53
this post was submitted on 08 Oct 2026
53 points (73.5% liked)
Technology
88677 readers
3726 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
It's the 80/20 principle.
Getting a piece of software done, the main work, the 80%, takes 20% of effort.
Refining the remaining 20% on the other hand takes 80% of all effort and time invested.
Another aspect is burnout. Sure, vibe coding means sped up development, but what I noticed was... I also lost interest in projects much quicker. Unless it directly delivers something I need - e.g. I've built a 3D printer stand for my new Bambu H2C - it's going on the pile of "80% finished crap".
And the thing is, you CAN build an AI setup that automates everything. I've put together a skill/agent/rule stack that is essentially an entire company's engineering department. It works autonomously, only looping in the user for major decisions (and those decisions get a proper report the user can read to contextualise the decision itself). It creates initiatives, epics, then refines those into stories, then stories get planned with tasks, subtasks, etc., which then get the design-develop-test-verify loop (test here stands for test writing, while verify is the QA smoke test). Stories get the appropriate level of detail, with a BDD approach (so happy path, unhappy path, and edge case scenarios are defined before work begins), experts are created on-demand, the list goes on. All of this with 6-8-10 parallel teams.
It works incredibly effectively.
It also burns through tokens like there's no tomorrow. I let it run over a weekend - Saturday morning to Sunday midnight. It correctly identified the project needs for a relatively small thing (something a sole developer would write over 2-3 months), created an appropriate enterprise approach, broke it down into 9 MVP stages, and in 2 days, delivered the first four of those.
At the cost of burning through an entire weekly allowance of Claude Max 20X tokens...
I guess I could pay another £150 for a secondary account and have it churn through my old projects to refine and finalise them, but it's hardly worth the money.
And no, while I say that it's a "full company", it's only because I am an actual software engineer, I understand the code being written, I understand the processes, the flows. It's not something you can toss to a complete beginner and have them build a successful business on top.
With LLM it's closer to 60/40, you can get an MVP in a day or two, but the next step is either impossible or so hard it will require so much resources that it makes it almost impossible. But if you start with a very clear idea of architecture and software stack - in that case you'll get 80/20. Then the real problem starts - you don't know your code, and have no idea how everything interacts and there going to be thousands or tens of thousands of lines of shit that does nothing or done something in the past that LLM forgot to remove, or it left a document that describes it so it won't remove it and so on.
It's kind of weird because it kind of acts like 80/20, but you end up doing a significant effort to unwind things back to 60/40 to start trying to massage it to what you really need.
Assuming it's one of the areas where it does relatively strongly, other areas it will flounder far worse and for some scenarios it can essentially one shot a working thing, if it is really a supremely common and simple thing.
No, it's not really closer to 60/40.
You can deliver the core of the product - the 80% - in the indicated 20% effort zone, aka a few days.
As for not knowing the code... I actually do read what's being output. Just because I'm not familiar enough with the language or framework to write it myself with confidence, it doesn't mean I don't understand it.
Beyond that, the review loop I described above incorporates experts that are language/framework specific to spot issues. Static analysis can easily spot unused code paths - plus that gets covered by tests too, so any leftover is spotted. And even further, after each MVP cycle (which I use in place of sprints as sprints are for human effort cycles which LLMs don't really have, so managing actual target goals instead of physical time periods makes more sense), I have a multi-model (as in, different models from different sources, not just e.g. "all Claude but using Sonnet, Opus and Fable", but more like "Opus 5.5 and GPT 6.1 Sol and [insert currently best model for language]") quorum of experts review the code, test it end to end, find unit/integration/e2e/UI test gaps, write up everything as big/chore tickets, and have another go of the main loop fix those. Same for documentation, it's kept up to date in the same loops.
You get 80% there by being a software engineer to start with, without proper discipline and experience in the field it won't be 80/20.