


🎙️ Podcast host @htmleverything 🧑💻 Tech enthusiast & web developer 🎙️ Podcaster @DOPMGaming ⛳️ Triple Bogey Specialist 🎮 Always gaming


Can't even talk to someone these days... Sometimes I'm just gonna have a chat with someone, I don't need an automation to fire off writing a blog post or some other nonsense about it.

Been working on a major re-design the @htmleverything website. Here's a sneak peek 👇 Open to suggestions and feedback

High time this become available on PS5, eh?
NOVEMBER 17, 2006 📅 Resistance: Fall of Man released 20 years ago today in North America as a PlayStation 3 launch title. Who was there when #ResistancePS3 launched two decades ago? 💀 🪖

This is a crazy AI limitation on Copilot. Copilot is inside the Outlook mobile app, but cannot add events to the calendar. I purchased tickets to an event a while ago and realized that it wasn't added to my calendar. So I asked Copilot to find the email and extract the details - which it did in seconds. I then told it to add it to my calendar, it replied that it could make some changes like rules and such to the calendar, but wasn't allowed to add events. The workaround, I told it to generate an ICS file, which I then downloaded and applied to my calendar. How is Copilot still not able to add to an Outlook calendar? Am I missing something here?

Your AI workflow doesn't need to be all AI. People forget that an AI workflow is still a workflow. You can have a real person be involved in one or more of the steps. Spending an hour or more to automate tasks that would take a human seconds is a waste of time, especially in the short term.

Context switching is absolutely killing my productivity lately, and I blame AI. My days feel like they consist of chipping away at 10 projects, 1% at a time - I have absolutely no focus anymore. If I'm curious about something, I fire off a prompt to get research started. When I have something on my plate that can be delegated, I engage my AI agents. And all of this culminates in me running from task-to-task all day, most of the time the tasks are completely unrelated to one another. To escape I might scroll my timeline a bit, only to find something else I'm curious about - and another research prompt gets fired off. I've mentioned on my podcast many times that the human brain is the bottleneck for AI's outputs and I'm feeling it more than ever before. Anyone have some advice on locking in when there's just so many things I feel I should be doing?

First episode edited in this new style, may tweak and hopefully get better at it, but I'd love some notes from the audience htmlallthethings.com/podcast…

Trying out a new podcast editing technique. I used to primarily edit for audio quality and cutting out dead air. On top of those two things, I'm now trying to "edit for pace" where I'm cutting out some re-iterations of points or hesitations while we're talking about more intense topics. And I'm trying to do so organically, where I'm not cutting all the filler words or hesitations, but removing some of them to keep the listening experience progressing. I'd love some feedback from the audience, the first episode like this dropped over the weekend 👇

Does technical debt matter in 2026? Some think that because AI has taken over much of the coding that we do, that having a codebase that is largely unmaintainable by humans doesn't matter. Personally, I think this is rubbish and equate it to having a messy workshop. In your workshop, if you leave a rake lying on the floor - then you know what to avoid... but it still poses a danger. If a stranger were to walk into your workshop without warning then they'd likely step on that rake. Okay so we'll just tell the AI that there's a rake there and it'll avoid it every time... right? Maybe, but what about when you're 100 prompts in and it starts to lose context? Once that context is lost, there's still a major risk of it stepping on the rake - so why leave the hazard there to begin with? In fact, why not tell the AI to pick up the rake and put it away so we have a clean workshop ready to continue working in? Personally, I think having a maintainable codebase that fights against technical debt is still the way to work in 2026.

For ChatGPT I'm finding that GPT-6 set to Medium is the best result to value ratio. I'm on the Plus plan. With Astra I starting hitting my limits when set to High (not really surprising) but it scared me back to GPT-5.6 initially. Lately, on Chat heavy days I've been rocking GPT-6 set to Medium and have been very happy with the results and speed I'm getting. In terms of agents though, I've only tested Codex a little bit so I'm not sure how it will affect my limits. But the results from my agents have also been great at this level.

I've been enjoying ChatGPT and Codex. But one major sticking point is the switching between modes: Chat, Work, and Codex. This has got to be fleshed out more. I'm able to get the same work done in all three modes often times. Sometimes there's tasks that clearly lean one way or the other... but in general I think this UX needs to be improved.

We value handmade furniture. Would anyone value "hand-coded" apps?

Is software dead? I think the whole "software is dead" movement is nonsense. Can AI generate you some software? Yes. But we've had site builders, templates, and other DIY tooling to let business owners make their own websites for years. And guess what? There are still web development agencies servicing hundreds or thousands of clients. People want stuff made for them when their time is better spent working on something else. The general public isn't going to want to update their app when an OS update breaks it. They're not going to want to install security patches and other fixes. Is software dead? No.

I'm digging this new Activity feature on @descript The old chevron with collapsible menu left a lot to be desired.

Definitely not a fan of the Microsoft 365 changes, lost terabytes of storage capacity Surprised they'd just take away OneDrive storage away that I've had for years...