from the all-the-information-is-in-the-task dept
That is half two of my collection on utilizing AI instruments to struggle again in opposition to tech firm management. Half one defined why we are able to get past simply begging billionaires to repair our instruments. This half reveals precisely how I constructed my very own software — with zero coding expertise and virtually no cash.
A couple of weeks again, I obtained uninterested in productiveness apps that don’t work the way in which I feel and determined to construct my very own. Not for tens of millions of customers or enterprise capital — only for me. Utilizing nothing however plain English prompts to AI instruments (what’s known as “vibe coding”), I now have a activity administration system that works virtually precisely how my mind works, prices virtually nothing to run, and may’t be enshittified by some growth-hacking product supervisor.
As a joke/homage to the completely good UK TV present Taskmaster, I name my app, “L’il Alex,” in honor of the present’s creator “Little” Alex Horne who’s, in spite of everything, the “Taskmaster’s Assistant.” Which is precisely what I want. An assistant to assist me with duties.
Like loads of individuals, I’m obsessive about productiveness however horrible at productiveness instruments. I’ve tried dozens of various techniques, however none caught. I’m a kind of terrible individuals who retains a psychological listing of duties, supplemented by e mail, calendar chaos, and browser tabs I swear I’ll get to ultimately.
This implies issues fall via the cracks — however extra importantly, it means I’m consistently combating instruments designed for another person’s workflow, not mine.
I’ve tried dozens of activity administration instruments, and so they all endure from the identical drawback: they’re constructed for the legendary “common consumer” moderately than how any particular individual really thinks. The closest I’ve discovered is Intend.do, which calls itself an “intentionality” software moderately than a productiveness software.
The important thing distinction with Intend is that moderately than loading up one million duties that you simply by no means get to (the downfall I — and plenty of others — have with most activity supervisor instruments), Intend is concentrated simply on at this time. You create a couple of bigger objectives, after which every day, you jot down which duties you need to do in direction of these objectives. It makes it quite simple to try this, after which as you go, you possibly can examine off what you achieved. It additionally asks you some questions, and checks in with you over time.
Intend will get each day planning principally proper, however it will possibly’t deal with future activity planning the way in which my mind works. There’s a hacky Workflowy integration, nevertheless it’s awkward. Extra basically, Intend has robust opinions about how I ought to work, and like each different software, forces me to adapt to its worldview. It obtained me perhaps 65% of the way in which there — which is why I stored abandoning it.
That is the core drawback with all productiveness software program: you’re renting another person’s imaginative and prescient of how work ought to occur. When that imaginative and prescient doesn’t match yours, powerful shit. You adapt otherwise you go away. However you by no means get to really management the software.
So I made a decision to construct precisely what I wanted: each day planning like Intend, however with built-in future activity administration, and not one of the philosophical baggage (a few of which can be nice for others, however didn’t mesh with me!). A software that adapts to me, not the opposite approach round.
Right here’s the place the “vibe coding” magic occurs — and it’s approach less complicated than you assume. I began with two fashionable AI coding instruments (Bolt and Lovable) and simply… requested them to construct what I wished. In plain English. No technical specs, no wireframes, no consumer tales. Simply this:
I want to create a sort of activity supervisor, with the main focus being on what it is advisable work on at this time, however which may even retailer “future” duties. The primary purpose can be to let me select which duties I’ll work on that day, put them in an inventory and examine them off, whereas additionally itemizing out future duties that may be saved in a separate display, however which might then simply be added to at this time’s duties as vital. It could even be helpful to have the system nudge me if I put a activity in “at this time’s duties” a number of days with out conducting it, maybe asking me if I need to break it up into a number of duties. The duties may also be separated into completely different classes of duties so I can arrange what kinds of duties they’re. The classes needs to be changeable/editable. Duties needs to be movable from class to class, and I ought to have the ability to prioritize duties. Additionally, it needs to be attainable to set duties to be recurring. Duties also needs to have the ability to retailer clickable hyperlinks and notes. Future duties can have dates connected to them. There needs to be a each day “examine in” on the finish of the day to see how I’ve finished.
I wished to attempt a number of platforms partly out of curiosity, however principally as a result of that is the great thing about vibe coding: if one AI interprets your request in a approach that doesn’t click on, you simply attempt one other. You’re not locked into anybody’s interpretation of what you want.
With that, each created preliminary variations of this software. Bolt referred to as its model “TaskIntent” whereas Lovable referred to as its model “Future Focus Every day.” I began enjoying round with each, making some recommended adjustments or explaining how I wished issues to work barely otherwise. Every service offers you a specific amount you possibly can immediate the system without cost every day (with Lovable it’s 5 free prompts, with Bolt it’s based mostly on the variety of tokens you employ, however is roughly 5 free prompts).
After three days, I spotted that the model Lovable had created was nearer to what I wished and determined to simply deal with that one (that is additionally the place I renamed it to “L’il Alex”; RIP “Future Focus Every day”). This highlights one thing necessary about vibe coding: completely different instruments interpret the identical request otherwise, and discovering the suitable match is a part of the method. It’s not concerning the “greatest” software — it’s about what works to your particular mind.
At that time, I additionally thought I’d see if I might construct and deploy the software for myself for no cash, based mostly on simply giving it 5 prompts per day. By that third day, I began utilizing the software to handle my duties. The one main drawback I bumped into early on was a realization that, regardless of my connecting this new app (at Lovable’s suggestion!) to a database at Supabase, it initially was storing all my duties regionally. So once I logged in on my cellphone to see the way it labored as a cell app, it was empty. However all it took was a single immediate to rectify that, and now my duties are saved in a database.
My plan to do the entire thing without cost hit a stumbling block a couple of days later, once I realized that whereas Lovable’s free account offers you 5 free prompts per day, there’s nonetheless a restrict of 30 whole per thirty days. So, after about six days of utilizing it, you’re probably on maintain till the subsequent month begins.
At that time, I attempted a couple of different approaches. I had related my Lovable mission to GitHub, so I had a repository with all of the code, and I noticed that Bolt had just lately built-in with GitHub as effectively. I attempted to open the repository with Bolt to see if I might proceed engaged on it there, nevertheless it… didn’t work. It actually would ask me which mission I wished to make use of in GitHub, I might click on on the repository, and Bolt would simply reload as if nothing occurred.
Then, on the suggestion of a pal, I attempted to import the mission into Firebase Studio, owned by Google, and constructed on prime of Firebase. As with Bolt, I first tried to drag in my GitHub repo, which labored… besides, for causes I don’t perceive, the AI options then all appeared disabled. After messing round with it for a bit and getting nowhere, the identical pal who advisable I attempt Firebase Studio stated “it would simply be sooner to ask Firebase Studio to recreate the mission from scratch.” I gave it the identical immediate and… it created “Momentum Circulation,” with its personal interpretation of the immediate.
I performed round with it a bit extra, however shortly realized that Firebase Studio positively expects a bit extra data/expertise to make use of and isn’t practically as seamless as Lovable. For instance, realizing I wanted a database backend, I requested it to set that up and (not surprisingly) it recommended I exploit Firebase. However… moderately than set it up, like Lovable did with Supabase, it simply instructed me I wanted to set it up manually. And I stored working into comparable points, the place it could recommend what I ought to do, moderately than simply doing it. I’m certain that’s nice for precise programmers, however my purpose right here was completely different.
For what it’s price, I additionally tried one other software referred to as Adaptive.ai (it named my app FocusDay), which appears fairly cool, however after working into a couple of extra points with it, I gave in and determined that my purpose of a very free app was too formidable and agreed to pay $25/month for 100 prompts at Lovable. It seems that you simply really get greater than that, as Lovable offers you 5 free prompts a day (use ‘em or lose ‘em that day), and after you employ these up, you get these 100 further prompts. So, in actuality, you are able to do someplace between 100 and 250 prompts, for those who time it proper.
Since then, I’ve been iterating the app to work precisely how my mind works. It’s most likely horrible for anybody else — however that’s the purpose. For the primary time in years, I’ve a productiveness software that doesn’t struggle me.
That is what whole management over your instruments really appears to be like like. Right here’s the “plan your day” web page, which I both use late at night time for the subsequent day, or early within the morning for the upcoming day:
Duties age visibly so I can see what I’ve been avoiding. Every little thing flows the way in which my mind really works.
I prioritize duties nonetheless is smart to me that day. I can mark what I’m engaged on now versus subsequent, and it mechanically organizes itself round my precise workflow.
At day’s finish, it checks in with me the way in which I need to be checked in with — not with guilt-inducing streaks or gamification bullshit, just a few easy stats and particulars concerning the day (and it lets me create follow-ups on duties I accomplished just lately).
The system handles my unfinished duties precisely how I would like: no disgrace, no penalty, simply transferring them again to the long run pile for tomorrow’s planning session, with the sort of notifications I would like. So every day begins contemporary, and I get to plan out how I would like my day to go.
The entire thing has labored brilliantly since about day three. After two weeks of each day use, I’m genuinely hooked — not due to gamification tips or habit-forming design patterns, however as a result of the software really serves me as a substitute of combating me. I’m nonetheless tweaking some issues right here and there, however the software simply works. For me.
Right here’s the true magic: once I need a new characteristic, I simply ask for it. Final week I made a decision I wished behavior monitoring built-in into my activity movement. A couple of prompts later, finished. No ready for the subsequent product roadmap, no hoping some PM thinks my use case issues. No upsell on it as a “premium” characteristic. I simply had the system construct it.
I additionally added a bookmarklet for capturing articles to write down about and turning them into duties in addition to an e mail integration for turning emails into duties. These weren’t deliberate options — they have been simply friction factors that aggravated me, so I eradicated them.
The e-mail integration required organising a webhook via IFTTT — the AI initially recommended Gmail might deal with webhooks instantly (incorrect), nevertheless it shortly pivoted to working options. That is typical of vibe coding: the AI makes occasional errors, however iterating towards working options remains to be sooner than begging some firm to perhaps add your characteristic request to their backlog.
Positive, there are nonetheless some small bugs to repair and options so as to add. However right here’s the distinction: when one thing doesn’t work the way in which I want it to, I can really repair it. I’m not caught submitting help tickets into the void or hoping the subsequent replace doesn’t break one thing I depend on.
There have been minor annoyances — often once I requested it so as to add a characteristic, it reverts older adjustments or will get bizarre design concepts — however I can simply inform it to sort things and it does. Evaluate that to each different productiveness software I’ve used, the place design selections that annoy me are simply everlasting info of life I’ve to simply accept.
Intend obtained me 65% of the way in which there. L’il Alex is already at 85% and climbing. Extra importantly: when my wants change, the software adapts. I’m not locked into another person’s imaginative and prescient.
That is precisely what vibe coding is for: fixing private friction factors with out ready for permission from product managers whose incentives don’t align with yours.
So, for all the parents who’re involved about centralized management and dropping company over their digital lives: there at the moment are extra choices past simply yelling at tech corporations or hoping for higher authorities regulation. For private productiveness instruments, workflow automation, and small group initiatives, you possibly can really construct what you want, no coding expertise wanted. The instruments exist, they’re accessible, and so they work. Generally the easiest way to flee another person’s management is to simply cease asking, and simply take cost.
Filed Beneath: company, autonomy, intentional software program, l’il alex, private productiveness, activity administration, taskmaster, vibe coding
Corporations: bolt, google, intend, lovable
