Souk's Notes - Unofficial Summaries
AI Analyst LabHow to Find the Moment that Transforms your Career #careeradvice #podcast #tech #jobsearch
- One chance encounter can snowball a whole career. The speaker says each of the people speaking caught a break somewhere that led to where they are now, and tells viewers to watch for theirs.
AI Analyst LabHow to Find Work-life balance and Love What You Do #podcast #shorts #trending
- Work-life balance matters when work isn't for yourself. The speaker says the question only arises when you are working for the job rather than for yourself; while they love what they work on they don't think about it, and only would if that stopped.
Polis Pandit$Where Has Joe Biden Been?
- This post includes paid content that is not available publicly. See the original source.
Mr. GlobalHow the minimum wage impacts everyone not just those earning it.
- The minimum wage anchors pay well above the minimum. Rebutting a viewer who said it no longer matters, he argues the $7.25 federal minimum is the starting point for wage negotiation, so people earning more still earn less than they should; he puts the average hourly wage at a little over $11. His illustration: at a $2 minimum, earning $4 is "twice the minimum" yet still leaves no money in your pocket.
AI Analyst LabHow to Pivot to Tech and Build a Path to Success #podcast #datascience
- A pivot into tech may mean taking a lesser role first. The speaker says they applied to everything, interviewed for probably five to seven data science roles and were rejected for all, but took an Amazon business analyst role at a 40% pay cut.
Polis Pandit$The Fall of al-Assad and the Major Implications for the World
- This post includes paid content that is not available publicly. See the original source.
Mr. GlobalYes, you can make gasoline from tar sands.
- Canadian oil sands crude can be made into gasoline. Rebutting a viewer's claim that it is too dirty and only good for byproducts, he says gasoline can be made from almost any fossil fuel, even coal; once upgraded, oil sands bitumen is basically oil and is refined like any other. He concedes it is much dirtier and more energy-intensive to produce.
- Any oil removed from the market moves gasoline prices. He argues the grade is irrelevant: taking any barrels off the market, Canadian oil sands or Venezuelan, raises the price of oil and therefore of gasoline.
AI Analyst LabEmbracing the Unexpected: From Super Bowl to Data Science to Entrepreneur #careeradvice #podcast
- An unexpected path can be worth more than the money. The speaker says that had they taken a certain offer out of college they would probably have been a millionaire before 30, but would not trade having danced at the Super Bowl for any amount of money.
Mr. GlobalThe Federal Minimum Wage is 4.93 an hour not 7.25. I explain why.
- Inflation has cut the real federal minimum wage by a third. The $7.25 an hour set in 2009 now has the spending power of about $4.93 (he also rounds it to $4.90), so in his framing the effective minimum wage is $4.93.
- Politicians promise lower costs instead of higher wages. He argues politicians and big business prefer pledges to cut gas, heating, energy and mortgage costs, saving people $100–200 a month, over paying them perhaps $400 a month more. In his reading this appeases workers with a little breathing room while keeping wages near subsistence.
Mr. GlobalThe Dispatch (And a promo code!)
- The media's main problem is self-importance, not bias. He relays a Jonah Goldberg article in The Dispatch arguing the media believes it can define what matters to the public; exit polling shows the issues it hammered during the election ranked at the bottom of voters' priorities. The video is also a promo: code "Mr Global" gives a free month of The Dispatch.
AI Analyst LabHow to Break Into Data Science in 2024 #podcast #datascience #careeradvice
- Break into data by going deep, then building toward real data. The speaker's framework: pick the easiest entry tool, such as SQL or Excel, and go deep; find the business domains you care about and build projects on Kaggle datasets; volunteer with real data; then apply for jobs.
Technology and FriendsMichael Feathers on Working with Legacy Code [EPISODE 833]
- Legacy code is code without tests, and the code supplies expected values. Feathers chose the definition because good test coverage tells you quickly whether a change broke something; without it you are "hunting in the dark". To test code you do not understand, he puts in a dummy expected value such as zero, runs it, and plugs in whatever comes back, so the test records current behaviour. He says writing the tests is usually easy and breaking dependencies enough to get them in place is the hard part, so tests should go as close as possible to what you are about to change, sometimes spanning four or five classes.
- After a refactoring bug, fear can make a codebase decay faster. The business reason to refactor, he says, is that code people cannot understand slows them down and leads to wrong changes. But a couple of refactoring bugs can lead management to stop it, and the resulting caution lets things get worse at an accelerating pace. Teams have to accept a few errors early, aiming for fewer over time, and refactor the areas they care about aggressively.
- Temporary instance variables and throwaway tests can unlock huge methods. For methods of 1,000–2,000 lines with up to 15 levels of indentation, which refactoring tools only partly handle, he saves intermediate state from five or six lines into instance variables and writes tests against them, then deletes the variables and tests once the refactoring is done. For large classes, his book suggests sketching which methods depend on which methods and data to find classes to extract, which in statically typed languages can be done before any tests exist.
- AI-generated tests are useful because passing tests characterise behaviour. Feathers says he resisted at first, since writing tests is how you build understanding, but a generated test that passes is correct in the sense that it pins current behaviour, and it can be kept or deleted. He calls chunk size the hardest problem: generating hundreds of tests or classes leaves you 10 minutes to half an hour reviewing output, so he breaks work into pieces that fit in his head, and treats hallucination as a reason to keep a human in the loop.
- Change concentrates in few files, so small improvements pay back quickly. He shows teams a plot of each file's commit count, sorted, where some files get 500-odd commits over a project's life and others one or two. On that basis he argues a 3-million-line codebase is in practice more like a 500-line one, so improving what you touch, the "scout rule", soon returns the benefit and helps morale.
GCastAutomating Microsoft Excel with VBA [GCast 189]
- Excel's VBA tools sit on a ribbon hidden by default. Giard enables the Developer ribbon under File, Options, Customize Ribbon, which gives access to the Visual Basic Editor; he says this is part of why he had forgotten about VBA after using it a lot years ago.
- Hand-written VBA works but demands knowing the language and object model. He writes a short sub that sets cells A1 and A2 to text values and then selects A3, with IntelliSense listing an object's properties and methods after a dot and Ctrl+Space auto-completing. His judgment is that it is very flexible, but without good knowledge of VBA and Excel's object model you are constantly looking things up.
- The macro recorder writes the code for you, then you fix it. Recording requires saving as a macro-enabled workbook; he records copying a data block to a new worksheet, pasting values only (no formulas), and formatting one column as a date and another as currency. The generated code lands under Modules in the editor, but it hard-codes the range A1:D15, assumes the right sheet is active, and refers to the new sheet by the name it happened to get during recording.
- Generalize recorded code with named ranges, sheet indexes and unique names. He selects the first sheet explicitly, swaps the fixed range for a named range so added rows are not missed, and refers to the new sheet by index 2 since it is created after the first (VBA arrays start at 1, unlike most languages). To avoid a name clash on reruns he builds the sheet name from "Gcast-" plus a random number times 10,000, converted to an integer, then a string, then trimmed; he notes a repeat could still eventually cause an error and suggests conditional logic to guard against it.
- View, Macros runs, debugs, edits and assigns shortcuts without the Developer ribbon. From the View menu, which he notes everyone has access to, a macro can be run, deleted, edited, or stepped into line by line with F8 in the editor. Under Options a description (documentation only) and a shortcut key can be set; he assigns Ctrl+Shift+B and runs the macro with it.
Scott HanselmanThe Return of Scott's Wife Mo
synopsis
Hanselman's wife Mo, a nurse, returns after three or four earlier appearances. The couple, together 24 years and approaching 25, end up agreeing that what has held a "mixed marriage" together is presuming they are on the same team and saying expectations out loud, while they still differ on how much latitude to give their teenage sons.
A mixed marriage. Mo says race barely registers between them; the difference is how each was raised. She grew up in Zimbabwe in a family of seven children with an ethic of ubuntu, respect for elders and a village that disciplined other people's children; she notes their sons still call every family friend "Auntie". Hanselman sets this against American individualism, and against his own exit to the garage at 13 and an apartment at 19, which to her would have meant leaving the family short-handed.
Staying together. Hanselman recalls an early agreement to assume they are on the same side and act with positive intent, and a later saying, don't attribute to malice what is incompetence. Mo connects this to trusting him through his years of travel.
Parenting teenagers. With the older son at university and the younger, 17, at home for about another year and a half, they talk about the coming empty nest. Hanselman, quoting his mother, says a 19-year-old dropping off laundry and leaving is selfish but normal. Mo holds a firmer line on respect and asks whether he would tolerate the same tone toward his own parents. He answers that a father should be the one person a child can be a jerk to. Both say they have grown: he now gives space, and she grants an hour to decompress.
Change versus constancy. Mo once wanted the man she married to stay the same; Hanselman values self-improvement. They settle on improving while keeping the fundamentals fixed.
Advice and what comes next. His advice is that unspoken expectations breed resentment; hers is to communicate and protect couple time, which they once did with date nights and alternating "mommy day" and "daddy day". Chores have shifted, with him cooking and her doing laundry. For later life he imagines classes, a trainer and mentoring; she wants more school and medical volunteering abroad.
- Their deepest difference is culture, not race. Mo, born in Zimbabwe and one of seven siblings, describes a "we culture" of ubuntu ("a person is a person through other people"), where the village and neighbourhood aunties take responsibility for children. Hanselman, with one sibling, contrasts American individualism and moving out at 19; in her upbringing you left home only for university or marriage.
- Presume you are on the same side. Hanselman says an early rule of their marriage, now nearly 25 years, was to assume positive intent when something looks wrong, later joined by "don't attribute to malice what is incompetence". Mo ties it to trust during his years of frequent travel.
- They split on respect versus patience with teenagers. With sons aged 17 and 19, Mo has little tolerance for disrespect and asks whether he would let them speak that way to his parents; Hanselman, citing his mother, treats a selfish teenager as developmentally normal and wants the root cause. Both cite their younger son saying he was hungry and needed space, then apologising an hour later, as a success for each of them.
- Unspoken expectations breed resentment. That is Hanselman's main advice for newlyweds: tiny unspoken assumptions, such as who makes dinner, compound over 25 years. Mo's is to communicate and make couple time a priority; they say chores have since shifted, with him cooking and her doing laundry, without feeling imbalanced.
- Mo wants to give back once the nest empties. After a decade as a nurse following a mid-life career change, she wants more schooling and medical volunteering in less fortunate countries, to "send the ladder back down". Hanselman would take classes, get a personal trainer and mentor people, half-joking about stand-up or a talk show.
AI Analyst LabHow to know When Hard Work Does (and Doesn’t) Pay Off #growth #shorts #trending #career #money
- Long hours pay off only where learning and management are good. The speaker says putting in a lot of hours doubled their salary and sped up promotions, and that they make that investment only where they can learn and are well managed; elsewhere you can sacrifice work-life balance and get nothing back.
Mr. GlobalBREAKING - OPEC TO MAINTAIN CUTS
- OPEC is defending a price floor, not unwinding cuts. It is reported that OPEC decided at a meeting that day to keep its production cuts through March; he predicts a further extension at the end of March. He reads the aim as letting demand growth catch up with supply in a market he calls mostly balanced but tipping toward oversupply.
- He still expects an oversupplied market in 2025. He says demand will not catch up by the end of March and the decision will not change the oversupply, but it will limit how far oil prices fall in 2025.
- US oil production growth is flat despite better annual averages. He expects the 2024 annual average to come in about 400,000 barrels a day above 2023, but says that reflects a much stronger first quarter of 2024 than of 2023. Output now is barely above December 2023, and he expects it to stay flat.
Mr. GlobalWhy “Drill Baby Drill” Is Not A Thing
- Record output came with far fewer rigs. He shows rig counts peaked two and a half years earlier and have since lost over 250 rigs, more than 30% of the fleet, yet US oil and natural gas production hit records in 2023 and he says will again in 2024. He credits technology and science, not rig numbers, and says people in the industry grasp this least.
- He expects oil rigs to fall and gas rigs to rise under Trump. He says this is not driven by any Trump policy but by lower oil prices after two years of record production, and by higher natural gas prices after a period of low prices from overproduction. He expects the overall rig count to hold roughly steady only because gas rigs are added.
Polis Pandit$Before You 'Eat the Rich', Consider This
- This post includes paid content that is not available publicly. See the original source.
Crosstalk SolutionsThe ULTIMATE Peplink Configuration Guide!
- Peplink WAN behaviour is set by drag-and-drop priority. WANs at the same priority load-balance; lower ones are backups. The presenter, using a dual-modem Max BR2 Pro, notes every Peplink starts at 192.168.50.1 with admin/admin, which it forces you to change, and that Crosstalk, now a Peplink Gold partner, sells the units.
- Cellular standby and APN settings matter most. "Remain connected" keeps a backup modem hot, while "disconnected" saves data but takes minutes to come back. Each of the two modems has two SIM slots (four SIMs, two active), with LTE/5G and band locking available. A missing APN means calling the carrier; SmartCheck is the recommended LTE health check.
- Outbound policy routes traffic by rule, under HTTPS persistence. HTTPS persistence stays on top so sites such as banks don't see changing public IPs and lock you out. The demo puts guests on VLAN 10, enforces them onto Wi-Fi-as-WAN to spare cellular data, and adds a firewall rule denying guest-to-LAN traffic. Priority and lowest-use are the other algorithms.
- SpeedFusion bonding needs a Peplink endpoint at the far end. Bonding adds the WANs' speeds and can send duplicate packets for resilience. The far end can be the paid SpeedFusion Cloud (the easiest), a second Peplink, or a SpeedFusion Hub on a cloud host such as DigitalOcean, Vultr or AWS. Usage can be limited by application, SSID or LAN client.
- InControl 2 adds remote management and InTouch access behind CGNAT. It is free for the first year then a subscription, adopts devices by serial number, can preconfigure Wi-Fi and SpeedFusion before shipping, and opens the local web admin remotely. InTouch reaches internal web or SSH services behind double NAT or CGNAT, much like a Cloudflare tunnel, or a serial console for out-of-band access.
AI Analyst LabNavigating Data Science Jobs Transition Effectively | Ep5
synopsis
This episode of the Data Neighbor podcast follows a guest who went from quantitative finance analyst to business analyst, business intelligence engineer, data scientist at Meta and, less than two months before recording, full-time entrepreneur. The conversation argues that such moves come from knowing your current priority, taking an imperfect first step, building skills on the job and promoting yourself loudly, supported by confidence and relationships that formal career ladders never mention. It ends with the guest's interview checklist and a step-by-step route into data for people outside it.
From finance to a foot in the door. The guest studied industrial engineering at UC Berkeley and, as an international student, needed any job that would sponsor a visa; about 40 interviews yielded one offer, at Bank of America. A junior-year internship at Facebook had ended without a return offer, which the guest blames on imposter syndrome: almost silence in meetings, no requests for help, and no relationships with the manager or team. The finance job took the guest to Charlotte, where they danced as an NFL cheerleader at the Super Bowl, an experience they would not trade for the millionaire-by-30 path a return offer might have brought. After failed interviews about two years out of college and rejections for five to seven data science roles a year later, the guest took an Amazon business analyst job at a 40% pay cut, which they say paid off in the long run.
Knowing your metric. A co-host frames this as knowing the metric you are optimizing, whether compensation, entry to a field or management. The guest agrees priorities shift every few years: tech at all costs then, compensation first four or five years later, once in tech with bargaining power.
Building a promotion case. At Amazon the guest found the job was mostly writing simple SQL pulls, pitched a self-serve tool to the director, and built it with an AWS database, pipelines and an Excel/VBA front end using free AWS credits. That became the evidence in a push to BIE, backed by a supportive manager; later docs arguing for headcount turned the side project into a team of five BIEs. Talking to product managers filled the product gap, and Meta made an offer on the third attempt. The hosts call this being "noisy"; the host describes brag documents at Stripe, the guest says Meta's performance reviews worked the same way, and the host says his staff promotion depended on people beyond his manager knowing him.
Confidence and superpowers. The host recalls staying quiet in consulting meetings and credits managers who amplified his work. A co-host describes feeling out of place among PhDs until owning executive storytelling as a superpower. The guest argues creativity matters more as standard problems are automated. Confidence lessons follow: a mentor's "confidence of a man", a hedged statistic that women apply at 90% of qualifications and men at 50%, and "no one here knows what they're doing."
Work-life balance. The host says long hours at Nextdoor doubled his salary, while consulting gave top reviews but no promotion or raise. A co-host argues balance matters only when work is not for yourself, and the guest says the best managers were the hardest critics. Their top lessons were raising the return on time: working on important problems, productionizing pipelines, getting feedback early and learning to say no.
Going solo. The guest left a full-time role to test a hypothesis that AI makes one-person businesses easier to start and run, supported by family, a partner and a friend who run businesses, and some savings; they caution the story may differ a year on.
How to get hired. As a coach, the guest ranks communication first, then SQL beyond advanced practice problems, then a short list of statistics applied well, then mock interviews, with large gains over about 10 sessions. Analysts should transfer internally, use referrals and keep resumes simple with quantified impact. Outsiders should master Excel or SQL before Python, build Kaggle projects from a list of 10 industries, volunteer on real data, and persist: the guest cites people applying to 6,000 jobs for 11 interviews and says their own route took four years.
- Imposter syndrome, not skill, cost the guest a return offer. The guest, a LinkedIn data content creator, says that in a junior-year internship at Facebook they barely spoke in meetings, did not ask for help (waiting on queries that ran two and a half hours), and built no relationships beyond other interns. They now repeat a mentor's line to walk in "with the confidence of a man", citing, as stats they say they may misstate, that women apply at 90% of a job's qualifications and men at 50%, and a Meta colleague's remark that "no one here knows what they're doing."
- Automate your job, then argue you already do the next one. After about 40 interviews for one finance offer that sponsored their visa, and later rejections for five to seven data science roles, the guest took an Amazon business analyst role at a 40% pay cut. They built a self-serve AWS-and-Excel/VBA data tool, pushed for promotion to business intelligence engineer by showing they did BIE work, wrote docs that grew the side project to five BIEs, and learned product from PMs; Meta made an offer on their third attempt. The host adds that his move from senior to staff depended on decision-makers outside his manager knowing him.
- Find your superpower and lead with it. A co-host says that among math and stats PhDs, their confidence came from owning business storytelling to executives, then working on weaknesses. The guest argues creativity is underrated in data science because problems with standard solutions are already solved or automated, and credits a designer-quality deck at Instagram with changing how they present.
- Long hours pay off only where you learn and are well managed. The host says grinding at Nextdoor doubled his salary and sped promotions, while a consulting job rewarded five-out-of-five reviews with no promotion or raise. A co-host argues work-life balance only becomes the question when work is not for yourself; the guest advises choosing a manager who will push you, saying the best were the hardest critics.
- Interviews turn on communication, hard SQL, statistics and mock practice. The guest, now a career coach, works on communication with 100% of clients, says at least 50% are worse at SQL than they think (practise complex joins, query optimization and window functions), advises a short list of statistics known well, and reports large gains over about 10 mock interviews. For analysts, internal transfers, referrals and a simple resume with quantified impact; for outsiders, master Excel or SQL before Python, build Kaggle projects in fields you care about, then volunteer on real data, a path the co-host frames as at least a year's commitment.
AI Analyst LabHow a Principal Data Scientist Discusses uses an Open-Minded in Your Early Data Career
- Don't fix early on one role at one company. The speaker tells aspiring data scientists set on being an applied data scientist at a specific company that they might not like it, or data science at all, and argues that because data skills transfer well you can change path without being pigeonholed.
Scott Hanselman.NET Aspire with Anthony Simmon
synopsis
Simmon, a staff software developer at Workleap in Montreal, ends up arguing that .NET Aspire's value lies in local development: cutting onboarding to minutes and absorbing networking chores. Because it is a framework you extend in C#, he has made custom HTTPS domains work across host and containers in a few lines, which he hopes Aspire adopts.
Blogging. Hanselman praises him for still writing long-form posts. Simmon says explaining Aspire to others taught him it; he writes in English with help from translation and grammar tools.
Why Aspire clicked. He was improving developer experience at Workleap and about to write an internal CLI when Aspire was announced at .NET Conf. The Glenn Condren and David Fowler talk listed the pain points he had already collected.
Onboarding. He says onboarding is the problem people relate to most. The eShop reference app ran fully after F5 in minutes, where it would normally take days of docs and diagrams. Aspire also takes over ports and service references.
Custom domains and certificates. Workleap uses .localhost domains with mkcert certificates. His extension automates hosts-file entries and certificates, then mounts the certificate authority into containers so they trust it. Someone on a GitHub issue adapted his code to a working solution. He is weighing a public repo or a NuGet package; Hanselman thinks short snippets beat a library people must maintain.
Production. He has not used Aspire's deployment side. For proxied traffic he points to forwarded headers and, where required, internal certificates for in-cluster traffic.
- Aspire's strongest case is cutting developer onboarding time. Simmon says a new hire can spend two weeks getting a project running; with Microsoft's eShop reference app (multiple .NET services, Blazor UI, Python, language models) he cloned the repo, found the Aspire app host, hit F5 and had every service running with AI-generated seed data within minutes.
- Aspire removes the networking chores developers dislike. Declaring services lets Aspire manage ports and inject references between them through environment variables or secrets. He says this ends cross-team port conflicts, such as every team defaulting its front end to port 3000.
- Local custom domains should use the reserved ".localhost" domain. At Workleap, with dozens of services, they use custom local domains because heavy localhost use bloats cookies. His Aspire extension edits the hosts file, creates matching TLS certificates with mkcert, and shows the URLs in the dashboard; Hanselman notes the IETF reserves .localhost so no one can buy it.
- Containers need extra wiring to trust local domains and certificates. Containers have their own hosts file and certificate store, so his extension bind-mounts the exported mkcert certificate authority into the distro-specific locations and registers the domains. He says it is a few lines of C#, which he sees as Aspire's main value as a framework rather than a tool, and hopes it lands in a future Aspire version.
- Cloud deployment needs different certificates or forwarded headers. Simmon has not used Aspire's deployment side and says self-signed certificates are not for the cloud; behind reverse proxies, forwarded headers let an app act as if served over HTTPS, and full end-to-end encryption depends on company requirements. Hanselman recalls his banking teams giving up and leaving traffic between containers inside the firewall unencrypted.
Crosstalk SolutionsSecure Your Network with a Raspberry Pi - Fing Agent Makes it Easy!
- Fing Agent installs on a Raspberry Pi through snap. Flash Raspberry Pi OS with the imager (Wi-Fi details preset), SSH in, update, install snapd, then run
sudo snap install fing-agentand its connect script under /snap. The host says it works on Pi 4 and Pi 5; the app's "add monitoring unit" then finds the agent automatically.
Technology and FriendsVenkat Subramaniam on The Evolution of Java [EPISODE 832]
- Java has outlasted the usual decade of popularity through its community. Subramaniam says most languages stay popular for about 10 years, and calls Java the ecosystem (JVM, JRE, language, tools) and the people more than the language; its 30th year is "next year" at recording. Java began as a way to run applets in the new browsers, the applet died out, and it pivoted to become a back-end technology, which he says it mostly remains, including cloud native.
- Decoupling release from development made Java agile again. He says he criticised Java heavily around 2010, receiving emails saying "I hate you because you hate Java", and saw Java 5 to Java 8 as its stagnant period until Java 8 brought functional programming. Java now ships every March and September, is at version 23 with 24 due in March 2025, and no longer promises features in advance, which he says used to make releases years late. Features ship in the next release once ready, after two to 12 years of work: virtual threads took 12, type inference five.
- Java is a language of implementation, not innovation. With more than 10 million programmers and billions of devices, and banks that equate Java with performance and stability, he argues it "cannot say oops", so it adopts ideas proven in smaller languages. He says it asks three questions of each: is it feasible in the JVM, is it useful to day-to-day programmers, and what will it make harder to do in the future.
- Virtual threads make asynchronous code look synchronous. He argues callbacks avoid blocking but create "callback hell" and accidental complexity, especially in exception handling. Virtual threads are separate from the underlying carrier threads, like riders and horses: an application can create millions of virtual threads, and one blocked on I/O releases its carrier thread for other work, so code is written as before but scales far better.
- Kotlin is a "cocktail" of many languages. He says its designers learned from C#, Java, Python, Ruby and others, which is why he sees no surprise in Google embracing it. He calls it multi-platform, running on the JRE and compiling to WebAssembly and JavaScript, and multi-paradigm, supporting imperative, object-oriented and functional styles. He agrees with the host that Android has moved somewhat towards Kotlin.
Crosstalk SolutionsTech Websites You Should Know About #1
- Passwords-generator.net beats built-in generators on options. Beyond the usual numbers, cases and symbols, it can drop look-alike characters (O and 0, I, 1 and l), forbid repeated or sequential characters, and generate several passwords at once (five by default). The host rates it one of the best for its simplicity.
AI Analyst LabHow to frame impact for data professionals
- Data work's impact takes one of three forms. The speaker defines impact as work that moved a metric, work that led to a change of direction in the business, or a process change resulting from its output.
Polis Pandit$Elon Musk Wants You To Have More Children, But There's a Problem
- This post includes paid content that is not available publicly. See the original source.
GeoHussarSyria losing Aleppo is disastrous for them
- Losing Aleppo is nearly a death blow for Assad. It is Syria's largest city, about 2 million people, and its economic hub; its full recapture in 2016 had ended most hope of defeating Assad. The new assault worked, the host says, because every rebel faction agreed to stop fighting each other and aim at one enemy.
- Assad's backers are too overstretched to save him. The host cites Russian losses of over 700,000 troops in Ukraine and Iran's proxies being decimated by the IDF. A fall would cost Iran its training ground against Israel and cost Russia Tartus, its only Mediterranean port, on top of the humiliation of its client being pushed back.
- Damascus is exposed and outside powers may be converging on Assad. Israeli officials said the regime would pay for aiding Hezbollah, and the host expects more Israeli incursions; Turkish-backed factions joined the offensive. The host speculates, explicitly unconfirmed, that Turkey and Israel may be cooperating to sever the Russia–Iran nexus.
- Any successor to Assad is a net win, despite HTS's ideology. The host acknowledges Hayat Tahrir al-Sham's fundamentalist leanings but cites reports it has grown more pragmatic; any post-Assad regime will be unfriendly to Iran and Russia. It is win-win if it reunites Syria with Kurdish autonomy, and the host predicts a chance of regime change in Russia, Iran and Syria under Trump.
AI Analyst LabHow a Meta Sr Machine Learning Engineer Leverages Skydives #shorts #career #careeradvice
- Skydiving's perspective made every problem look solvable. The speaker, a senior machine learning engineer at Meta per the title, says that from 13,000 ft you see only mountains and rivers, and that this taught them every problem has a way through if they stay calm and work at it slowly.
Mr. GlobalClearing up myths about oil imports in the U.S.
- US crude imports track refining, not domestic need. Since 2005, when he says US consumption peaked, production has risen by 8.5 million barrels a day while imports fell only a little over 3 million barrels a day. Refiners import crude, refine it and ship products back to the source countries; imports have risen slightly over the last 18 months as refining capacity grew.
- Oil tariffs would hit refiners and invite retaliation. Rebutting a viewer who said domestic drilling means no tariffs on oil, he argues all US refiners importing crude would pay the tariffs regardless of where products go. He expects retaliatory tariffs on US oil exports to follow.
- More US production means more exports, not fewer imports. He compares refiners to an alternator maker that imports parts globally, assembles them and sells worldwide: some output serves the US, some does not.
GCastCreating a Copilot Agent with Copilot Studio [GCast 188]
- Copilot Studio builds an agent through a natural-language dialogue, without code. At copilotstudio.microsoft.com, Create offers templates (travel, finance, weather, a website) or "New agent", where Copilot Studio's own copilot asks what to make. It then asks in turn for a name, what to emphasize or avoid, a tone, and a public website as data source. Giard describes a health-and-wellness chatbot, renames it "GCast Wellness", tells it not to answer medical questions to avoid giving bad medical advice, and picks a friendly tone.
- A website data source must be two levels deep or less. Giard explains that a domain with two subfolders is accepted but a third level is not. He chooses none for now, and says the source can be changed later.
- Topics are trigger-driven workflows, some created automatically as system topics. Conversation Start shows a greeting built from the bot.name variable as soon as the agent loads. Greeting fires on phrases like "good morning" or "hi", and Giard adds "good evening" to its list. Start Over asks a yes/no question: yes resets the conversation context that otherwise lets a follow-up like "tell me more about that" work, and no carries on. In the test panel, "let's start again" triggered it although that is not one of the listed phrases. Fallback asks the user to rephrase up to three times, then hands off to an Escalate topic ("talk to a person"), which he says is not yet configured beyond a message.
- Uploaded knowledge replaces public-internet answers with your own content. By default the agent answers from the public internet. Knowledge can instead come from a public website, a SharePoint site, Dataverse or uploaded files. After Giard uploads a Word document of healthy-eating Q&A, which takes a few minutes to parse and index, the same question on eating out returns the document's answer with a link to the source.
- Publishing to a website first requires changing authentication to none. The Channels tab warns that only Teams channels are available under the default Teams authentication. Giard switches it under Settings > Security > Authentication to no authentication, essentially anonymous. He then publishes to a demo website rather than a custom HTML page; at its URL the agent answers the greeting and returns the uploaded document's answer.
Scott HanselmanEPISODE 5 - Scott and Mark Learn To... Undocumented APIs
- Mark's undocumented-API career began with reverse engineering. He studied a published disassembly of the Apple II ROM in about 1985 and added a computed GOSUB to Applesoft BASIC, which he wrote up in Compute! magazine. His Windows work grew out of his Carnegie Mellon PhD on "application transparent fault management": he tried to apply its checkpoint-and-restore idea to Windows 3.1 and 95, which took him deep into their internals.
- Regmon required reverse engineering the native API beneath Win32. Mark describes Win32 as a facade, one of the "personalities" Windows NT was designed to carry over an undocumented native API. He worked statically, with a disassembler he wrote that merged in the private symbols Microsoft accidentally shipped on an NT 4 beta CD for hardware developers. He also worked dynamically with the SoftICE kernel debugger, at whose maker, NuMega, he briefly worked.
- APIs stay undocumented so Microsoft keeps the freedom to change them. Mark says that once an important app depends on an API, it can no longer change without breaking that app. He accepted the risk because he maintained his commercial tools and simply fixed a diagnostic tool when it broke, and the core internals move slowly anyway. He says Microsoft must document any API its own products use, so there is no hidden trove of exciting APIs. Hanselman's answer was "document everything". Mark's was that internal functions, open source included, are internal for a reason.
- Compatibility shims exist to get enterprises onto new Windows. Hanselman argued that Microsoft gets too little credit for running old software, citing VisiCalc running on 32-bit Windows 10. Mark countered that the credit is the billion PCs Windows runs on: enterprises will not upgrade unless their old IT software comes along. He adds that apps lock themselves to a release in subtle ways, such as checking whether a function sits at a given address. Windows no longer has a 16-bit subsystem.
- Mark's NTCrash fuzzed Windows before "fuzzing" had a name. In 1996–97 he wrote it to barrage the NT system-call interface with garbage, inspired by the crashme tool he had used on BSD Unix at Carnegie Mellon. Microsoft fixed the faults he published. A later version that also hit the kernel-mode Win32 graphics and windowing interfaces found more. When he joined Microsoft he found NTCrash and NTCrash 2 checked into the Windows NT codebase as test tools.
Crosstalk SolutionsA Ghost Town Goes Online: 1800s Silver Mine Gets WiFi
synopsis
The host, Dave Barger and YouTuber Techno Tim spent three days wiring Cerro Gordo, an 1800s silver-mining ghost town at 9,000 ft in California. They left it with wireless coverage across town, a backup WAN, and a 10 Gbit fiber link 700 ft down a mine shaft. Ubiquiti sponsored all the network equipment, TrueCable donated the outdoor Cat6, and Infinite Cables donated the fiber.
The core moves. The plan put the core on the General Store, but it had no power. The host moved the 4U rack, 1U UPS, UniFi Dream Machine SE and a Rocket Prism with an omnidirectional antenna into the Gordon House instead, and he later judged that the better site. Internet comes from a wireless ISP in Lone Pine through a solar relay pole. The crew left that relay alone ("if it ain't broke") because a Starlink Mini now gives a second WAN.
The Bunk House and the hill. Until then, guests' Wi-Fi had come from an indoor AP hanging out of a window. The Bunk House got a NanoBeam, a U6 Mesh Pro outside and a U6 Plus inside. It was also the only place with line of sight past the tailings pile, so the crew ran 300 ft of shielded direct-burial Cat6 up the hill to a NanoBeam aimed at the Hoist House. They painted visible gear brown with Plasti Dip so it would blend in.
The mine shaft. The old coax could carry Wi-Fi down the 900 ft shaft but couldn't power cameras or more APs. The crew chose a hybrid: 244 m of armored six-strand OM3 fiber to an Enterprise 8 PoE switch at the 700 level, and pre-terminated Cat6 for the UK-Ultra APs at 100, 200 and 900 ft. A G5 Flex camera watches the water source. Brent, the owner, installed the run about three weeks later, and the switches' SFP+ ports gave a 10 Gbit link, which the host calls overkill.
The smaller sites. At the recording studio, a NanoBeam's PoE-out port powers a mesh AP over one cable. The General Store got a Switch Ultra 60W, a hidden NanoBeam and two G5 Flex cameras, running off an extension cord until permanent power goes in. Brent's house got a NanoBeam and a U6 Mesh Pro bridged with PoE injectors. Dave measured about the ISP's full 50/40 Mbps inside the tin-sided church, so they skipped it and kept the spares.
The owner's view. Brent recalls a 1–2 megabytes-per-second hotspot and driving to Lone Pine to upload videos. He names the hotel and future cabins as the next needs.
- A site without power forced the network core into another building. The plan put the UniFi Dream Machine SE and a Rocket Prism point-to-multipoint antenna on Cerro Gordo's General Store, which had no power, so the crew moved them to a locked 4U rack with a 1U UPS in the Gordon House. The host thinks this ended up better than the original plan. Ubiquiti sponsored all the network equipment.
- They kept the old solar-powered ISP relay and added Starlink as backup. The wireless ISP's signal from Lone Pine reaches a solar relay pole (NanoBeam, MikroTik router, LiteAP). They brought new NanoBeams to replace it but left it alone because it worked, and a Starlink Mini now plugs into the UDM SE as a second WAN beside the roughly 50/40 Mbps ISP link.
- A tailings pile meant relaying to the Hoist House via a hilltop NanoBeam. The Bunk House was the only spot with line of sight up the hill. A NanoBeam there picks up the Gordon House link, and 300 ft of shielded direct-burial Cat6 donated by TrueCable runs up the hill to a NanoBeam aimed at the Hoist House. They painted the outdoor gear brown with Plasti Dip, which the host says doesn't hurt Wi-Fi.
- The 900 ft mine shaft got fiber plus Cat6, replacing coax. The old coax could carry Wi-Fi down but couldn't power cameras or more APs. Infinite Cables donated 244 m of armored six-strand OM3 LC-LC fiber, giving a 10 Gbit link between Enterprise 8 PoE switches at the top and at the 700 level. Pre-terminated Cat6 feeds UK-Ultra APs at 100, 200 and 900 ft, plus a G5 Flex camera on the water source. The owner, Brent, installed the shaft run about three weeks later.
- Some sites needed far less than planned. Dave Barger measured close to the ISP's full speed inside the tin-sided church, so they skipped it and kept the spares. At the recording studio, a NanoBeam's secondary PoE-out port powers a mesh AP over one cable. The General Store got cameras running off a temporary extension cord. Brent says the town went from a 1–2 megabytes-per-second cellular hotspot to working internet, and he sees the hotel as the next need.
Polis Pandit$Give DOGE a Chance, But Be Realistic
- This post includes paid content that is not available publicly. See the original source.
Crosstalk SolutionsStop Network Intruders: Automatic Device Blocking with Fing Agent!
- Fing Agent blocks new devices by default after you approve existing ones. You first scan the network and confirm which devices belong. With automatic block turned on, any new device is kept off the network until you allow it, and Fing Agent notifies you when one tries to connect.
- Each device can be blocked outright or paused for a chosen time. A block removes the device from the network until you unblock it. A pause cuts its internet access for a period you choose or indefinitely, which the host says suits businesses that want control without constant manual work.
- The host relays a Black Friday price on Fing's Premium plan. As captioned, it costs $449.99 a year and covers up to three Fing Agents. The short presents this as a limited-time offer and says nothing either way about sponsorship.
AI Analyst LabAn Unconventional Machine Learning Career - Story of a Meta ML Engineer | Ep4
synopsis
This Data Neighbor Podcast episode interviews Karthik, a senior machine learning engineer at Meta who previously worked at Amazon and interned at Google, and argues through his story that a move into machine learning is a sequence of deliberate, adjacent steps rather than a lucky break. He ends on two claims: that his apparent confidence is imposter syndrome turned into a habit of seeking out smarter people, and that skydiving taught him every problem is solvable.
Getting in the door. Karthik's interest began around 2013 or 2014 with Google search and the PageRank algorithm, which led him to data mining and machine learning courses. From Jaipur he cold-emailed alumni and professors; he estimates 99% never answered. He interned with a professor at another Indian university to find contacts, and eventually a Minnesota alum's lab took him on, which led to a master's at the University of Minnesota in 2015. He calls himself an introvert, and credits a mentor's line that the worst case is a "no" that leaves you where you started. The hosts' reading is that what looks like luck was a 1% chance worked for relentlessly.
The detour through software engineering. A student loan, a visa clock of a few months and the dominance of PhDs in ML hiring at the time pushed him toward software engineering. He chose an Amazon pricing team next to an ML team over a better-paid Google Android offer, citing Bezos's regret-minimization framework. Once there he kept asking his manager for ML work, and later declined a fast track to senior software engineer in about four years, taking smaller ML projects instead because a senior-level switch would carry higher expectations.
Building the path. He contributed a feature the scientists lacked the bandwidth to build; it went into the next model iteration and built the team's trust in him. His route ran through ML infrastructure (an inference platform for dynamic pricing), feature engineering, ML engineering, search within Amazon, and then Meta. He says most software engineers who move to ML follow this route because infrastructure skills transfer. For someone outside tech, such as in healthcare, he stresses domain knowledge as the hardest part to earn, then coding and a network of practitioners; the hosts frame it as a series of bridges, and Karthik adds that some people have to switch companies to make a jump.
What the job is. He describes ML engineering as a spectrum from infrastructure to research, with titles that differ by company and team. As an applied engineer he owns search and search ads: diagnosing why some queries fail, then choosing between a heuristic, an optimization or a model, and evaluating through A/B tests.
Confidence and perspective. Karthik says he struggles with imposter syndrome at every new team, and would again in a field such as computer vision, and that it pushes him to find people smarter than him. He took up skydiving in 2020 and is now licensed. He says seeing the world from 13,000 ft eased his anxiety, and he recommends a tandem jump to everyone, putting the risk of death at under 0.001%.
- Cold outreach pays off only through volume and persistence. Karthik, a senior machine learning engineer at Meta (previously Amazon and Google), says that as an undergraduate in Jaipur he cold-emailed alumni about machine learning and roughly 99% never replied, until one alum's University of Minnesota lab took him on, which led to his move to the US in 2015. His advice to students is to work with an adviser on a project, because Indian advisers often have collaborators and former students in the US.
- Take the offer with a path to your goal, not the prestige. Karthik turned down a Google offer in Android development, which paid more and was at his dream company, for an Amazon pricing team that worked alongside a machine learning team, using Bezos's "regret minimization framework"; he calls it the best decision of his career. A host draws the lesson that someone pivoting should take a job next to the role they want, such as an analyst working closely with data scientists.
- Delay promotion so the pivot happens while expectations are low. Karthik says he could have reached senior software engineer within about four years but turned down bigger-scope projects for smaller machine-learning ones, reasoning that a move at senior level carries bigger expectations and that two or three years' delay is nothing in a 30-plus-year career. He adds that you earn the right to ask by being good at your current job first, then doing "5% extra".
- Career changers should jump in adjacent steps, carrying domain knowledge. His route ran from software engineer in pricing, to building an inference platform for Amazon's dynamic pricing team, to feature engineering, to machine learning engineer, to search within Amazon, then to Meta. For a healthcare worker wanting into tech he says domain knowledge is harder to earn than a CS degree; learn coding (Java, Python) and build a network of people doing ML in healthcare, which a host frames as crossing one "bridge" at a time.
- Applied machine learning starts with analysis, and may need no model. Karthik describes machine learning as a spectrum from infrastructure to pure research, with titles defined differently by every company and team. In his current work owning search and search ads relevance, he starts with data analysis of why some queries return poor results, may solve it with a heuristic or optimization rather than a model, and says everyone in the field ends up evaluating changes by A/B testing.
AI Analyst LabHow to be Concise in Writing #career #careeradvice #datascience
- Cut finished work in half, then in half again. The speaker gives this advice to their team for any email, presentation or other artifact: find a way to halve it while keeping what you are trying to say, and after doing it twice "you're probably in a pretty good spot".
AI Analyst LabHow to Measure the Impact of Your Work #career #careeradvice #datascience
- Tracking your work's impact also makes the work more exciting. The speaker says this is what is looked for both in interviews and when promoting people, and that understanding it changed their mindset about their own work.
AI Analyst LabHow to Communicate with Executives
- An executive deck must pass a child-retelling test. The speaker tells their team a deck should be clear enough that a seven-year-old shown it could explain it to another seven-year-old; if it is not at that level, an executive will not understand it.
GeoHussarTikTok Live 10/15/2024
synopsis
In this solo TikTok question-and-answer live, held shortly after the 2024 election, the host ends where the evening's questions keep returning: Trump's announced agenda is largely unworkable and, if carried out, inflationary; immigrants are an asset America depends on; and Hamas's October 7 attack has handed Israel, now unrestrained by Biden, licence to clear Gaza. Despite all this, the host maintains America remains the best country in the world for an immigrant to start from nothing.
Platform and crypto. The host opens by recounting TikTok cutting creator funding without explanation, then restoring it, and pushes viewers to Patreon and YouTube. On crypto, the host refuses to tip meme coins or a Bitcoin top, discloses an average Bitcoin purchase price of about $24,000–25,000 with nothing yet sold near $90,000, and says the US government has no need to buy Bitcoin when it already holds tens of billions in seized coins.
The Electoral College. Following a response video to a large creator, the host defends the Electoral College as protection against mob rule and foreign influence campaigns, rebuts a viewer's math with the ten-largest-states simulation, and notes states' partisanship changes across generations.
Personal finance and housing. Asked how to reach financial security, the host describes saving four to five months' rent by walking to work and eating the cheapest food, then automatically investing paychecks in appreciating assets. On housing, the host speculates that boomers downsizing could produce a glut, and says private equity owns perhaps 1% of homes.
Trump's plans. Recounting a podcast appearance that day, the host calls the plans unrealistic and hopes they are posturing, since even moneyed interests would lose if all were enacted. The host details the missile-shield supply chain and the farm-labour math of deportation, and dismisses talk of invading Venezuela, whose oil is plentiful but poor quality.
Immigration and America. In a long exchange, the host argues immigrants work harder, commit less crime and educate their children, and that childless Americans are worth less to the government than immigrant families. Challenged by viewers, the host concedes US flaws — about 2 million people imprisoned, private prisons, an insurance-lobby-driven health system, widening inequality — but blames voters who turn out only for presidential races.
Gaza. The host argues pro-Palestinian protests that celebrated Hamas pushed wealthy Jewish donors the host knows to Trump, and asks whether 40,000 dead, possibly 100,000 or more, is "worth it". Pressed by viewers, the host traces the conflict back to the 1947 partition, calls Netanyahu a butcher who funded Hamas, labels Israel's treatment of non-Jews apartheid, and insists on a two-state solution. The host calls Tulsi Gabbard's nomination a joke.
Closing questions. The host sees Russia–Ukraine as a stalemate, says ending it will not lower prices if tariffs arrive, backs nuclear power as the cleanest energy, and lists China's compounding crises.
- The Electoral College guards against an easily swayed popular vote. The host argues foreign disinformation targets the masses, who once accepted slavery and denied women the vote. By the host's own simulation, a 51–49 split in the ten largest states would outweigh the other 40; state leanings shift too, as New York, Texas and California all voted for Reagan in 1984.
- Trump's all-American missile shield and mass deportations are unrealistic. The host says such a shield would need TSMC chips, from a lead of a decade or two, Dutch ASML lithography, and rare earths mostly from China. Deporting 10 million undocumented workers, including about 1 million of 2.5 million farm workers, would leave crops rotting and raise food prices; "big money" is already pricing in inflation.
- Undocumented immigrants are a net gain the government quietly values. The host argues they pay state and local taxes, spend locally, are 1.6 times likelier to start businesses, and have more children while the US birth rate is 1.67. If elites truly wanted them gone, the host says, they would fine employers a million dollars a day rather than close the border.
- Hamas's attack gave Israel carte blanche, the host argues. The host blames Hamas's October 7 assault for the deaths that followed and predicts that under Trump, Gaza may be cleared of Palestinians within four years. The host also calls Israel's treatment of non-Jews apartheid, saying West Bank Palestinians get water below the UN minimum of about 100 litres a day while settlements have pools.
- China's stimulus is too small for its stacked crises. The host cites $23 trillion of local-government debt against an $18 trillion economy, record youth unemployment, deflation, a shortage of women from the one-child policy, and little immigration. The host predicts India will eventually reach China's level, perhaps in 30–40 years.
Mr. GlobalCan Trump Re-Open The Keystone XL?
- Keystone XL cannot be "reopened"; it never operated. The original Keystone has run for 14 years, while the XL was 8% complete when Biden cancelled it and the pipe that was laid has been dug up. He expects TC Energy would need a new permit, then face lawsuits and perhaps 10 years in court.
- No pipeline opening has ever lowered prices, he argues. He says gas rose over a dollar a gallon in the four years after Keystone opened and rose significantly after the Alaska pipeline opened. He calls the lower-prices promise a line oil companies give politicians to win approval.
- Trump, not Biden, was the one begging the Saudis. Biden asked Saudi Arabia only to undo the OPEC 2020 deal, while Trump's 2018 pressure crashed the market, drove up US oil and gas bankruptcies and cost 30% of drilling capacity. The OPEC 2020 deal cut production by 10 million barrels a day, and when the Saudis refused Biden, he used the SPR.
- Fracking was not limited under Biden. He cites a 2023 record for Permian wells drilled and completed, and over 2,000 wells on federal land in the last two years against a level Trump never reached. He offers the commenter a bet of any size that Trump's gas-price plan is to ask Saudi Arabia for more output.
AI Analyst LabHow To Provide Value
- A student can offer helpers the good feeling of helping. The speaker says many people want to help their community and feel good when they do, so a reasonable message lets them feel that; the trick is twofold — find the people who feel that way ("that's what the two hours is for"), since not everyone does, and craft a message that draws the feeling out.