I also find creating my own little shortcuts to be super satisfying and of course very transferable since they work out of the box.

```some examples " Quick access to commonly edited config files (and a directory for my Shell scripts!) map <leader>v :e ~/.vimrc<cr> map <leader>V :source ~/.vimrc<cr> map <leader>w :e ~/Workspace/myCo/tmuxp-session.yaml<cr> map <leader>W :e ~/.tmux.conf<cr> map <leader>z :e ~/Shell<cr>

" Super simple in editor note setup, amazing map <leader>x :vs<cr>:e ~/Documents/notepad.txt<cr> map <leader>X :vs<cr>:e ~/Documents/notes<cr>

" Quick terminal pane map <leader>t :vs<cr><c-w>l:term<cr><c-w>j:q<cr> " Pull file path into clipboard nmap <leader>b :let @+ = expand("%")<cr> " Pull current line into clipboard nmap <leader>B "*yy ```

Quick disposable terminals, tons of short cuts to get me into the config files that make it all happen (vimrc, zshrc, tmux.conf, a tmuxp session file) and to reload them, and super quick access to a well organized directory of notes are all huge boons during my workday.

I believe there's a fair amount of tax implications involved with that bucketing though. Capex is taxed at a lower rate than opex is my understanding but I may be wrong on the specifics of it all.

You got downvoted for the snark, but damned if it ain't a reasonable opinion.

If you read the seminal "Design of Everyday Things" by Norman Rockwell you'll come away annoyed at half the physical _doors_ you walk through... here in 2025.

I've been pushing these terms to help us talk about and design better interfaces at work...

Static Interfaces - Your supermarket's pretty much a static interface. The frame of whatever website you're looking at. These are static. They've very powerful and were pretty much all you had before digital interfaces became ubiquitous. There's an initial learning curve where you figure out navigation, and then for the most part it's fairly smooth sailing from there provided the controls are exposed well.

Adaptive Interfaces - These interfaces attempt to "adapt" to your needs. Google is probably one of the most successful adaptive interfaces out there. A query for "Shoes" will show a series of shopping results, while a query for "Chinese food" will show a map of the restaurants nearby. The interface adapts to you.

I call this narrow adaptive because the query triggers how the UI adapts. I think "wide area" adaptive interfaces where the interface attempts to meet your needs before you've had a chance to interact with the static interface around it are tremendously difficult and can't think of examples of them being done well.

Adaptable Interfaces - This last interface bucket includes controls which allow a user to adapt the interface to their own needs. This may include dragging icons into a particular order, pinning certain view styles or filters, or customizing the look or behavior of the applications you're working with.

Finder, the iPhone's basic UI, terminal, basic music catalog management (e.g. iTunes)... these are interfaces which are created once with an initial curve of varying difficulty to learn and then live on for decades without much change.

Conclusion - The best interfaces combine an intuitive static frame, with queried adaptive elements, and adaptable features to efficiently meet the needs of a diverse group of user flows instead of attempting the one size fits all approach (which leaves 2/3rds of people annoyed).

Another category, searchable interfaces, may fit into one of these or may be it’s own separate category. But tools like MacOS Spotlight or the command palette in some editors are very useful for power users. Having every command available through a minimal set of fuzzy keyboard strokes is a significant productivity boost, while also allowing some degree of discoverability.

As an aside, if anyone at Adobe is reading this, this sort of tool would be an excellent addition to Illustrator, Photoshop, etc. InDesign already has something like it, although that implementation leaves a little to be desired.

Or you can focus on a single class and produce the best UI for that class.

Static Interfaces for the common actions that everyone does. Best as basic utilities in the operating system (Notepad, The calculator)

Adaptive Interfaces where you have a few advanced layouts for people that wants a bit more. (Wordpad, Notepad++, Kate,...)

The expert tools (Blender, matlab, Adobe Illustrator,...) You will have a small userbase, but they're often willing to pay for a good tool that will solve their needs.

Small nitpick, "Design of Everyday Things" is written by Don Norman.

Thanks, got a feeling when I typed that that there was something off about it... can't edit now!

The fact frequent, repeated contact with the customer isn't the norm is why so many interfaces suck and so many engineers could't design a decent one with a gun to their head (although, frankly, that level of stress might not induce thoughtful design patterns).

Instead engineers get hit with micro view after micro view, and they build it using test flows that don't mimic the real world, and then they all tie it in to create a tangled macro view that's a shit show for the user.

I've been working to bring recurring Shadow Sessions here at my workplace by creating a basic scheduler (which is really the pain point at scale) that just sends you and somebody working in the tooling you're building (we're internal tooling) every three weeks and the feedback is overwhelmingly positive and we're working to expand the functionality a bit.

So, all you out there who want a nice win, set up a little scheduler and get your Product, Design, Engineers, Managers, and TPMs in rotating sessions with actual customers at a lightweight pace with minimal asks to create greater empathy which translates to all of us potentially ending up with better software in the world as a whole.

A fella can dream.

Just wrapped up reading "The British are Coming" by Rick Atkinson (great book!) about the start of the American Revolution and half the letters from the generals were in request for or talked of shipments of rum and mead and wine. Many captured soldiers were just the ones too drunk to flee when they needed to.

It was a brutal endeavor. They did not have ice cream.

Basically any... autopilot behavior for me is going to turn into some insight at some point or another (although it's good to be in the present too! Now I try to think in Spanish which forces me to be present cause it's such a slow process for me still lol, but actually super useful I've found for continuing to flesh out those neural pathways is to just sit and talk to myself outloud in Spanish about whatever I'm doing all day)

Also love writing things on paper. I love using different pens, and write in different styles sometimes (cursive, block, normal, whatever). I just take quick throwaway notes on paper or I'll sometimes sit down and outline things for an hour over a few sheets just to put organization to some large task which I'll translate into something for consumption by others later. Almost always throw these away. Sometimes I work through a notepad, sometimes it's a little notebook, sometimes it's printer paper... getting away from the screen for a bit is lovely.

And I set up a notes section on my computer which I can pop open with a quick <space>X in ViM which has a directory structure giving me spaces for recurring meetings, architecture work, or notes on whatever initiatives I'm trying to push through. I index file names with 01-, 02- etc and if I have a section that branches out I'll just add another index so 01A-, 01B-, etc...

Having everything in ViM with a tmux conf setup that scaffolds out a ton of different spaces I've set up mnemonic keyboard shortcuts to navigate to is really helpful in reducing the overload of having Slack, IDEs, browsers, notepads, music, etc etc etc all open at once (since on a good day I just have Slack, Outlook, iTerm, Firefox and maybe Figma or Music open at most points). I'm also quick to wipe away tabs in the browser...

Next up I'm diving into Figma so I can turn these ideas into beautiful diagrams and even mock interfaces my partners can ingest visually for easier understanding of complex spaces

Finally, and I'm thankful I can afford to do this, but I turn all notification sounds off, Slack'll bounce once from the hidden dock...

It's great, distraction free, tailored to my needs, a fair amount of variety, and some really nice productive times.

I'm slightly confused as to why you'd use a LLM to sort structured data in the first place?

The goal is to do more complex data processing, like build dashboards, agentically figure out which tickets are stalled, do a quarterly review of things done, etc. Sorting is a tiny task in the bigger ones, but hopefully more easily exemplifies the problem.

I don’t understand how this can work. Given probabilistic nature of LLMs the more steps you have more chances something goes off. What is good in the dashboard if you cannot be sure it was not partially hallucinated?

> What is good in the dashboard if you cannot be sure it was not partially hallucinated?

A lot of the time the dashboard contents doesn't actually matter anyway, just needs to look pretty...

On a serious note, the systems being built now will eventually be "correct enough most of the time" and that will be good enough (read: cheaper than doing it any other way).

>On a serious note, the systems being built now will eventually be "correct enough most of the time"

I don’t believe this would work. File a “good enough” tax return one year and enjoy hefty fine 5 years later. Or constantly deal with customers not understanding why one amount is in the dashboard and another is in their warehouse.

Probability of error increase rapidly when you start layer one probabilistic component onto another. Four 99% reliable components sequenced one after another have error rate of 4%.

We're all just going to be glorified debuggers trawling through reams of generated code we've never seen before to root out the hallucinations.

Everything you described is already solved by Metabase and few other tools. It takes a few hours to make daily reports there and the dashboard of your dreams.

And its not like it changes every day. KPis etc stay the same for months. And then you can easily update it in a hour.

So what exactly does llm solve here?

The on a log scale is doing a fair amount of lifting here I think

There was a primary, it just wasn't a hotly contested one. I voted in it...

That being said, there was no contest post Biden drop out, although there was a party alignment

The person that ran for office was not the person that won the primary.

Even short day dreams can be incredibly productive which is why I keep my phones in the other room away from me.

This forces me to get up and walk into the other room every time I have to do 2FA at work which has a ton of benefits. I'll bring dishes or cups to the kitchen on the way, very frequently have useful thoughts about whatever I'm working on, get up out of my chair more frequently, and look at things farther away than my screen which relaxes the eyes.

In general, I advocate for avoiding any product with an infinite scroll as I find them detrimental to my own health, extremely addictive, barely rewarding, and frequently enriching to people I barely have any good impressions of.