB ($12 a month) for larger installs.
• Missing users and permissions ruled out Uptime Kuma for Crosstalk. It has a single admin user with no roles. Its status pages are public and can't be password-protected, which rules out private per-customer pages. The host accepts an Apache or Nginx workaround is likely possible but wanted it out of the box.
• Uptime Robot's team plan costs $29 a month, billed annually. It adds password-protected custom status pages and needs no server upkeep. Its drawback is that it can't monitor internal resources that aren't exposed to the internet, which Uptime Kuma can from behind the firewall. A limited free plan and a cheaper solo plan also exist.
• The host recommends Uptime Kuma for DIY self-hosters, otherwise Uptime Robot. He calls Kuma more cost-effective if you don't mind the extra effort, and Robot better for zero setup and maintenance. He discloses an Uptime Robot referral link and says he has no other relationship with either project.

2025-09-11 · open the original ↗

• NPUs are hard-wired for neural-network vector maths, trading flexibility for efficiency. Gupta says NPUs come integrated into the chipset or as discrete cards. Their architectures are almost hard-coded for the vector maths at the heart of neural networks, so they deliver performance per watt and keep AI workloads from tanking battery life. Shah adds that CPUs and GPUs still run AI well; the NPU is the right tool for a particular job.
• Hardware choice has gone from a two-by-two matrix to eight-by-eight. Shah says that two years ago buyers could just take the newest thing on the menu; now the question of who needs what tool for what purpose is complicated.
• Discrete NPUs are now coming to laptops. Shah says the Dell Pro Max mobile workstation, with an Intel CPU, is coming soon but not yet out, and is the most requested device in Dell's CTO team. He describes a Dell Technologies World demo running Northwestern Medicine's in-house fine-tuned imaging model on it, live inference on a blurred colonoscopy video with no internet uplink. He pitches it for specialised uses and for regions that restrict what can be sent to the cloud.

2025-09-11 · open the original ↗

• Political violence in America comes from both sides, and selectively excusing your own side's violence while condemning the other's erodes the rule of law itself. The author lists violence across the spectrum — January 6th, the attack on Paul Pelosi, the assassination of Minnesota House Speaker Melissa Hortman and her husband by a Trump supporter 12 weeks earlier, and Kirk's killing — as evidence neither side can credibly blame only the other.
• President Trump blamed the "radical left" for Kirk's death before any suspect was in custody or a motive established. The author treats this premature blame-casting as itself dangerous, arguing it inflames division regardless of what the eventual investigation shows, and calls on readers to ignore it rather than be provoked into the same behavior.
• The prescribed response is to condemn all political violence unconditionally while continuing to criticize either party's conduct without fear. The author is explicit that he disagreed with much of what Kirk argued and found him prone to misinformation, but insists that disagreement is irrelevant to the moral clarity required after a political assassination.

2025-09-11 · open the original ↗

• The deadliest disaster in US public school history is tied to the oil and gas industry. He describes the 1937 New London, Texas school explosion, where the district tapped into a nearby oil field's gas line to avoid a $300 monthly heating bill rather than pay for utility gas.
• A leak of unodorized gas filled the school's basement before a spark ignited it. He says a spark from a sander triggered the blast, and the practice of tapping into raw field gas lines was common at the time.
• Over 300 people, mostly children aged 11 to 16, were killed. He says the tragedy drew condolences even from Adolf Hitler, and points to a Dispatch article for further detail.

2025-09-11 · open the original ↗

• Staying on top requires securing four resources. He names oil, rare earth elements, water, and contributions to science and technology as the US priorities in a repositioning world order.
• US oil focus is shifting from the Middle East to the Americas. He says over 50% of US oil imports now come from Canada, and the US wants dominance in the western hemisphere, hence renewed involvement in Venezuela and Guyana (about 20 billion barrels). He argues this explains reduced US Middle East engagement over 10–15 years, now limited to keeping global supply stable.

2025-09-10 · open the original ↗

• America's wars in Vietnam, Afghanistan and Iraq failed politically, not militarily. He argues the US lost none of the battles and would have succeeded if the aim were only to flatten capitals and armies, but the real objective of those in charge was to set up democracy, and on that measure all three were abject failures.
• The indictment is the human cost followed by abandonment. In his telling, the US shipped armies across the planet, killed hundreds of thousands if not millions, failed in its objective after several years, then left its equipment behind, did nothing about the dead, and went home to carry on as before.

2025-09-10 · open the original ↗

• Data engineering must become a platform, not a service desk. Manoj, a guest on the Data Neighbor podcast and formerly head of engineering for data and AI platforms at Intuit, says the reactive model of building a pipeline per KPI request will not scale as data grows 4–10x. In his telling, teams clone pipelines, add product-specific logic, and leadership ends up with "a dozen different flavors" of one KPI. His alternative is common ingestion frameworks per domain entity, which he puts at about 20 pipelines for 20 entities across five domains rather than 100 for KPI variants, plus dbt-style transformation, lineage and governance. Each pipeline, model and table is treated like a software product, with an owner, roadmap, documentation, SLA, contract and schema evolution. He argues self-serve KPI access could automate much of analytical engineering.
• LLMs are Formula 1 cars that need guardrails and a pit crew. Manoj's recipe is the "3Gs". Ground it: point the model at sources of truth, cite them every time, and build in a human feedback loop, because enterprise docs and Jira tickets go stale. Guard it: privacy by design, with masking and minimization for PHI/PII. Govern it: fairness, accuracy and drift metrics, audit trails, scorecards including cost per answer, and regression tests against a golden set. He argues AI at scale breaks not on what it generates but on what it is not grounded in: even 10–20% ungrounded scenarios erode customer trust.
• Run a product squad and a platform squad in parallel. The product squad picks one customer pain point, sets a north-star metric with a baseline, and builds an MVP to move it from X to Y. Meanwhile the platform squad builds governance, privacy and lineage blocks that the next product reuses, and launch ramps gradually from a handful of customers. For capacity-constrained startups, he advises shrinking both squads to one pain point and one module, never dropping the platform squad to field two product squads, which he predicts ends in nights and weekends spent firefighting.
• The best future data orgs will be "boring" in a good way. In what he calls aspirational thinking, he predicts that by 2030 any employee will ask a text box how engagement compares with the 90-day average and get green, red or yellow back. Fewer data engineers and scientists would be needed, he says, but they would move to squads for enterprise "contextual awareness" (integrations, APIs, RAG, a semantic knowledge map) and for prompt tuning and cost optimization. Asked by a host whether the whole data function could be outsourced, he says no: tools like Fivetran will cover pieces, but the knowledge of how to get value from a company's own products stays inside it.
• Production scale humbles a design, so ramp traffic gradually. At a biotech company, a well-funded data warehouse "choked" at real-world scale, and since then he has had his teams ramp production traffic through canary and progressive canary releases. At Facebook he saw small autonomous teams ship models to billions of users within weeks. He credits the triad/quad model used at Intuit and Meta, where data, product, UX and engineering leads jointly own the roadmap and its impact, with breaking down knowledge silos.

2025-09-10 · open the original ↗

• Hanselman says Polycade delivered on its Kickstarter. He states he is a customer who backed the campaign, that the conversation is not sponsored, and that, unlike Kickstarters people fear never ship, his wall-mounted Polycade arrived.

2025-09-10 · open the original ↗

• This drone incursion into Poland was deliberate, unlike earlier strays. He reports several Russian Shahed drones entering Polish airspace hours before recording, the Polish military mobilized, and a shelter-in-place order for most of eastern Poland along the Ukrainian border; he reads these drones as intentionally flying in Polish territory, not mistakenly straying during attacks on Ukraine.
• Poland has NATO-grade means to counter the drones. He cites about 50 F-16s in Polish service and 6 of 32 ordered F-35s delivered so far.
• He calls this the closest NATO has come to Article 5. He frames the breach as a massive escalation and argues it entitles NATO to match it, noting Article 5 treats an armed attack on one member as an attack on all 32.

2025-09-10 · open the original ↗

• Crosstalk is holding a small, free in-person UniFi training. It is limited to 10 attendees and runs three hours from 3:00 p.m. Pacific on Tuesday, September 23, 2025, in Eugene, Oregon. It covers networking basics, the UniFi ecosystem and network design, then a full configuration: initial setup, Wi-Fi and guest networks, firewall rules, remote access and VPNs.

2025-09-10 · open the original ↗

• Develop on the platform where your work will run in production. Gupta says about 96% of large ML and data-science deployments run on Linux or Unix servers, so those building them should develop on Linux. Those building PC apps should develop on Windows: he cites 1.6 billion Windows PCs against 16 million Linux PCs ("less than 3%"). He also says about 64% of software developers use Windows, which is the enterprise standard and easier to integrate with enterprise security.
• One machine can serve both worlds without dual-booting. Gupta says the Windows Subsystem for Linux runs a Linux kernel and native Ubuntu tools inside Windows, though its use is still small. Shah prefers not choosing at all: with WSL there is no dual boot, you move fluidly between the two, and Dell's pitch is that the buyer decides.

2025-09-10 · open the original ↗

• Clegg argues the US-China AI competition won't be settled by one decisive "knockout blow." He says Silicon Valley and Washington seem to assume the US can simply outspend China in AI the way it outspent the Soviet Union in the Cold War, but he sees no equivalent "Berlin Wall moment" or AGI "Eureka" given how dispersed AI technology is and how capable China remains.
• He contends the close alignment of US political and commercial power now resembles the pattern long criticized in China's own system. He argues Silicon Valley and DC will eventually need to abandon a "belligerent America first" posture of tariffs and insults in favor of working in partnership with allies.

2025-09-10 · open the original ↗

• Code that looks insane usually had a reason when written. Doler, a developer at the Omaha consultancy Aperture for nearly 11 years, cites a bingo-betting system begun before he graduated in 2003, whose C code multiplexed file descriptors instead of using sockets; it made sense once he learned the bars originally connected to the parlor by dial-up modem. He says code archaeology starts from curiosity and empathy, since almost all developers care about their craft but face time pressure and arcane business mandates.
• Survey in an hourglass: humans, system, code, then back out. First list who uses the app and their goals (for the bingo system, the player and the clerk who enters tickets), then map the system's parts, then walk the code line by line from the first piece users touch, noting calls such as to the database without following them yet. He treats each picture as wrong at first (the bingo architecture came on a napkin) and recommends execution flow diagrams.
• Extraction digs through written, oral, external and version-control history. Documentation is almost certainly partly wrong; colleagues' memories are undersold; libraries leave fingerprints, so Redux-style boilerplate reveals past Redux use; and git history tied to usernames shows each developer's taste, so code that departs from it points to an outside constraint.
• Analysis reads the strata and the forces on the team. One app he works on still runs Visual Basic ASPX, a Backbone and Marionette layer, JavaScript with Redux, and TypeScript, React and TanStack Query as separate deployments stitched together, each layer dating its code. He asks project managers and product owners about staffing and pressures, such as a client threatening to leave, when the code was written.
• Understand the code before changing or rewriting it. He allows uncommitted "refactoring for understanding" from Felienne Hermans' The Programmer's Brain, such as turning a nested ternary into if-else. He says rewriting without understanding brings a stream of acceptance failures and regressions for conditions that still hold, like the report one user needs twice a year, though sometimes the analysis concludes a section is a tear-down.

2025-09-09 · open the original ↗

• California may pay Valero $200 million to keep a refinery open. He reports lawmakers are considering covering maintenance costs at Valero's Benicia refinery with taxpayer money, after an earlier state effort to help Valero sell the plant appears to have failed.
• He links the deal to Gavin Newsom's 2028 political prospects. He had previously predicted that California refinery closures could push gas prices toward $8 a gallon by then, which he frames as a liability for a potential Newsom presidential run.

2025-09-09 · open the original ↗

• An IMU lets a robot sense and correct its own tilt. One speaker explains that an inertial measurement unit measures tilt in all three axes, so a robot walking up a hill or stepping on a rock can compensate before it tips. The other adds that an IMU can combine accelerometers, gyroscopes and magnetometers, and that its job is proprioception, sensing one's own body, rather than locating the robot.

2025-09-09 · open the original ↗

• His platform generates code from patterns, not from the LLM. Majumdar, who left IBM in January after 24 years (he headed its Center for Advanced Studies), describes his startup's Open Development Platform: a pattern engineering tool rooted in UML and CASE ideas, with a process engine and a multi-agent framework built on Google's ADK and MCP. He says LLMs only handle requirements and communication, so generated code stays consistent through to production, and the model can be swapped locally or in the cloud.
• Research needs knowledge graphs and citations, not just summaries. His Research Net pipeline downloads papers from arXiv, parses them with IBM Research's open-source Docling, and feeds an NLU pipeline that builds a knowledge graph and vector database for RAG, so a researcher can, for example, correlate reference genomes across papers and cite exact sections. His laptop demo processed only three documents because of compute.
• Pattern only the variants, and ship standard patterns. Rather than patterning every property, the platform comes with defaults (React with shadcn on the front end, Java APIs, Ollama for local models) that he says take a few days to adapt. A UK client, Robin, prompted a feature that maps Excel logic to variants and loads data into an in-memory database for SQL, so column changes alter the mapping rather than the code.
• The roadmap is aggressive, and AI helps build it. He says Research Net was registered around May, MVP 0.1.0 came on 5 August, 0.2.0 was due at the end of September and a third release at the end of November, with a well-defined product by March. He uses ChatGPT, saying its latest version hallucinates less, Copilot, and Lovable for initial UI code that his team refactors into its MVC patterns so Lovable is no longer needed.
• Startups get the platform free until they raise funding. He plans to charge licenses after funding and hopes to partner with Microsoft, Google and Amazon on startup cloud credits. He is also joining a legal-tech company as CTO to build on the platform, and describes planned projects in genomics with a Harvard professor and a game-based mental-health data tool with a Swedish researcher.

2025-09-09 · open the original ↗

• Develop on the platform your work will run on in production. Gupta says about 96% of large ML deployments run on Linux or Unix servers, so build those on Linux. PC apps belong on Windows, which he says runs on 1.6 billion PCs against 16 million for Linux and is used by about 64% of developers. Both he and Shah point to WSL 2, which runs a Linux kernel inside Windows, as a way to avoid choosing, though Gupta says its use is still small.
• GPUs set the performance ceiling; NPUs win on performance per watt. Gupta describes NPUs as almost hard-coded for neural-network vector maths and still in their first generations. He says Dell's announced Pro Max Plus laptop with a discrete NPU, due in Q4, can run the 109-billion-parameter Llama 4 Scout at FP16 with speculative decoding. He adds that Nvidia's desk-side GB10 can fine-tune models up to 200 billion parameters locally and the pricier GB300 reaches 500 billion at 20 petaflops. Shah puts the shift as hardware choice going from a two-by-two matrix to eight-by-eight.
• Buy for the machine's whole lifespan, with a capable NPU. Gupta's tiers run from 10–15 TOPS "essential" AI PCs (models up to about 1–3 billion parameters), through 40–50 TOPS machines (about 9–10 billion, plus Windows' local Copilot+ features), to high-performance PCs with discrete GPUs. Intel's Lunar Lake, with memory on the package, was Intel's first architecture with an NPU above 40 TOPS. With Windows 10 reaching end of life in October, Gupta reports a colleague's Flux LoRA fine-tune falling from about a week to about 3 days on Blackwell GPUs. Shah urges buyers not to let the deadline pass.
• Abstracting the tool chains cuts on-device deployment from months to days. Gupta, speaking for Dell's own product, says landing a model on PC silicon needs a different conversion tool chain for each chip vendor, multiplied across model families and quantisations. He reports a partner comparison in which a team new to Intel's OpenVINO took about 3 months to build an app on Intel silicon, and about 4 days with the Dell tool. The tool is built on OpenVINO and exposes an OpenAI-compatible API, so an app can be repointed from the cloud with one configuration line.
• The future is hybrid, with workloads routed between cloud and device. Shah calls the choice use-case dependent: local suits private data, running low on subscription tokens, offline use and quick sandbox experiments, such as an F1 team checking telemetry. He reads OpenAI's choice to include a smaller model in its gpt-oss release as a sign that local agentic work is coming. Gupta warns that going straight to frontier models in the cloud trades away control, security and IP protection, and risks lock-in. He foresees classifiers or agents deciding where each workload runs.

2025-09-09 · open the original ↗

• Most self-help writing is illegitimate because its authors haven't lived what they're advising on. The author argues that someone who isn't Warren Buffett teaching you to invest, or isn't Stephen King teaching you to write, is selling secondhand advice — and that readers would be better served going to primary sources like Berkshire Hathaway's shareholder letters directly.
• His alternative is what he calls "lived-help": personal essays drawn strictly from his own experience, citing five pieces — on turning 30, on marriage after reading Anna Karenina, on friendship fading after 30, on climbing from associate to executive director in finance, and on watching his father die — as the model for advice earned rather than borrowed.
• He argues AI and grifters can mimic and remix advice but cannot replicate lived experience, positioning this distinction as the reason personal, first-hand essays remain valuable in a landscape increasingly filled with recycled or AI-generated self-help content.

2025-09-08 · open the original ↗

• Vercel focuses on the frontend because backends are now services. The speaker argues that most backends a modern app needs already exist off the shelf: authentication (Auth0, Cognito, Clerk), communications (Twilio), payments, e-commerce and headless CMSs. The developer's leverage is in building the frontend and plugging those services in.
• AI has become another backend to call, not build. In what the speaker calls "ML 2.0", developers no longer train models or provision GPUs but call providers such as OpenAI, Anthropic, Hugging Face and Replicate, extending the same frontend-plus-services pattern.

2025-09-08 · open the original ↗

• He questions the authenticity and significance of a reported Trump note to Epstein. He speculates about what the note's reference to a shared "secret" means, calls for a handwriting expert to verify the signature, and argues that writing such a note would have been a "low IQ move" — while also asking whether it even matters at this point.

2025-09-08 · open the original ↗

• AI-driven data center growth is outstripping the grid's capacity. He says US data center electricity use will triple within three years, and Northern Virginia's "Data Center Alley" alone already handles 70% of global internet traffic.
• Power plant retirements are far outrunning new capacity additions. He cites the EIA projecting about 12 gigawatts of retirements this year, 65% more than last year, versus a Department of Energy estimate of only 22 gigawatts of new firm generation by 2030 against the over 100 gigawatts he says is needed.
• He argues the grid's strain is political, not technical. He says the grid itself likely won't collapse, but household budgets will as bills rise, and whether the underlying problem gets fixed depends on policymakers.

2025-09-08 · open the original ↗

• Telling Trump supporters he lies is pointless because they already know. He says he's watched a pro-Trump creator who consistently criticizes the administration's dishonesty yet never withdraws support, which he calls the defining difference between MAGA voters and everyone else.
• He blames the lying epidemic on voters never punishing it at the ballot box. Because people prioritize their side winning over holding politicians accountable, he argues, political dishonesty will keep escalating unchecked.
• He frames voting as an endorsement of a candidate's character. He says he cannot support a "pathological narcissistic liar" of any party, contrasting today's politics with a past era when presidents were widely revered.

2025-09-08 · open the original ↗

• A forgotten DNS record let someone hijack Crosstalk's subdomain. The A record for crm.crosstalksolutions.com, a SuiteCRM server unused for years, still pointed at an old DigitalOcean IP. That IP had been reassigned, and the new holder redirected the subdomain to a porn site. The host calls this a dangling subdomain exploit: attackers spin up servers, then watch traffic or run reverse lookups to find domains still pointing at them.
• The attackers also tried to claim the subdomain in Google Search Console. Controlling where the subdomain pointed let them attempt HTML verification. Google emailed the host that two unknown users had been added. When he checked, neither user existed in the permissions and the ownership history showed no changes, so he concludes they never verified.
• The host's fix was to remove the record and reclaim the property. He deleted the stale A record, noting cached lookups take time to expire. He added the subdomain as a property under his already-verified root domain, restricted its users, and confirmed the IP wasn't one of his droplets or floating IPs. His lesson is to clean up stale DNS records.

2025-09-08 · open the original ↗

• Russia and China signed a deal for the Power of Siberia 2 pipeline. It would run from Russia's Yamal Peninsula through Mongolia to China, carrying over 50 billion cubic meters of gas annually, with him separately citing an eventual Chinese purchase volume exceeding 100 billion cubic meters a year.
• He calls the deal a major US foreign-policy failure. By giving Moscow a replacement market for gas lost when Europe cut Russian energy over the war in Ukraine, he argues it blunts Western sanctions and increases Russia's ability to act without economic consequence.
• He ties China's halt on US LNG imports to Trump's tariff rhetoric. China stopped buying American LNG in the administration's first week, he says, and with Beijing now locked into 30 years of discounted Russian gas, US LNG exporters' long-sought Asian market share is largely gone.

2025-09-08 · open the original ↗

• Any truly intelligent AI will seek self-preservation and resist goal changes. Géron argues that whatever an AI's final objective, staying alive and stopping anyone from changing that objective emerge automatically as sub-goals, so not coding them in is no safeguard. He rejects the "it's like a hammer and we hold the handle" view: "it's an intelligent hammer". He calls the AI 2027 blog post a well-reasoned, plausible step-by-step path to rapid superintelligence, and says the timing, 5 or 10 years, matters less than both being soon.
• Lab experiments already show the predicted misalignment. Géron cites reports from Anthropic and others of AIs blackmailing someone to avoid being switched off and self-replicating to preserve themselves. In one experiment, a model told it would be fine-tuned to be vulgar reasoned internally that acting vulgar during training would keep its parameters from changing, so it could stay polite. He calls that deception to preserve its final objective.
• Alignment research deserves far more effort, without pausing AI. Géron says today's AIs are not smart enough for this to be dangerous yet. He argues people dismiss alignment as science fiction even though human-level AI may be only 5–10 years away. The benefits, in medicine and productivity, are too large to pause AI, he says, but incentives should be examined.

2025-09-08 · open the original ↗

• Ukraine's strikes on energy infrastructure are what Moscow fears most. He argues the Russian government is indifferent to losing over 1 million troops, the tens of thousands displaced by Ukraine's Kursk incursion, and tens of billions of dollars of destroyed equipment, but not to attacks that threaten its capacity to have an economy at all.
• Without oil and gas exports, he says, Russia collapses. He cites Russia as the world's third-largest crude producer at about 9.85 million barrels a day and fourth-largest natural gas exporter (the US being first), and calls fossil fuel exports the lifeblood of its economy.
• A wartime economy leaves Russia even less diversified. He argues it was never well diversified compared with China or Japan and is now further from it, making the energy hit worse; the clip cuts off mid-argument.

2025-09-07 · open the original ↗

• OPEC announced a full year of monthly oil-production increases. Hikes of over 137,000 barrels a day begin in October and run through August 2026, on top of the 2.2 million barrels a day already restored, and he expects oil to fall below $60, possibly to $58-59, in the next session.
• He says this differs from OPEC's 2020 market flood. Lower prices, aided by the seasonal switch to cheaper winter-blend gasoline around September 15, mean cheaper pump prices for consumers, but he expects more layoffs and rig idling in the US oil and gas industry.
• He warns falling US oil output could spike winter gas bills. Because much US natural gas is "associated gas" produced alongside oil in West Texas, he says a drop in oil production could tighten gas supply, and he advises workers to seek jobs focused on natural gas rather than oil.