tech

2024-04-04 · open the original ↗

• A zoomed-in UniFi Protect live view pans as you tilt the phone. In landscape, after zooming into any camera's live view, turning the phone left, right, up or down shows more of that side of the image. In portrait mode, sliding down enlarges the image and allows the same left-right movement. The host says he learned of it only the day before recording.

2024-04-04 · open the original ↗

• A custom extraction model reads fields from same-format documents. Giard says Document Intelligence Studio, at documentintelligence.ai.azure.com, is the rebranded Forms Recognizer with new features added. It offers prebuilt models for invoices, receipts, health insurance cards and tax forms. He uses a custom extraction model instead, for a set of five PDF invoices that share one layout but carry different text. He says it also accepts JPEGs, Office documents and images.
• Creating a project provisions the service and its storage. You don't have to set anything up in the Azure portal first. The project wizard asks for a subscription and resource group, a service name, a region (he picks East US) and a pricing tier. He says the free tier limits calls per minute and document size, and calls standard "pretty cheap" for a small sample. The API version defaults to the most recent. The wizard then creates an Azure storage account (he takes the cheapest tier) and a blob container for the documents. He expected about 60 seconds and says it took about a minute and a half.
• Label one document by hand, skipping auto-label. After the upload he picks "run layout", which skips auto-labelling, because auto-label needs an existing model that matches the form. He clicks words to highlight a span, then names it as a field. The type options are field (static text), selection mark (a true/false checkbox) or signature (handwriting). A table icon detected the column names on its own. He makes the table dynamic rather than fixed so it can handle invoices with more than one row. He labels only one of the five documents.
• Neural training fits this case; template suits handwriting. Giard says the template option needs more than one labelled document, takes longer and is better for handwritten notes. He trains a neural model on the single labelled invoice and says labelling more documents would probably raise accuracy. Training on five documents took about 15–17 minutes.
• The test pulled the right fields, with uneven confidence. On a test document the model found the invoice number and date correctly, with about 96.8% confidence on the vendor address and 98% on the customer address. Other fields came in as low as 36% and 29.5%. Giard says real use would call the model from an application through its API. The studio shows the endpoint, API version, service resource and container, and he says he will cover the API calls in a future video.

2024-03-28 · open the original ↗

• Seniority is organisational game theory, not better code. Obasanjo, a lead product manager at Meta and formerly at Microsoft, says success as scope grows depends on treating other teams, companies and customers as parties optimising for their own needs; product management is "organizational programming". His example: once streamers were known to be unprofitable and writers and actors wanted more pay, he reasoned that studios would pull back content like NBC keeping The Office for Peacock.
• Tech feels disempowering because the ecosystem has matured into gatekeepers. Twenty years ago no permission was needed; now an app answers to Apple or Google and a website to Safari and Chrome, and the other route is VC money. He adds that big tech has stopped launching new services every weekend, and that a breakout like Figma is now rare.
• At web scale, social media reduces to content moderation. His exercise: a murder rate of a few per 100,000 means a million-user site hosts dozens of murderers, and bullies and spammers scale the same way. Against Hanselman's hope for self-hosted sites and the fediverse, he argues ordinary users won't run servers or delete spam, so moderation recentralises on whoever does it best. Hanselman adds that free long-tail services rot because their running costs come with no revenue, citing a Blogger backend change that broke Open Live Writer.
• Advertising is a progressive subsidy the industry spoiled. He calls ads the best business model yet devised, since heavy spenders subsidise a student in Nigeria using the same free service, unlike subscriptions, which take the same few dollars from everyone. But he says the industry, his employer included, failed to self-regulate and got creepy without being useful, chasing people with mattress ads after they had already bought one.
• "Positive vibes over business value" was a zero-interest-rate phenomenon. He argues cheap money pushed everyone into risk, so no one asked whether work made money, and it funded products like Snapchat's selfie drone (Hanselman adds the Rabbit AI device). Google's and Facebook's success fed the myth that growth finds a business model eventually; he says it is now clear a giant company can simply lose money, and "the free lunch is over."

2024-03-25 · open the original ↗

• An Azure OpenAI resource is created from the portal's Create a resource. Giard searches for "OpenAI", then sets a resource group, a region (East US for him, chosen for proximity) and a globally unique name. Standard S0 was the only pricing tier offered. Network restrictions and tags are optional, and the deployment took about 30 seconds.
• Models are deployed and tested in Azure OpenAI Studio, outside the portal. Its Models page lists the available OpenAI models and their dated versions, and which ones can be deployed depends on your region. Giard deploys GPT-3.5 Turbo and raises the tokens-per-minute limit to the maximum so a demo won't run out. He thinks the default was about 4,000.
• The chat playground keeps conversational context across prompts. By default it includes the past 10 messages. Giard asks for a 10-slide outline on Azure OpenAI, then asks to "expand slide 3" without saying what it was about, and it wrote about key features. "Shorten this to one paragraph" acted on the previous answer. He says GPT-3.5's knowledge stops at 2021, so it would not know about the NCAA tournament under way when he recorded, and he advises checking answers before deploying a bot publicly.
• The DALL-E playground generates images from a text description. Opening it deploys a DALL-E 3 model (DALL-E 2 was also offered), which Giard says can take about five minutes. A prompt for a polar bear eating ice cream in a top hat produced an image. The prompt can be regenerated as-is, or extended with a style ("in the style of Van Gogh") to restyle it.

2024-03-21 · open the original ↗

• Traditional embedded tooling makes IoT an economics problem. Kattenhorn, a CEO whose company does bespoke .NET work, describes taking over a client's PIC-based firmware for pay-over-time solar systems in the developing world. The team found C code overwriting memory and had to write their own bootloader and over-the-air updater, and the first version bricked 60,000 units that had to go back to service centres. Firmware labour was a critical part of unit price.
• Meadow lets ordinary .NET developers build firmware on the desktop. He says engineers with no embedded experience had the app's workflow and state management running on a PC in about two days, with sensors stubbed out to logs, then had it on a Meadow F7 a couple of days later. Dependency injection swapped the current sensor between the Project Lab V3 prototype boards (about 15 sent into the field) and the production CCM board. Hanselman disclosed he has personally invested in Wilderness Labs, Meadow's maker, and Kattenhorn's company backed its 2016 Kickstarter.
• Kenyan cooker monitors pair cost visibility with carbon-offset aggregation. The devices measure current, show households what a cooking event cost to ease fear of electricity over charcoal, and send data by GSM over MQTT to an IoT hub. The plan is to batch the small offsets into tonnes for sale and return the money to households. Anomaly detection quarantines odd readings.
• Treat a device fleet as a distributed system. He frames each device as a service with its own state, like digital-twin thinking. When a cell tower comes back, the reconnect rush needs exponential backoff, and devices queue data on one thread while another checks connectivity. Time is a hidden problem: TLS needs a correct clock, and their early devices were refused by the IoT hub until they fixed clock drift.
• Reliable over-the-air updates change how a business ships. He argues that unreliable updates freeze a company into waterfall, because nobody wants the Saturday call about thousands of bricked devices. Phased deployment through Meadow Cloud makes shipping fixes and features less risky. He sees a market in small HVAC and floor-heating makers who now need connected devices and feel excluded from IoT.

2024-03-20 · open the original ↗

• A third-party mount makes a standard Starlink dish usable in motion. The Trio flat mount takes a Gen 2 or Gen 3 dish; the host's Gen 3 standard version with suction cups costs about $475, and adhesive-pad, magnet and bolt options are cheaper. Trio sent the mount free, but the host says he was otherwise uncompensated and has no referral code.
• Dish plus mount undercuts Starlink's own in-motion dish by about $1,500. Starlink's high-performance flat dish, which is the one it sells for in-motion use, is $2,500, against roughly $600 for a standard Gen 3 dish plus the Trio mount.
• The suction-cup mount held through rough roads and a long drive. It took about 10 minutes and an Allen wrench to frame the dish, about 20 pumps per side to fix it to a glass sunroof, and it survived an 8-mile dirt road with an 8,000 ft descent. Power came from the vehicle's rear AC outlet into the Starlink adapter and router.
• Service held unbroken across a 700-mile trip through cellular dead zones. The host reports never losing internet over several days, with car orientation making no apparent difference and speed tests no different from stationary use; one in-motion test with zero cell service measured 278 Mbps down by about 18 Mbps up.

2024-03-19 · open the original ↗

• A firm's accumulated knowledge becomes queryable when put before a model. Nicholas, who leads the Cloud Solutions Group at the consultancy Trace3, gives his own firm as the example: about 1,400 professionals working roughly 48 weeks of 40 hours, around 2.5 million hours a year over two decades plus. Coupling that body of deliverables with a model through retrieval augmented generation would let a new or experienced consultant ask, say, for a summary of the last six healthcare strategy engagements, their outcomes, roadblocks and financials, instead of taking months to learn the firm.
• Enterprises must adopt several kinds of AI, so governance comes first. Nicholas lists personal assistants (Microsoft's Copilot), repeatable custom copilots published internally (Copilot Studio), RAG over enterprise data, and full agents such as Devin, and says AI governance is missing or immature at most organizations Trace3 talks to. He defines it as a published AI stance and risk tolerance, chosen control frameworks translated into controls, guardrails you are informed by rather than hit, and a federated committee drawn from finance, people, business and IT rather than IT alone.
• Many proposed use cases are matched to the wrong type of AI. In one four-hour ideation session a client produced over 300 use cases for AI assistants, and over half were found not suited to them. His recruiting story shows one problem split three ways: Copilot screening 300 emailed applications against five criteria, which he says gave over 90% efficiency on the recruiter's time in early testing with human spot checks for bias; a Copilot Studio version other recruiters can reuse; and RAG over five years of past applicants to find, for example, a ColdFusion expert.
• Sensitive data management is the foundation for successful AI. Nicholas says knowing where sensitive data is, who can access it, how long to keep it and what controls apply is the precondition, alongside data readiness: curating data into a form where models can be grounded properly. He adds that within RAG, Trace3 is exploring both vector-database and knowledge-graph architectures.
• Start small in all three areas at once, not one big goal. He recommends an MVP approach to security, using the OWASP top 10 for AI as a starting point rather than first interpreting the full NIST AI RMF, plus one small use case each for assistants, custom copilots and RAG, run through a pilot group of trusted, early-promise staff. His prediction is that enterprises will need all of these, and he warns against spending a year or two on controls before touching AI; Giard likens the approach to agile's iterate-and-fail-fast.

2024-03-14 · open the original ↗

• The big-ego model of leadership is archaic. Clarke, an executive coach, former executive recruiter and former Google recruiting leader, says women entered workplaces built on a dominant, "charge the hill" male energy and were expected to copy it. She argues for balancing it with feminine energy: leaders need some ego, but narcissism serves no one, and self-awareness and empathy are the new orientation. She agrees with Hanselman that the C-suite suffers from "the perils of homogeneity".
• Replace the personal brand with a personal narrative. She distrusts social-media branding as people presenting themselves as what they are not. Her answer to "tell me about yourself" is what is not on the résumé: how you think, your style, what moves you, your aspirations. She says she can spot someone uncomfortable in their skin immediately.
• Withholding critical feedback does people a disservice. She says leaders avoid feedback because they recall it hurting them, but a leader owes it to people. She calls herself still "brutally honest", but says she has learned to soften the tone. Hanselman's example is his own boss, who re-stated what he had wanted rather than saying the team had failed; she credits that boss with leaving room for having been unclear.
• Conscious leadership means self-reflection rather than autopilot. She traces her deliberateness to a spiritual and meditation practice that began after her father died and helped her stay present as a single mother at a demanding firm. Her prescription is noticing whether you are making things better, reading people's faces, breathing, then forgiving yourself and adjusting. Hanselman adds his friend's "Monday vision, Friday reflection", which he makes "Friday forgiveness".
• Layoffs delivered without empathy are a choice, not a necessity. Asked whether leaders must be emotionally detached for hard decisions, she says no; they choose to turn awareness off, at their own peril. She calls tech layoffs by email "inexcusable" and says bad news takes the courage to face the person.

2024-03-13 · open the original ↗

• Platform engineering fixes the bottlenecks DevOps itself created. Burns, a senior staff platform engineer at Grubhub, says DevOps rightly made dev teams own their code in production, but in doing so required every team to hold its own expertise in CI and CD pipelines and deployment. A platform team instead builds self-service tools with those problems already solved, while each team still owns its process; he describes it as a practice of about the last 10 years that emerges where companies run microservices with a dozen or more teams.
• Offer a paved "golden path" but allow off-roading. At Grubhub, Burns says, the platform team supplies pre-built Jenkins templates (Jenkins DSL, which is Groovy, behind the team's own YAML layer) and standard Gradle plugins, with parameters teams can tune and room to extend for services that need something custom. About 170 microservices run on the platform.
• The payoff only appears at scale. Burns says a handful of services or teams gets little benefit, but cutting each team's build-and-deploy toil by even an hour a week across 20 or 30 teams saves almost a full-time developer. Centralizing the under-the-hood expertise in the platform team also lowers the learning curve of adopting DevOps.
• Start by settling the microservices platform, then extract one problem at a time. Burns says first decide the standard way to deploy (today often Kubernetes and Docker containers, which did not exist for Grubhub 10 years ago), build, run CI, and handle metrics, logs and service discovery; implement it for a service or two, then pull out the common code. He points to Netflix's open-sourced tooling (Spinnaker; runtime libraries that fed into Spring Cloud), the Netflix tech blog, Gradle's "developer productivity" material, and a public platform-engineering Slack run by a consulting company he could not name.
• Kotlin linting needed a Gradle plugin, which Burns now maintains. He says KTLint, a Kotlin linter owned by Pinterest, exists because JetBrains' own linter lives inside IntelliJ and is not built for command-line or CI use; the Gradle plugin adds caching and tool download. He took it over about a year before recording because its original author was busy with open-source security work, and has also contributed to Spring, Micrometer and Google's Jib.

2024-03-13 · open the original ↗

• Copilot in a Teams meeting needs only transcription turned on. Giard says recording is not required: "record and transcribe" or transcribe alone is enough, and the transcript then appears alongside the meeting.
• Mid-meeting, Copilot can catch up a late joiner. He types "I just joined. What did I miss?" during a live meeting, pitching it as a way to catch up after arriving 15 minutes late without asking anyone or reading the transcript; Copilot returns a summary of the meeting so far.
• After a meeting, Copilot summarizes the recording with timestamped links. On a recorded mock conference-planning call with two colleagues, he clicks the suggested "summarize the video" prompt beside the recording; each summary point links to its spot in the recording, and clicking one jumps playback to the 6-minute 10-second mark.
• Copilot can narrow the summary to a single topic. Asking "what did we decide about food?" returns only that discussion, again with a link to where it occurs (stated as 3:29, then 3:20).
• Copilot can answer questions the meeting did not explicitly address. Giard offers "what are the next steps?" as an example (while noting the phrase was in fact used in this meeting); it returned action items grouped by person, plus a team meeting the following Tuesday, each linked to its spot in the recording. He calls it a powerful productivity tool.

2024-03-12 · open the original ↗

• Copilot Studio builds chat apps with little or no code. Giard describes it as a low-code tool for publishing a chatbot that answers in natural language, runs workflows (follow-up questions, conditional logic, multiple-choice answers), and answers questions from your own websites or documents. It was in preview as of March 2024. It is reached at aka.ms/copilotstudio, and "Try Free" walks you through creating an account.
• A new copilot takes a name, a language and a website to index. Five languages were offered. Giard points it at his own blog, which he says holds hundreds or thousands of articles from almost 20 years. Creating it generated seven topics, from a simple greeting (replying to "hi" or "hello") to a sample ordering workflow.
• The sample ordering topic shows branching on the user's answers. It asks for the shipping state. California, Washington or Oregon go straight to product selection, and any other state gets a yes/no question about additional shipping charges, $27.50 in the test. "No" ends the conversation. Product choice is multiple-choice, and the flow ends by asking for a rating.
• Questions that match no topic fall through to generative answers. A system topic searches the given data sources, up to four public websites, and shows an answer if it finds a decent one. Asked about data visualization, the test bot summarized Giard's own article series on it (one part dated 2011), with links to the articles. He suggests pointing it at your own site to answer questions about products, HR policy or success stories.
• The test pane's tracking shows which topic and step is running. With tracking on, Giard's "hi" visibly triggered the greeting topic. "Purchase item" triggered the ordering topic even though it was not one of that topic's listed phrases, which Giard calls smart enough to understand.

2024-03-11 · open the original ↗

• Draft with Copilot writes and then refines a Word document. It opens from a toolbar icon or Alt-I, or sometimes pops up on its own in a new document. Giard asks for a three-paragraph article on AI tools in Azure. He could keep the result, regenerate it or delete it. Instead he follows up with "expand this article to six paragraphs" before keeping it.
• Copilot in PowerPoint can build a deck from a Word file's link. Saving the document put it in OneDrive, so File > Share > Copy link gave a link. Copilot in PowerPoint offers to create a presentation on a topic, add a slide, or create a presentation from a file. Giard pastes the link into "create a presentation from a file", and it produced a slide deck with imagery based on the article.

2024-03-09 · open the original ↗

• Open a file in your preferred style and GitHub Copilot copies it. In a C# sample from an open-source "Copilot Playground" repository, comment prompts in one class first produced a bare check that threw an exception on any non-200 status and returned the content string. With a second class open that formats and logs an error message (ID and status code) to the console before throwing, and logs success before returning, the same prompts produced code following that logging pattern.
• More explicit prompts give better Copilot suggestions. Giard replaces "handle the response" with "handle the case where the response is not HTTP OK", saying Copilot does better the more explicit the prompt, then accepts the suggestions with Tab.
• An open unit test steers Copilot's code toward the expected behavior. With a test open that expects getOrder(-1) to throw "Negative values are not allowed", the comment "handle negative ID value" produced a check at the top of the method throwing an ArgumentException with that exact message, so the test passes. Giard calls this useful for test-driven development.

2024-03-07 · open the original ↗

• Gamers already hold transferable skills for tech careers. Jackson, a penetration tester, Air Force veteran and founder of Alpha Esports and Technology in Suwanee, Georgia, says he avoided tech because he thought he wasn't smart enough. He now tells young gamers their tinkering with computers and consoles is a foundation for help desk, tech support or system-building jobs. The idea came in 2019, when his son's Fortnite programme reminded him of security-awareness work.
• The centre teaches the esports ecosystem, not just play. He says esports earns more than the film and music industries combined, and the centre covers production, video editing, shoutcasting, coaching and management, for which scholarships exist. A stage is used for strategy talks and presentations. Ethical-hacking and API-security courses are his own, Unreal content comes through a partnership with Epic Games, and a game-development course is planned.
• Respect is the first rule, enforced like martial arts. Matches open and close with the same show of respect, there is zero tolerance for toxic behaviour, and sessions begin with 30–45 minutes of strategy plus wrist, finger and eye stretches. He tells players that lashing out makes them worse, and his rule is "it's not how you mess up, it's how you recover", with an apology needing action behind it. He sells parents on it as equivalent to organised traditional sports.
• Online safety is taught as practical cybersecurity. Players are told to block, leave lobbies rather than engage (to avoid escalation such as swatting), limit personal details and use multi-factor authentication. He warns them off currency scams in FIFA and NBA 2K. On shared machines they untick "remember me" and clear Battle.net settings, and because management software is expensive he wrote a boot-time batch script that wipes temp files and sign-in data.
• Girls Game Too creates a safe space before co-ed play. Launched in September, the Monday programme addresses girls' discomfort with confrontation and harassment on headsets, and with self-doubt. He says progress has led to the centre's first co-ed team, launching within weeks with two girls. He reports that expanding is hard because many leagues, often franchises around Atlanta, "just let them play".

2024-03-07 · open the original ↗

• AI hype is running far ahead of planning. Dojutrek, a tech coach, says "the most dangerous thing a CIO can do is actually read an article," and that the push is consumer-driven, trickling up from household use of generative AI to product owners. Kramer, a consultant, likens it to cloud 10 years earlier and expects adoption to take years; neither sees customers going heavy into AI yet, only experimenting and "learning what we want to do."
• Security and legal exposure come first. Dojutrek names security as the number-one concern, including bad actors poisoning the large pools of data models draw on. Kramer adds court rulings that AI-generated output is not copyrightable and royalty claims over training data, so a healthcare client such as Cigna might refuse AI coding help over where its code goes and its lack of legal recourse if the output is stolen; his firm pitches on-prem models trained on a company's own material, with some headway.
• Measure whether AI makes people more effective, not just more productive. Dojutrek says Copilot let him, after about 30 years away from coding, write code much faster but not better: "having a faster car doesn't make you a better driver," and over-reliance could make coders lazy. Giard, who says he works for the company behind GitHub Copilot, adds that it gets you perhaps 80% of the way and the developer must review the rest; Dojutrek urges defining the outcome wanted and working backwards.
• Begin with quick wins on mundane work. Dojutrek suggests writing test cases as you go, getting entry-level hires up to speed faster, and repetitive code (a month's logic replicated for all 12 months in one line with Copilot). He thinks retail benefits most because it handles less sensitive data than healthcare.
• AI will change tech jobs rather than eliminate them. Kramer expects tech jobs to keep growing long term and blames the current slowdown on high interest rates and lower consumer spending, not AI, noting he knows of no 200–300-person projects rewriting systems with AI. Dojutrek says companies are waiting for someone else to go first, and predicts value shifting from tool skills to prompt engineering ("garbage in, garbage out") and engineering principles such as continuous quality, holistic stewardship and continuous learning, with college curricula needing to follow.

2024-03-06 · open the original ↗

• GitHub Copilot runs in VS Code as an extension. It needs a GitHub Copilot subscription and the extension published by github.com. Suggestions appear greyed out and Tab accepts them. In a PowerShell file, typing $jan = 1 and $feb = 2 was enough for it to suggest the remaining months. Giard picked PowerShell because he does not know it well.
• Comments describe what you want, and Copilot writes the code. "Create a hash table with months and their corresponding numbers" produced the table. "Log into Azure" offered both az login and Connect-AzAccount, then went on to suggest creating a resource group, a VM, opening port 80 and getting the current subscription.
• Copilot fills in assumptions unless the prompt states them. Asked for a resource group, it made up a name and picked East US. Naming the group and asking for West US made it use those values. Giard's caution: it is not always accurate, "you're still the pilot", so double-check its output.
• The inline chat dialog rewrites code to a request. Giard asks it to create the resource group only if it does not already exist, which is safer, since creating it fails if the name is taken. The result can be accepted, discarded or regenerated.
• Copilot Chat's slash commands generate tests, explanations, fixes and docs. With the code highlighted, /tests produced Pester tests that mock Azure. Copilot warned they were basic, and Giard calls them a good start. Right-click

  Copilot offers "explain this", which he finds useful for inherited code or a readme, plus "fix this" (it caught a misspelled az group create) and "generate docs". The same options are under / in the chat window.

2024-03-05 · open the original ↗

• The host built a free filtering tool for choosing UniFi hardware. The UI Product Selector, at uinotify.net, which the host says he created, filters UniFi access points, switches and gateways by features such as Wi-Fi 6 versus Wi-Fi 7, PoE or PoE+, and built-in UniFi Network. Devices go into a list with a shareable link, and a free account saves lists for later.

2024-03-05 · open the original ↗

• The Cloud Gateway Ultra should be the default single-site gateway. The host rates the UCG-Ultra ($129 US MSRP) right for homes and small businesses on internet up to 1 Gbps, noting most customers he sees run 300–600 Mbps. It has a 2.5 Gbps WAN, four 1 Gbps LAN ports (one usable as a second WAN for failover/load balancing) and onboard UniFi Network for 30+ devices and 300+ clients; he calls it the long-awaited USG replacement.
• Its spec limits are deliberate trade-offs for the price. The host reads the 2.5 Gbps WAN as headroom so a 1 Gbps connection loses nothing to Ethernet overhead; Ubiquiti states full 1 Gbps with IDS/IPS on. Faster LAN ports or multi-gig IDS/IPS would mean more cost, heat and size, which is the $379 Dream Machine Pro (10G SFP+, 3.5 Gbps routing with IDS/IPS). Over 1 Gbps internet, the UCG-Ultra is not for you.
• Each small UniFi gateway fits a different use case. UniFi Express, the priciest of the three, adds a Wi-Fi 6 AP but manages only four more devices and about 60 clients, with no IDS/IPS: for apartments or two-to-three-person offices. The $129 Gateway Lite has no controller or Wi-Fi but full gigabit IDS/IPS: for installers running a hosted or multi-site controller.
• The Dream Router is a step up from Express. A later insert notes the $199 UDR adds a 4x4 Wi-Fi 6 AP, four 1 Gbps LAN ports (two PoE), a 128 GB SSD plus microSD, and runs Protect and Access alongside Network, which the UCG-Ultra and Express cannot. Downsides: rarely in stock, and its 1 Gbps backplane won't reach full gigabit with IDS/IPS.
• The Switch Ultra's PoE budget depends on how it is powered. The $129 USW-Ultra has one PoE+ input and seven PoE+ ports: PoE+ input gives 16 W of output, PoE++ (802.3bt) 42 W, an optional 60 W adapter 52 W and a 210 W adapter 202 W. It has a magnetic mount; one of each Ultra device makes a starter network for about $347 (US pricing), or about $377 with the 60 W adapter.

2024-03-03 · open the original ↗

• Scrum's ideas were published before Schwaber and Sutherland presented it. Petersen, a former lawyer and one of the first Scrum.org professional Scrum trainers in 2010, traces the rugby metaphor to Hirotaka Takeuchi and Ikujiro Nonaka's 1986 Harvard Business Review article, "The New New Product Development Game". He reads the preface of their 1995 book, which says the article used the Scrum metaphor for how Japanese companies developed products, with the ball passed within the team.
• A 1990 book already described "Scrum" and "sashimi" approaches. Petersen shows a 1990 Yourdon Press book, Wicked Problems, Righteous Solutions, containing both terms and citing Takeuchi and Nonaka, not Schwaber or Sutherland. He says the 1995 OOPSLA paper that introduced Scrum lists it as reference 9 but never cites it in the text, which he takes as proof nobody seriously reviewed the paper.
• His objection is to erasure, not borrowing. Petersen says the 1995 paper's wording "We call the approach the Scrum methodology" claims a name the earlier authors used, and that over time their credit has been erased. He argues that applying an idea is not creating it, as applying the Gang of Four's adapter pattern would not make him its co-creator.
• Today's Scrum is a certification mill, unlike the original idea. In his reading the original was about knowledge creation, the ball being the knowledge passed within a team, with Scrum itself an aside. He says Scrum has been co-opted into certification and training, that its dogma rejects home-grown variants, and he counts himself as part of that.
• That gap explains why Scrum fails in many workplaces. Petersen says what organizations call Scrum has nothing to do with what it was meant to be, and calls on Scrum Alliance and Scrum.org to give credit where it is due. He names Prentice Hall as the likely rights holder of the earlier material.

2024-02-29 · open the original ↗

• Postgres is winning on openness, versatility and "it just works". Srirampur, CEO and co-founder of PeerDB, says it has grown from a metadata store into transactional, analytical and search use, and that enterprises are migrating off Oracle and SQL Server to avoid lock-in. He says Postgres revenue has gone from 1% to 10% of Oracle's in five years, and cites DB-Engines for it being the only database still growing in popularity.
• Migration is a snapshot; replication keeps two stores continuously in sync. Offline migrations that can afford downtime use pg_dump/pg_restore; online ones take a snapshot and then read the transaction log (logical decoding / change data capture). Continuous sync suits cases like offloading fraud detection from an OLTP Postgres to Snowflake, where freshness is a cost choice: one customer syncs invoices about every minute, another every two hours, and the interval can be edited on a running job.
• PeerDB says parallel snapshotting turns days of copying into hours. It logically partitions large tables by Postgres tuple identifiers, which map to data on disk, and streams the partitions in parallel. He claims a 1 TB move that takes a couple of days with pg_dump/pg_restore takes a few hours, and that customers who waited four to five days for a 5 TB initial load into Snowflake get it within a day; the repository claims 10x over existing tools.
• Fault tolerance comes from a state machine built on Temporal. The initial load is idempotent at the partition level, so a crash resumes rather than restarts. He reports the service's first outage, five minutes the previous night over an IPv4/IPv6 transition, recovered with no manual intervention; across customers PeerDB moved about 3 TB in 3 hours.
• Open source builds trust; the managed service makes the money. He argues self-hosting at scale takes months of setup, so production users will pay for the managed offering, and he is not worried about copycats because his team works on it full time. PeerDB is Postgres-only by design, a "peer-to-peer" approach of quality over breadth rather than hub-and-spoke; MySQL or MongoDB might follow once the Postgres story is nailed.

2024-02-26 · open the original ↗

• dotnet-outdated is Dependabot for the command line. Hanselman says it "honestly should be built into .NET"; it installs as a dotnet global tool and, after the original author's project changed hands, now lives in a dotnet-outdated GitHub organisation run by two maintainers. It lists each package's available update, coloured by whether it is a major, minor or patch bump.
• Upgrading is one flag, but never upgrade blindly. Running it with -u updates every package (sometimes it needs more than one pass); he then checks the csproj diff and runs unit tests, code coverage, and headless and "headed" Playwright browser tests to confirm the site still works. Patch bumps he usually trusts to the test suite; major ones such as packages moving from .NET 6 to 8 need more thought.
• Pair it with Upgrade Assistant for framework upgrades. With the transitive-dependency option it shows what his dependencies depend on, which is where he switches to the .NET Upgrade Assistant, project by project from the leaf node, then reruns dotnet-outdated to catch stragglers. He prefers that to the whole-solution upgrade a viewer asked for.
• A footer showing commit and runtime tells you what's in production. Following his previous video, his sites print the .NET version, commit hash and build number, which links straight to the deployed commit; he notes some people would show it only to admins.

2024-02-25 · open the original ↗

• Moving from .NET 6 to .NET 8 LTS is an easy call. Hanselman notes .NET 6 is in support only "until this November", while .NET 8 (8.0.2 at recording) brings new features, performance, and support moved out a couple of years.
• Upgrade Assistant makes a simple Razor Pages site a one-line change. He runs the .NET Upgrade Assistant CLI on a new branch; it offers .NET 7, 8 or a .NET 9 preview with their support dates, and for his site it only changed the target framework. He used his GitHub Copilot CLI alias to recall the git command for making the branch.
• The cloud must be upgraded too, not just your machine. The build pipeline needs the .NET 8 SDK (a YAML step on one site, the old visual Azure DevOps pipeline on this one) and the Azure App Service runtime setting must change. For a real production site he says he would deploy to staging behind something like Azure Front Door and A/B test; here he deployed straight to production.
• His only failure was forgetting to push. The site failed until he realised the merge had never been pushed to trigger CI; once pushed, the release deployed and the footer showed .NET Core 8.

2024-02-22 · open the original ↗

• A gateway buys small teams a Stripe-grade API experience. Twist, co-founder and CEO of Zuplo, lists what shipping an API now demands: docs, self-serve key management, analytics and protection, since "your partner or your internal developers will take you down with a clumsy for loop". For large enterprises the pitch is one choke point instead of hundreds of teams' own auth and rate-limiting sprawl; he compares it to Auth0, which he says 95% of the startups he talks to now use rather than building identity.
• Zuplo deploys to the edge, everywhere, by default. He says a git commit deploys in about 6 seconds, always under 20, to 300 data centres, within 50 ms of everyone except the ISS and Antarctica; environments are unlimited (one customer runs 250 previews), and fewer data centres costs more, on the Enterprise plan. Bare-metal and VPC backends connect through a WireGuard-based secure tunnel, which lets SecOps own connectivity and API engineers own the gateway.
• Distributed rate limiting is the hardest thing Zuplo built. Syncing limits across 300 data centres fights the speed of light, so it fails open and may let an extra call through; an async mode lets requests proceed during the check to avoid added latency. Limits can be set dynamically in a few lines of TypeScript, and he says per-node limiters behind an autoscaler, a mistake he attributes to an unnamed product, multiply the limit as nodes are added.
• Zuplo rebuilt itself to run natively on OpenAPI. Twist was not a fan in the Swagger 2 era but saw design-first users adopt it, so Zuplo stores its settings as OpenAPI vendor extensions; customers can make Zuplo's file the master, or re-import their own and have a smart diff keep the extensions while overwriting the rest.
• API keys suit machine-to-machine calls; JWTs suit people. He argues opaque keys are easier to revoke than JWTs, which are part of OpenID Connect and fit calls made on behalf of a person. Zuplo's managed keys are covered by GitHub secret scanning, so a leaked key is revoked automatically with a webhook, Slack or email alert, free tier included.

2024-02-21 · open the original ↗

• Ubiquiti's USW Ultra can run entirely on PoE input. The first switch in the Ultra line has a gigabit PoE++ input port on the back and seven gigabit PoE+ ports on the front, and the PoE budget left for devices depends on how much power comes in. It costs $129 in the US without a power adapter, with a 60 W or 210 W AC adapter offered as an add-on.
• A magnetic mount makes the switch easy to place. The host likes that it holds to metal surfaces without needing a permanent install.

2024-02-21 · open the original ↗

• Local models keep your data off third-party servers, even offline. Hanselman's case is privacy, cost and working "in airplane mode"; he says you need a decent GPU, typically an RTX 2080 or 3080 (some manage on a 1080), and that Intel-only laptops probably won't cope. His machine has a 3080 with about 10 GB of GPU memory.
• LM Studio downloads and chats with quantized models locally. He loads Microsoft's Phi-2 (about 2 GB quantized), Mistral and Code Llama (a 7-billion-parameter model of about 4 GB) and shows GPU memory rising as each loads. Setting GPU offload to -1 puts all 32 layers in GPU memory and is visibly faster, with little CPU load.
• The developer payoff is a local server that impersonates OpenAI. LM Studio's server on localhost:1234 answers OpenAI-style calls such as /v1/models and chat completions from curl; he hit connection-refused errors and had to restart it, and multi-line curl calls failed until put on one line. Responses vary between runs, a reminder that the models are not deterministic.
• Ollama does the same from the command line or Docker. The Windows preview runs from the tray, pulls models such as Llama 2 (a few GB each) and serves them; it also runs as a Docker container with WSL. Typing "bye" in the chat does not free GPU memory, so you have to quit the app. His cache already held 17 GB of models and he warns disk use climbs fast.
• Expect confident fabrication. Asked about Scott Hanselman, Llama 2 produced a wrong birth date, a wrong hometown, a job at Borland and a consulting business called Hanselminutes. He says to think of model output in terms of grounding in reality rather than hallucination.

2024-02-21 · open the original ↗

• VLANs layer separate subnets onto the same physical network. Each Ethernet frame carries a VLAN ID (tag), so devices are grouped by tag rather than by the hardware they plug into. The host's example adds guest on VLAN 10, IoT on 20 and VoIP on 30 beside the default LAN on VLAN 1, each a /24 (254 usable addresses), with the VLAN ID matching the third octet by his own convention. He gives three reasons: security, organization and performance.
• The untagged VLAN decides where VLAN-unaware devices end up. A PC or IoT device lands on whatever VLAN a port has untagged, while a device configured for a tag, such as a VoIP phone looking for VLAN 30, joins that VLAN only if the port carries it tagged. He recommends an exposed jack in a public space, like a doctor's waiting room, carry only the guest VLAN untagged, so a plugged-in device gets guest access and a tagged phone gets no address at all.
• Trunk ports match manual tagging but pick up new VLANs automatically. A trunk carries the default LAN untagged and every other VLAN tagged, which works the same as listing them by hand. The difference is that a VLAN added later (say 40 for cameras) must be added to each manually configured port. Trunks are typically used between switches; an access-point port carries the secure LAN untagged plus the guest and IoT VLANs tagged, each mapped to its own SSID.
• UniFi sets up VLANs in two steps from one console. Networks are created under Settings with a manual VLAN ID, then the Port Manager shows each switch port's native and tagged VLANs, and port profiles apply a set to many ports at once. The host enables client isolation on the guest network but not on IoT, because casting, AirPlay and Philips Hue-style hubs need IoT devices to reach each other; the networks still need firewall rules between them, which he does not demonstrate.
• Mixed-vendor gear works but each device must be configured separately. VLANs are a standard, but off-the-shelf consumer routers may offer only a guest network. The host's Synology RT6600ax allows each of its four ports only to be a trunk or carry a single untagged VLAN, and his Aruba 24-port PoE switch had to be told about each VLAN. Plugging in, his PC received 192.168.20.x on the IoT port while the VoIP phone on that same port received a VLAN 30 address.