Workflow Wishlist
A sober wishlist of processes I’m working to automate with AI.
Identifying and contextualizing product sales
Why: …because I have a history of obsessive online window shopping. And I like the thrill of getting the best for less, buying something that will last. Even if I were among the cohort for whom a luxury expenditure is a drop in the bucket, that’s simply not the case for most, and I’d like to think I’ll always be price conscious.
Status: In progress. Currently a local app that serves as a deal-finding engine, utilizing everything from emails, to scraping, to Shopify data feeds. I’ve been thinking about making something user-facing from the start, but whether that’ll be an app (just typing that feels like flirting obsolescence) or I use the app to power a newsletter, I’m not sure. At any rate I think that the specificity of what I’m doing (menswear…for nerds, arguably) might prove durable even as AI gets better. Or it might be futile. For now I’ll continue to burn time on this. I’ve made a lot of progress, done some crafty things, and will certainly write a post about this in the future.
Managing and monitoring local development servers
Why: my having access to Claude Code is like putting a paintball gun into the hands of an eight year old. I want to pull the trigger. A lot. Like a maniac. And so I quickly realized I needed a better way to manage local dev servers, restart projects, stop servers lest my laptop get choked out by RAM use.
Status: Done-ish…also in progress. I’ve built Harbr, a menu bar app. I’ve made it open-source for now. You can find it on my github here. I’m also working on expanding its features.
Wrangling a f*ckload of writing, notes, journaling, bookmarks, etc.
Why: Clumsy though human-machine interfaces may be (look at you, slapping that keyboard like a cybernetic monkey, staring at a glowing rectangle), I’m about as enmeshed with my devices as you can be, at least on some fronts. I’m no biohacker, but I am an incessant, obsessive capturer of thoughts and information. But for all of the “second brain” systems out there, none suit me, none impress me. I desperately need a way to connect, parse, surface, and most importantly utilize what I save and write more effectively. For fuck’s sake I better figure this out, lest I be a mere digital hoarder.
Status: In progress. My main tool for information capture is Obsidian. There are LLM-powered plugins for it, but none of them are quite there yet relative to my needs. So I’m making my own.
Tracking fitness progress – volume, loading, RPE, nutrition, sleep, etc. – more effectively, and via good science.
Why: there are great workout and nutrition apps out there but, to begin with, these two areas are disjointed (separate apps), and don’t do a great job of helping you contextualize your activity relative to goals. My brain is always thirsting for the bigger picture. I also want to incorporate more than just workouts and food, i.e. I’d like to see correlations re: sleep, even self-reported mental health.
Status: In progress, code name (not actually a code name) Thorp – an amalgamation of Ed Thorp (an unlikely fitness role model), and Jim Thorpe. So far so good; my first attempt at mobile dev. In this nascent stage the dashboard takes in data from Hevy, Cronometer (and Apple Health) – best to test before going all in on duplicating the functionality of these apps.
Automating content (mostly video) ingestion, and summary output/storage.
Why: Life is more or less a process of taking in, assimilating, and utilizing information. And there’s a lot of good information in them there videos points to YouTube. But watching video is a time suck. Better to have a computer do it, especially since I’m not watching things for emotional effect, or to trigger revelation. I’m just hoovering info, and often what an LLM might not surface…or, at any rate, call it curation and leave me be.
Status: In ideation, but to be honest I’ll likely be raw-dogging a [paste resource link into NotebookLM] -> [copy output and paste into Obsidian] workflow for the foreseeable future. I haven’t yet sunk a ton of time into this.
Making multi-platform e-commerce operations easier
Why: I have a teeny-tiny e-commerce business. Tying listings, lots of SKUs, prices, COGS, consignment, discounts, inventory, finances, etc. together was driving me insane. Shopify and its plugins admittedly do a lot of things very well (including cross-posting to eBay), but have their limits.
Status: Operational, and tweaking constantly. I built myself a pretty comprehensive e-commerce command center. Product listings sync, as do accounts from my business bank account via a very handy bridge. Maybe I’ll fortify and modularize this app, make it available to others eventually.
Domain-specific digests (and maybe knowledge graphs)
Why: I need to keep up to speed on multiple of industries, AI included. But it’s a drag to go hunting for news, hop between substack and emails and youtube and whatever else. So, similar to the video ingestion item above, I need to sic agents on a multitude of information sources, have that info “read”, parsed, and put into some kind of digest. It might also be valuable to extract what’s most novel in the content, match it against a greater, pre-existing graph of knowledge, and extend said graph with it.
Status: In progress. I’m maybe a bit slow to actually implementing RAG type systems and clusters of agents a la OpenClaw, but at the end of the day, the more I can have valuable information come to me – instead of having to go to it – the better.
Personal and business finance
Why: money can be confusing, and unnecessarily so. Cue the adage about how school doesn’t teach you to invest or do your taxes. Cue my frustration, too, with the design of popular finance apps…par for the course when designing for a common denominator, I suppose. I just want an absolutely juiced up finance app that’ll let me get an overview of my financial health, and execute financial planning with ease, and into the nitty gritty when I want to.
Status: In progress. I’m building something that I think is quite powerful. When developing an app I always do my due diligence, see what’s been done before and well, see what UI/UX patterns might be useful. But because this is my own app, I get to go outside the lines, push into functions and areas you won’t find in any other app. Already, being able to see both business and personal accounts, subscriptions, and transactions categorized by things like tax category is helpful.
Unifying work across AI/LLM tools
Why: as marvelous as the likes of ChatGPT and Claude are, they can’t talk to each other. Claude can’t see what’s in ChatGPT, GPT can’t see what’s going on in my terminal with Claude Code. Moreover, I simply have a years-long conversation trail with multiple LLMs at this point, spread apart, locked into each respective app. As of this writing (5/30/26), Claude Chat is separate from Claude Cowork, too. Some of this can be remedied with a bit of diligence, i.e. being deliberate about what work you’re doing where. But I’m already behind the eight-ball.
Status: In progress. This impulse to unify and centralize is a theme in many of the above projects. Unless personal computing shifts overnight and renders these issues solved and therefore null, there’s value in not letting work/outputs fall into the digital abyss – so I’m working the problem. Part of remedy will be going through old convos, exporting or simply pasting into Obsidian. Another part might involve file architecture. Maybe I’ll find a solution once I have some agents running on a second machine. I’ll be working the problem in time.
Designing and modeling a physical product
Why: I’m designing a physical product – very slowly, I admit, but I’m beyond square one – and some 3D modeling is par for the course because, 1. I’m creating something that simply doesn’t exist right now and, 2. it can be 3D printed, at least for prototyping. Thing is, I know next to nothing about 3D modeling. I’m not necessarily trying to skirt learning a new skill, and remaining completely ignorant makes me feel uneasy anyway, but I also want to explore what might be available in terms of AI-aided design. This is less some recurring automation, more a one-shot, ‘can AI produce this?’ or ‘can I avoid losing a lot of time to mistakes?’ exploration.
Status: In progress. Some time ago I found a startup (web app) offering prompt-to-design as a service, but it didn’t seem like I could demand the sophistication I needed. In other words, it worked for designing a desktop flower pot, not so much for what I’m designing, which has a spring, a hinge, will interact with other materials that aren’t uniform, i.e. a small amount of complexity. I expect that there has been progress in this area, and maybe some design apps have incorporated prompting, but I’ve yet to go off and get a grip on the state of things.
Monitoring and managing system resources
Why: Vibe-code close enough to the sun, and you’re going to put a strain on your computer. Especially if, like me, you’re hard on your computer to begin with. There are resource monitoring apps out there – Stats (free; I do currently use this for network monitoring), iStat Menus (paid) – that are pretty good. Some apps even allow fan control. But none mesh with my mind all that well or, more to the point, don’t offer the granularity I’d like where I’d like it.
Status: In progress. I’m working on a menu bar app called Whirr. So far so good. Its chief feature is the simple narrative it puts to things when you need to see what’s going on at a glance, e.g. “System is swapping heavily”; “Big processes that look unused…”, and a tight display of top processes and their consumption.
More to come soon, I’m sure…