Souk's Notes - Unofficial Summaries

2025-05-15 · open the original ↗

• An enterprise is differentiated by its proprietary data and its unique skills. Roese argues these two are what AI must be applied to. Generative AI went after the first: RAG chatbots that embed proprietary data, such as service information, in a vector database and make it answerable, which he says almost all of Dell's production projects do, at industry scale for about a year.
• Agents digitise narrow skills, not whole humans. Roese says agents go after the second source, taking an objective and reasoning through it autonomously to deliver an outcome equal to or better than a human's on that skill. Like geofenced self-driving cars or driverless airport trains, today's agents do one job; a well-rounded entity with ethics and morality is AGI/ASI and "a long ways away".
• Agentic AI moves the human from in the loop to on the loop. In Roese's framing, reactive generative AI is a tool with the human doing the work, while with agents the human sets objectives and delegates, and work may stay "below the machine line" indefinitely. He says the architecture differs too: the LLM is the brain, with a knowledge graph holding memory and learned skills, and tool use and function calling giving it a body to act and perceive.

2025-05-15 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-15 · open the original ↗

• Learn adjacent skills inside assigned projects, not on the side. Robert Chang joined Airbnb as a data scientist in early 2016, after Twitter, and worked on supply growth. To build a listing lifetime-value model he bet on Airbnb's then-early ML infrastructure rather than backfilling features by hand, partly to learn how it worked. He says time set aside to study unrelated topics felt productive but was quickly forgotten. A host adds that data scientists should close the loop on their own projects rather than wait on others.
• Infrastructure PM work is mostly invisible reliability work. In 2019, after four or five lunch conversations with the data platform's head of product, Chang became PM of Airbnb's metrics infrastructure team (Minerva), starting with a major incident and a moratorium in week one. What surprised him was how much access control, SLAs, SLOs and instrumentation outrank user-facing features. The role runs on what Paul Graham calls the manager schedule, which Chang found wore him out.
• The move to engineer was about fit, made workable by transparency. In 2021 Chang moved to engineer on the same team. He wanted maker time, and the team was about to rebuild Minerva from scratch while running the legacy system, because cascading backfills under data versioning had become prohibitively expensive. His director sponsored the switch. Chang got through the awkwardness by naming it openly, taking small projects, pair programming, and showing proof of effort and proof of work, which he calls leading indicators of proof of worth.
• Each role should borrow the others' perspective. Chang says engineers must sell what they build: a launch announcement that only restates the feature will flop. PMs should look for the class of problem behind several requests, as an infrastructure maintainer would. Data practitioners add most value from "0.5 to one", bringing measurement rigour and knowing how logging and instrumentation work. A host adds empathy for the PM's triage load, and the engineer's habit of thinking through edge cases.
• Excel at your current role, then find a sponsor. Chang's sequence: do your present job very well, because people extrapolate from it; build social capital; explore adjacent work as proof of work; then find a sponsor with the authority to bet on you, which he calls necessary though perhaps not sufficient. A host argues that an internal move beats switching companies. On GenAI, Chang uses coding assistants for unit tests and edge cases and LLMs as tutors, and calls vibe coding probably still too good to be true.

2025-05-14 · open the original ↗

• The US may have neutralised China's nuclear bargaining chip with Riyadh. China had long dangled nuclear energy before Saudi Arabia, which wants to run on renewables and export its oil, and is its largest oil customer; the Saudis are also angry Russia overtook them as China's top supplier. He says this reverses his previous day's view that Riyadh could play both powers, but the captions end before he explains how the US outmanoeuvred China.

2025-05-14 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-14 · open the original ↗

• The captions list the rack's contents but give no cost. The host walks the UniFi Enterprise rack at Ubiquiti's RSAC 2025 booth, top to bottom: an Enterprise 24 switch, a dual stack of EFGs, 100G aggregation switches, four 48-port Enterprise switches, three AI Keys, four Enterprise NVRs and a Power Distribution Pro. No prices are stated, despite the title.

2025-05-14 · open the original ↗

• Weigh feedback by the expertise and motive of its source. Hanselman said a TED speaker-coaching session had just told him "there's not a talk here". He decided to defer to the organizers' 13 years of running the event, since they want a good event and are not trying to hurt him. He argued younger people tend to read such feedback as an attack rather than ask what the giver wants.
• "Data-driven" can be weaponized to block decisions. Mark said "where's the data?" often works as a smoke screen from teams that simply do not want to do something, followed by one objection after another. He added that new product categories cannot be backed by data in advance, citing a keyboardless phone.
• Assume good faith until there is real evidence otherwise. Mark argued the people who assume everyone is acting in bad faith lead a miserable life, though he does not deny that people have hidden motives. Hanselman said how much good faith you can extend depends on culture, age, confidence and your standing during things like layoffs.
• Separate feedback on the idea from its pitch and context. Mark recounted pitching what became Azure Resource Manager, a declarative, extensible application model, to the heads of Azure engineering and product, and getting a lukewarm "interesting". He blamed his selling and their priorities, kept his conviction, and the idea was built anyway. He also relayed a roughly 20-year-old study finding judges sentenced more harshly after lunch, and Hanselman said about half his podcast guests thank him for asking whether they are in a good headspace.
• Receiving feedback is a skill, and you can choose what to take. Mark said holding your breath makes criticism harder to absorb, and that "feedback bookends" around criticism signal positive intent. Hanselman called feedback "a cafeteria". Mark said a brutal speaker coach some 10 years earlier caught tics he never knew he had, and he left some of them alone as "just who I am."

2025-05-14 · open the original ↗

• ROI is the first question before funding any AI project. Roese, Dell's global CTO and chief AI officer, says AI is useful only in service of a measurable outcome tied to an improvable process: profit, revenue, margin and cost for a company, educational outcomes for a university. AI that gives salespeople more time with customers, which Dell did, counts; AI that makes them slightly happier is meaningless unless measured.
• Ask where AI should be applied, not where it could be. After the ChatGPT launch Dell received about 800 generative AI ideas, many of them, Roese jokes, inspired by Star Trek, and found it could not pick winners from them without context. Dell flipped the process: start from what it measures (profit, revenue, cost, regulatory risk), pick four areas (supply chain, sales, services and engineering), and find the biggest lever in each, such as sellers' content preparation and core coding in engineering. The host's research says the 800 were narrowed to eight.
• Every enterprise is stuck at the same starting problem. Roese says every customer he talks to has hundreds of ideas and struggles to choose where to begin. Running 800 concurrent AI projects is impossible even for the largest companies, and debating them all leaves nothing in production.

2025-05-13 · open the original ↗

• Saudi Arabia is extracting the most from both America and China. He likens it to a reality-TV love triangle, with the one difference that only the US can defend Saudi oil fields.
• The US tie rests on oil security, but interests have partly decoupled. Standard Oil found and built Saudi oil nearly a century ago; the Saudis later took control and twice brought the world to its knees in the 1970s. The 1974 petrodollar deal priced oil in dollars in return for US protection of the wells and royal family, possibly against Iran.
• China's pull is trade and, decisively, a willingness to offer nuclear power. China is the top buyer of Saudi oil and top trading partner, with roughly $100 billion in annual two-way trade; 50 of Saudi Arabia's 100 overseas projects involve China, and Beijing brokered the Saudi–Iran embassy deal two years ago. He reads the nuclear offer as China's clincher for now.

2025-05-13 · open the original ↗

• Many social media feuds are staged for growth and monetization. He says he has been approached about four times by creators proposing a fake online beef and refused each time; he names no one but says the bigger and more aspirational an account, the likelier it has been asked or taken part.
• Social media will drift toward scripted content as reality TV did. He expects coordinated, scripted conflict between creators to become more prevalent over time, and says being asked to join one is a quick way to lose his respect.

2025-05-13 · open the original ↗

• Fund only AI that moves a measurable business outcome. Roese, Dell's global CTO and chief AI officer, says Dell's roughly 800 post-ChatGPT ideas could not be ranked without context, so it asked where AI should apply: profit, revenue, cost and regulatory risk, across four pillars (supply chain, sales, services, engineering). His gate to production is a material ROI, alignment with how Dell wants to run in future rather than "a big blanket" over bad processes, then technical viability and security and compliance. That gate is how he says companies escape "PoC prison".
• High-ROI first projects start a flywheel that funds the rest. Roese argues later projects reuse the first one's technology and so cost less and ship faster, while starting with a goodwill project produces no fuel; Dell left G&A functions for later. Sales went last because its data needed fixing first, after which Dell Sales Chat went live for 20,000 sellers and, he says, exceeded expectations. Before ChatGPT, Dell ran 500–1,000 AI experiments a year, 80% with zero business impact, which he defends as building familiarity.
• Agents digitise narrow skills, putting humans "on the loop". Roese says enterprises differ by proprietary data, which RAG chatbots have unlocked and which accounts for almost all of Dell's production projects, and by unique skills, which agents digitise. Like geofenced self-driving cars, agents do one job autonomously from an objective; whole-human AGI/ASI is "a long ways away". He says agent architecture differs too: an LLM brain plus a knowledge graph for memory and tool use to act.
• Agent interworking standards will emerge by consensus this year. Dell has run multi-agent systems since its first two-agent research-report writer, once flexing to 1,600 agents that could "hire" others. Roese says authentication, authorisation and cross-company knowledge sharing are unsolved, praises MCP for model-to-data and Google's Agent2Agent, announced the day before recording, and is "100% confident" de facto approaches arrive before year-end, ring-fenced at "level four autonomy" and reusing tools such as homomorphic encryption.
• AI will create named new jobs, but construction is the biggest. Roese lists software composers who decide architecture without writing code, "thermal plumbers" who manage liquid-cooled GPU clusters, and AI explainers who deliver AI findings empathetically, as with genomic risk results, and says such roles are forming inside Dell. Above all, he says, the AI infrastructure build-out exceeds the US infrastructure bill of several years ago. He also forecasts that viable quantum computing, "a few more years" away but not decades, will be a bigger disruption than ChatGPT, making AI three to five orders of magnitude faster.

2025-05-12 · open the original ↗

• The U7 Pro Outdoor's 6 GHz radio only matters in high-density deployments. It adds 6 GHz to 2x2 MU-MIMO on all three bands, which the original U7 Outdoor lacks. The host expects the gain at venues like festivals with many Wi-Fi 7 clients, not in a home backyard; 6 GHz is opt-in at adoption since it is not legal everywhere.
• It is a $279 outdoor AP with weatherproofing and PoE+. That is the US MSRP at launch, after a long wait for FCC approval. It has a 2.5 GbE port powered by PoE+, IPX6 rating or IP67 with the included cable gland (which the host recommends), wall and pole mounts, and roughly 300-client support.
• The bundled external antennas make only 2.4 and 5 GHz omnidirectional. Without them all three bands are directional, about 90° per Ubiquiti's included diagram. The host can't say why 6 GHz stays directional and guesses the form factor lacked room for a third antenna; the antennas must also be enabled in UniFi, not just screwed on.

2025-05-12 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-12 · open the original ↗

• Trump and Xi may be staging a crisis to engineer a boom. He calls it "tinfoil hat" but harder to dismiss after both sides cut tariffs by 90–100% for 90 days, leaving an effective US tariff on China of 40%. His speculation: raise barriers, unwind them slowly, then remove them for a market surge; he says he is not fully convinced.
• Soaring stocks mean little while Treasuries keep falling. Twenty minutes into trading the Dow was up 970 points, the Nasdaq 677 and the S&P 500 151, but bond prices kept dropping and yields rising, making US debt costlier to service. He allows that could change within the 90-day window.

2025-05-12 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-12 · open the original ↗

• The authors benchmarked Polars on GPUs themselves, with vendor hardware. Janssens says Nvidia approached them on LinkedIn and Dell supplied a machine in its US lab where five professional workstation cards, including the RTX 6000 Ada generation, could be swapped. The pair insisted on producing their own numbers rather than copying promotional material, and the results are in the appendix of their book. The host notes Dell and Nvidia are two of the podcast's biggest 2025 sponsors.
• Polars' GPU engine sits after its query optimiser. Nieuwdorp explains that Nvidia's RAPIDS team, maker of the GPU dataframe library cuDF, did not simply map the Polars API onto cuDF functions, because much of Polars' speed comes from its optimiser. Instead it built, with Polars, a GPU engine that takes the optimised plan; it was in open beta at recording.
• GPU gains start on modest data, even on smaller cards. Nieuwdorp says data transfer to the GPU adds overhead, so speed-ups appear as data grows, but already from around 1 GB, sooner than the promotional material suggested. Even cards with fewer processors showed big speed-ups.

2025-05-12 · open the original ↗

• De-dollarisation talk tracks the cost of dollar-priced oil, not the dollar's decline. He notes about 85% of oil trades in dollars; China pays dollars except for Russian crude, bought in yuan at deep discounts. The talk peaked late 2022 with oil near $120 and persisted while the DXY was high, around 106 even at $75 oil.
• Cheaper oil and a weaker dollar have silenced BRICS boosters. With the DXY near 100 and oil about $60, briefly in the high $50s, he hears almost nothing; he rejects the idea that critics are quiet because the dollar is already finished. He bets the propaganda returns if oil climbs to $80–85 with the DXY at or above 100.

2025-05-11 · open the original ↗

• uv replaced Poetry and Docker for the book's reader environment. Nieuwdorp says the authors started with Poetry plus separate Python-version tools and considered Docker for the companion notebooks repo, but settled on uv because readers need only install it and run uv sync, which sets up the right Python version and dependencies on any system.
• Tooling performance is a feature in itself. Nieuwdorp places uv in the trend of Rust-based tools, like Polars, and says it can be more than 10 times faster.
• uv's speed makes throwaway environments practical. Nieuwdorp says uv can create an ephemeral environment for a single command, which he used to benchmark successive Polars versions. He found a roughly 10% slowdown around version 1.2-something, traced it to two queries, and used a git-bisect script to identify the commit. The Polars team refactored that code the same week, which resolved it, though they had not pinned down the cause.

2025-05-10 · open the original ↗

• The administration wants the outrage over ICE raids. He argues the nationwide raids, including in Worcester, Massachusetts, galvanise the left, please or resign the right, and frighten the middle about civil unrest. His prediction: if protest turns into riots and property destruction, it produces another Republican wave in 2026 and a more Republican Congress.

2025-05-10 · open the original ↗

• Tables are underserved by Python's visualisation tools. Janssens says that despite a myriad of plotting packages, Great Tables, created by his colleagues Rich Iannone and Michael Chow, is the only one he knows for producing presentation tables with colour, currency formats and mini graphs. Polars exposes it through the df.style accessor.
• Format how data is shown; never change the data. Janssens objects to a post he saw that added dollar signs by converting numbers to strings. Styling should be a separate representation layer, he argues, so the underlying floats and integers stay intact for later calculations. He corrects the host's reading that the point is computational cost.

2025-05-10 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-10 · open the original ↗

• Falling Treasuries alongside rising Bitcoin signal lost trust in US debt. In uncertain times investors normally buy Treasuries, especially the 10-year, pushing yields down while risk assets like Bitcoin fall. The reverse is happening, which he attributes to trade wars and tariffs. He discounts the new US–UK trade deal's easing on steel and aluminium, since a 10% baseline tariff on UK imports remains.

2025-05-09 · open the original ↗

• The India–Pakistan clash is not yet a war. India has launched air strikes into Pakistan and is braced for retaliation, but he rejects influencers calling it war. He cites three wars, in 1947, 1965 and 1971, and over a dozen minor skirmishes over Kashmir and other disputes since.
• A real India–Pakistan war could tempt China to grab Himalayan land. The two share about 2,100 miles of border, and he notes China did it before: a clear 1962 victory took part of Jammu and Kashmir, though China was "wiped out" in a 1967 clash. The captions end as he argues China's military is far stronger now.

2025-05-09 · open the original ↗

• Nvidia packages models as microservices so they can be swapped. In episode 883, Nvidia's Sama Bali explains that NIM (Nvidia inference microservices) containerise open, proprietary and Nvidia models pre-tuned to the GPU, because models turn over every couple of months, such as Llama 3 to 3.1 to 3.2. Dell's Logan Lawler says a NIM loads with one line of code. Bali says CUDA libraries such as RAPIDS cuDF can speed pandas- or Polars-style preprocessing by 100x on RTX 6000 GPUs without code changes, and cuML up to 50x for scikit-learn tasks.
• AWS's custom chips rest on Annapurna Labs' Nitro work. Emily Webber of AWS (episode 881) says Amazon acquired Annapurna in 2015 to break up the monolithic hypervisor. Its Nitro system physically separates customer data from instance control, and every modern EC2 instance runs on it. She says more than half of new AWS compute is Graviton Arm CPUs, and calls Trainium 2 the most powerful EC2 instance for AI/ML, chosen mainly for price-performance and energy efficiency.
• Data scientists outnumber the engineers who deploy their models. Greg Michaelson, Zerve's co-founder and chief product officer (episode 879), calls the hand-off a problematic dependency where mistakes creep in. He says Zerve gives every canvas a Docker container so environments are shared, serialises every variable so a trained model can be pointed at for an API route, and deploys on serverless Lambdas, letting data scientists ship their own software.
• Heterogeneous integration scales chip performance beyond shrinking transistors. Kai Beckmann (episode 875) describes combining dies into one system, such as stacking memory or placing it beside a GPU (CoWoS-style structures) to shorten data transfer. He says it needs front-end-like precision, heat and power management, and new materials, which his company develops with AI-driven "materials intelligence" to cut unnecessary experiments.
• "AIPC": accelerated, individualised, private, cost. Shirish Gupta (episode 877) gives his mnemonic for on-device AI. His examples: an ER physician's report drafts that learned the doctor's style, data that never leaves the device, and a financial-services customer spending 15% of data-centre compute on developers' code generation, which it wants offloaded to laptop NPUs. He concedes cloud inference costs are falling fast but argues for "the right engine for the right workload".

2025-05-09 · open the original ↗

• India's strike on Pakistan is too early to call a war. He reports India struck what it calls terrorist and militant sites after an attack on Hindu tourists in Kashmir, while Pakistan says mosques and civilians were hit, calls it a blatant act of war and is retaliating. He notes three wars since 1947, the third in 1971 over Bangladesh, and nearly a dozen strikes since.
• A full war would pull China and the US in as mediators. He expects both powers would have to step in quickly between two nuclear states. The captions end before the promised history of the Kashmir dispute and the global consequences.

2025-05-09 · open the original ↗

• The India-Pakistan strike may not amount to a war. India struck Pakistan after an attack on Hindu tourists in Kashmir blamed on suspected Islamabad-linked militants; India says it hit only terrorist and militant sites, while Pakistan says mosques and civilians were hit and has vowed retaliation, calling it an act of war. He argues that's premature — the two countries have fought three wars, in 1947, 1965 and 1971, plus roughly a dozen further armed clashes since.
• Kashmir's Hindu ruler traded autonomy for Indian protection in 1947. Maharaja Hari Singh, ruling a Muslim-majority princely state, asked for Indian army protection against Pakistan and ceded Kashmir to India in return, triggering the first India-Pakistan war; a 1962 border war with China then left China holding a further slice, so Kashmir today is split between three claimants.
• China leans on Pakistan as a buffer against India's rise. He notes Pakistan owes China roughly $30 billion and depends on it to counter India, while the US relies on India — which he puts as the world's fourth-largest economy — as a counterweight to China, even as it separately depends on Pakistan for counterterrorism cooperation, leaving Washington reluctant to fully back either side in a war.
• He expects a full war would alarm both Washington and Beijing. A full Indian mobilization would worry China, and an Indian victory that further destabilizes Pakistan — which he says has defaulted on its debts three or four times in the past five years — would erode China's leverage against India; he predicts the two powers would rather broker peace quickly than let the conflict escalate, and calls a slide into a world war unlikely given how many regional conflicts already run concurrently.

2025-05-08 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-08 · open the original ↗

• Montemagno's vibe coding is a reviewed flow, not ignoring the code. He rejects "forget the code exists" and describes reviewing agent output as it streams, stopping and correcting it in real time. Hanselman set this against Simon Willison's line that code an LLM wrote but you reviewed, tested and understood is not vibe coding at all, and said he would not vibe code something into production. Both work at Microsoft, and Montemagno acknowledged "some bias" from working closely with the VS Code team.
• He shipped a production app writing almost none of it. His sentiment-analysis tool for Reddit and YouTube comments, with a Blazor front end, Azure Functions and Azure AI, came to about 17,000 lines in his PR to David Fowler, of which he wrote 20, the Reddit JSON parsing the agent could not solve. It went from zero to deployed in three days of spare time. He said it would otherwise have about a tenth of its features and saved "weeks" of work, and that the tools had transformed how he works in the previous three to four months.
• Persistent context is what makes agents reliable. He keeps a custom-instructions file, sent with every call and so kept moderate in size, that records what he likes and bans, such as "never write a test unless I tell you". Reusable prompt files carry decision trees for new services, and a co-worker's instruction makes the agent commit after every prompt. Hanselman compared carrying these prompts and MCP servers to his decades-old utils folder and dotfiles.
• Trust tracks expertise: review your strong areas, delegate the toil. Montemagno controls the C# structure, services and dependency injection himself but leaves CSS and JavaScript to the agent; a full restyle took 15 to 20 prompts over about 30 minutes. He called a React app he generated without reading the code a prototype, and he picks models by temperament, Claude for exploring and GPT-4.1 to "just do the thing".
• Beginners should learn fundamentals and use the tool to teach. Montemagno said a new developer should start in ask and edit modes and instruct the agent to explain, ask questions and link to code, and that data structures still matter. He conceded the setup is non-trivial and that maintainability a year or two out remains open. He used MCP servers (GitHub, Perplexity, Context7) to have the agent review his codebase and file its top 10 issues.

2025-05-08 · open the original ↗

• Polars rescued a pandas pipeline that could not scale. Janssens says Alliander, the Dutch grid operator which runs electricity and gas infrastructure for a third to half of the Netherlands, needed to cut cost and process far more data, but its Python and pandas code was inefficient and ran on a single AWS instance with over 700 GB of RAM. He believes it was one of the first companies with Polars in production, over a year before Polars 1.0.
• Win over a sceptical team with one small benchmark. Janssens says the team was understandably wary of a library promoted by three consultants from Xomnia, where Polars was being developed. The consultants rewrote a piece of low-hanging-fruit pandas code in Polars and showed the speed and memory numbers. The full migration then went piece by piece, idiomatically rather than line by line, until R and pandas could be dropped as dependencies.
• Memory per job fell more than tenfold. Nieuwdorp puts it at about 500 GB down to 40 GB, and attributes the drop partly to Polars and partly to restructuring code along the way.
• Production use and book-writing exposed Polars' gaps. Janssens says real work surfaced missing features, such as weighted random sampling that pandas has, while filling in the book's reference tables revealed naming inconsistencies and missing operators. The authors fed these back to the project as issues and some pull requests.

2025-05-08 · open the original ↗

• This post includes paid content that is not available publicly. See the original source.

2025-05-07 · open the original ↗

• Quiet biases are more worrying than obvious ones, being harder to find. Will Landecker, founder of the consultancy Accountable Algorithm, separates clear, measurable harms, such as loans approved at different rates, from "whispers", such as slightly different interest rates that only a large pile of data and a strong hypothesis can reveal. A tiny gap multiplied by 8 billion users is an enormous effect, and finding these gaps takes proactive searching.
• "Good" usually means accurate on the majority, which entrenches gaps. If 90% of applicants are men, an accurate model can still give women worse rates, and Landecker says executives everywhere he has worked could ask "what does the data say?" without anyone scrutinizing it. A portrait generator that worked for him hallucinated the features of a mostly Black nonprofit's members, which he calls an evaluation-set failure as well as a training-data one. Evaluating only on current users locks a product out of new markets, and a host adds that A/B tests should be read by segment.
• Machine-to-machine feedback loops strip out the human checkpoints. Landecker's main agentic concern: when agents interact directly, independent errors may cancel out, but correlated errors push the system in their direction, faster than human-speed review or user reports can catch it. He notes that long-running projects that collect human-written text now say they can no longer guarantee incoming text is human-written.
• Using one LLM to test another for bias is fraught. Subtle bias, such as different responses when pronouns or racially coded names are swapped, needs massive varied datasets, and generating them with an LLM leaves no clear party to blame when a disparity appears. Landecker recalls Word2Vec at Twitter around 2013 completing "man is to woman as doctor is to" with "nurse". He calls the current shotgun of approaches (LLM judges, human red teams, interpretability) passable but not enough.
• Every AI organization should ask how its success metric lies to it. This is Landecker's one recommended habit: enumerate the compromises behind a measurement, such as a cheap, high-coverage data source or a quick evaluation function with no human in it, and ask how they mislead, rather than whether the number went up. He also advises finding an executive sponsor and treating internal policy as the main guardrail, since he is not hopeful about government action. He recommends the FAccT conference proceedings.

2025-05-07 · open the original ↗

• OPEC+ output increases will hurt Russia more than America. He concedes to Mr. Global, the energy creator, that higher OPEC+ output will hurt US energy firms and raise US natural gas and pump prices. But he argues Russia is far weaker: official March inflation of 10%, probably understated, a 21% interest rate against the US's former 5.5%, and Russian economists warning of serious risk.

2025-05-07 · open the original ↗

• Polars code reads as declarative paragraphs, not nested brackets. The authors say pandas code is full of brackets and often hard to reason about, whereas Polars expressions (in their book's metaphor, recipes whose operations are steps and whose functions are cooks) compose into queries that read like a paragraph. Users declare the end result and leave processing and optimisation to the engine.
• Polars was shaped by pandas' failures and strengths alike. Relaying conversations with creator Ritchie Vink, the authors say a formative frustration was pipelines crashing 20 minutes in on errors that could not be seen up front. Polars kept what Vink liked in pandas and borrowed from Spark, whose syntax it resembles, and from Rust. They stress they are not bashing pandas: without Wes McKinney's library there would be no Polars.

2025-05-06 · open the original ↗

• OPEC's faster output increase is crashing an already weak oil price. OPEC holds about 70–75% of world oil reserves and supplies about 40% of world oil, and it has started raising supply faster than expected into a market he describes as already falling on weak demand.
• Cheap oil starves Russia's war funding, as in the 1980s. He says 63% of Russia's total exports are fuels and energy products, crude the largest share, and that this revenue funds the war against Ukraine. He reads it as most likely a repeat of how America brought down the USSR in the 1980s and '90s.

2025-05-06 · open the original ↗

• A point-to-point bridge is a "wireless wire", not client Wi-Fi. An access-point radio on the gateway's LAN shoots to a station radio that terminates into a switch, and phones or laptops need a separate Wi-Fi AP on the far side; it is also distinct from mesh kits like Orbi or Deco. Point-to-multipoint is the "Y cable" version WISPs use, as in the firm's Cerro Gordo ghost-town deployment.
• Band choice trades speed against range and line-of-sight tolerance. 2.4 GHz reaches far and tolerates light trees; 5 GHz is faster but stricter; 60 GHz gives gigabit-plus but needs precise alignment and suffers in fog and rain. For a detached garage the host says a roughly $100 pair of 2.4 or 5 GHz LOCO radios suffices.
• Ubiquiti's Wave line is 60 GHz with a 5 GHz backup. Point-to-point: Wave Pico $179 (1.3 km), Wave Nano $279 (5 km), Wave LR $379 (about 8 km, mostly a far-customer station), Wave Pro $599 (15 km paired). Wave AP Micro (90°) and Wave AP (30°, longer reach) are $499 PtMP access points; the 2 Gbps rating is 1 up plus 1 down, falling with distance, and backup runs about 200–300 Mbps.
• Wave radios now take DHCP, but should be fixed. Unlike older gear defaulting to 192.168.1.20, they lease an address; the host pins them with a reservation or a static IP outside the DHCP pool. Setup on firmware 3.4.0 means switching wireless mode from the point-to-multipoint default, setting one side to main mode, and matching link name and strong security key.
• The host's demo link ran near his gigabit line rate. Two Wave Picos side by side showed about 1.5 Gbps of 60 GHz link capacity and a speed test of about 900 Mbps down and 720 Mbps up. The video promotes design help from Rogue Support, Crosstalk's sister company, for picking among the overlapping models.

2025-05-06 · open the original ↗

• One bank's Copilot rollout reached near-total adoption within a year. Peterson, a Microsoft customer success account manager covering banking, says the customer, a Midwest bank with a 10,000-person workforce using M365 Copilot with company and web data, started a pilot about a year before recording and has activated over 1,000 licenses at about 97% adoption. She reports use in its call center, finance and IT teams, and quotes the customer calling Copilot "a game changer" and an SVP calling it "my new best friend".
• Addressing oversharing fears unblocked the rollout. The bank's initial restrictions came from security concerns, such as whether an employee could use Copilot to find the CEO's salary; Peterson says Microsoft's security controls and configuration, plus third-party solutions, resolved them. She credits the speed, in an industry Giard calls slow to adopt, to the account team's trusted-advisor relationship, contrasting organizations that build in-house chat tools until they find they are not getting the benefits.
• One team's visible success spread demand across the bank. Peterson says the finance team was among the first to embrace it, using prompts to pull data from various files into dashboards of trends, forecasts and volume, work she puts at one to three days without Copilot. Seeing that team's results, others wanted to use it too.
• Structured programs, not the tool alone, drove the adoption. After the bank's AI council approved the tool, Microsoft's AI Experience Accelerator put it in a room with peers from other banks to identify use cases and rank them short, medium or long term; around October of the prior year the customer asked for a "promptathon", an on-site prompt-design hackathon led by Microsoft's subject matter experts. A partner was brought in to roll out at scale because the bank did not want to take six months.
• The rollout's estimated value is deliberately conservative. Peterson says a Microsoft business-outcome engagement put the potential value at $21 million, from time saved, process automation and customer-service impact, and that the real number was "probably bigger". She describes the next step, already starting, as custom agents built in Copilot Studio for specific business processes.

2025-05-06 · open the original ↗

• Polars' declarative style makes pipelines easier to read than pandas. Janssens and Nieuwdorp, authors of O'Reilly's Python Polars: The Definitive Guide (released on the recording date, 1 April 2025), say pandas code is bracket-heavy and hard to reason about, while Polars expressions compose into a query that reads like a paragraph, leaving optimisation to the engine. They credit pandas as Polars' foundation, and relay that creator Ritchie Vink was shaped by pipelines crashing 20 minutes in, borrowing from Spark and Rust.
• A production migration cut memory needs by an order of magnitude. At Alliander, memory per job fell from about 500 GB to 40 GB. The Dutch grid operator's pandas pipeline ran on a single AWS instance with over 700 GB of RAM. The authors won over a sceptical team by benchmarking one small rewrite, then migrated idiomatically until R and pandas were dropped. Nieuwdorp says the same time frame that once fit 25 simulation samples later fit the 50 stakeholders wanted, with code restructuring contributing alongside Polars.
• The GPU engine pays off on modest data sizes. With five Nvidia workstation cards in a Dell lab, both companies being podcast sponsors, the authors ran their own benchmarks for the book's appendix rather than reuse promotional figures. Nieuwdorp says Nvidia's RAPIDS team built an engine that takes Polars' optimised plan instead of mapping the API onto cuDF. Speed-ups appeared from around 1 GB, sooner than promotional material implied, and even on smaller cards.
• uv is fast enough to change how tools get used. Nieuwdorp says uv replaced Poetry and a planned Docker setup for the book's repo, since readers need only uv sync. Its speed allowed throwaway environments per command, which he used to benchmark Polars versions, find a roughly 10% regression around version 1.2-something, and git-bisect it to a commit.
• Format a table's presentation, never its data. Janssens says Great Tables, by his colleagues Rich Iannone and Michael Chow and exposed in Polars via df.style, is the only Python table-styling package he knows. Turning numbers into strings to add dollar signs breaks later calculations. He also recounts rewriting most of the visualisation chapter (16) when Marco Gorelli switched Polars' plotting backend from hvPlot to Altair, a feature then marked unstable.