DEV Community: UnitBuilds

We are no longer continuing the interview.

UnitBuilds — Mon, 05 Oct 2026 17:08:35 +0000

So late last night, I got an update from Wasmer, that they will no longer be continuing with my application. A bit sad, ngl, especially given friday wasnt by choice that I had to cancel the interview. But all the best to them.

So what now? I've been applying to more jobs and prepping my pitch deck, so I can look at funding UnitBuilds CC and V.E.L.O.C.I.T.Y. IDE! Which honestly would be amazing!

Right now, I'm taking a break though. Since I left my job, I've been non-stop busy and couldnt rest. So I decided I need a hiatus, rest and relax, then get back at it in a week or 2. In the meantime, I'm going to get back to my roots and relax, play some games (Tainted Grail has been REALLY fun!), went to the beach today, gunna do a BBQ now and go to bed, without the sound of 2 laptops and 2 computers running in the background.

Honestly, not very excited taking a break, but I think that's all the more reason to take 1. I wont stop developing during this time, infact, I think I'll be developing even more, but on the things that matter to me. IDE has gotten to a pretty solid place now, so it's minor refinements and testing. I DEFINITELY need to update my website (yeah, it's bad). Probably should release the health and fitness app, then beyond that, I'm not sure. Maybe refine pitch deck, set up cloud compute and model routing for the IDE, etc. That way it can live up to the dream I envisioned: "A lightweight IDE, that scales across all your available computers and if it has to, cloud". I kinda wanna refresh UnitBuilds itself (the original foundry) and open it up, I mean I have token plans to cover it, might as well open it up so people can test it.

Overall, I think I've gone too deep down the rabbit hole and I need some self-reflection time. Maybe finally read the Iliad and disconnect for a bit. But that's not quite me, I like software development way too much to quit cold turkey.

On other fronts, Doccit is doing well, I added regional support for practically most of the world, I wanna expand on it though. How many of you track your time and need time tracking apps like clockify to log your time. Imagine of the ide integrated with Doccit and automatically logged what you're busy with and where you spent that time... Optionally, with agent time tracking (not for billing, for production oversight), to see how much time the AI is doing the work and you're actually free to do whatever you like (lets face it, most of us either browse, or work on something else if we're invested).

So to finish up on this scatter brained spill of my mind. I have a question... If your IDE could do 1 thing, 1 magic thing, no matter how insane, what would it be? Genuinely, what is the thing you'd want most?

I had a few ideas, like a 'bottleneck detector' that highlights high traffic code that could be improved. Or a 'get it done' mode, where an AI team fully takes the product from 0 to 1, including comprehensive testing (not what passes for it by today's AI standards... Actual 24h+ usage to test it in real-world settings). Maybe a 'test suite' criteria checklist, where you state what you want tested and and what part of development, like E2E testing, integration testing and manual site traversal testing. I mean the IDE has full windows automation, so it could theoretically even use AutoCAD for you, if the model was smart enough to know what it's making. Or the core concept that UnitBuilds was founded on, 'prompt to binary', a true E2E production lifecycle in a single click, where your idea becomes a reality, no matter how many iterations it needs to go through to thoroughly complete and test it. Though as I experienced with the original foundry, sometimes, we're just a bit to vague on what we want and that leads to over or under engineering.

I'm sweating bullets

UnitBuilds — Thu, 01 Oct 2026 11:58:19 +0000

Well, about 4h30m to the interview and I'm stressed. Honestly, this is why I hate interviews, because it seems the more I study up, the more I feel like I dont know what I'm doing, even though I built it...

Mostly, because during the tech interview, I was asked really in-depth questions and expected to talk through implementations, quoting method signatures, which isnt really something I consider a skill of mine...

Not to mention, last time was just with the Tech Lead... This time it's with the CEO, Tech Lead and a Senior Dev... So a simple 'I did it this way, because that's the right way to do it' wont cut it. I need to be able to explain why I chose it vs what alternatives could have been used. All things I dont generally give a hoot for.

So needless to say, I'm staring down a barrel of a gun and I know I'm gunna dance like a chicken and make a fool of myself. Idk about any of you, but I work on multiple systems at once and I dont much pay attention to anything in particular, more the overarching architecture and making sure things actually work, are tested and benchmarked. If the benchmarks are unsatisfactory, then I optimize, simple as that. I dont much consider alternatives until I have the time to explore them and I dont consider them till they're tested and benchmarked. Not quite the kinda thing that shines in an interview...

What makes it worse, is it's an hour... For the AFS and the pgrust port combined... They wont beat around the bush, they're gunna hammer me with hard questions that I know I'll choke on. That being said, I have an ace up my sleeve... My SASOS can finally run and render native Notepad++. If I can find a gap to show them that, it might distract from the fact that I'm scatter brained and cant explain why --net during build is the exact fix needed 2 layers down, in order to make pgrust print times accurately. I'm stressed and I'm anxious and rightfully so, because I feel like I dont know what I'm doing. It's kinda like a carpenter who learned by trade, being asked 'what's this chisel' and they just call it 'thin chisel'. I can write solid code, but dont ask me to explain it, just say thank you that it works and it's tested.

In other news, IDE got a major upgrade. I added nodes! You can now host a headless running on an ubuntu server, then from your GUI instance, allocate work to be done on that system, so if you have a few old systems lying around and you wanna boost your productivity, you can now work on 1, execute on many. Next is cloud VM scaling, where a cloud instance can trigger expansions (think Kubernetes) for each agent it spawns, so ephemeral state agents can execute in their own sandbox environment, while maintaining connection to the sitemap. That way like you have a local LLM and a cloud LLM, you can just 'offload to cloud' for heavy builds or if you really want to power through a complex system via the teams studio.

Idea for that 1 being, wasmer's edge is actually perfect for it... Think about it, if an agent spawns as a .wasm instance, you can scale thousands of agents and scale back down to 0, cuz each agent is ephemeral, they just have their skillset and objective.

See... Scatter brained, I'm meant to be stressing about the debrief and I'm talking about my side projects! It just feels to me like I'm suited for working, not for interviewing... Cuz these are the kinda things that add value on-the-job, not before you land the job. They're looking for competence and I'm giving them experimental 'I'm not quite sure's' and that's what stresses me. Relaxing didnt help much either, clearing my mind came up foggy at best and I worry that I'm gunna trip up on the finish line. It took them a day to review the work... In that day, as the people who maintain the system, they most definitely knew exactly where to dig to find the flaws and the things I never thought of. An 'I dont know' or 'well that's why you're the lead' wont work... This time the interview isnt just with the lead (he's nice, during manual code review, he actually gave me hints, cuz he wanted to help), this time, it's him + 2... I dont know the other 2... I dont know if they're entering constructive, or destructive, if their goal is to weed me out, or to see my potential. Cuz I'm pretty sure I'm not the only candidate in the running at this phase. Probably am the only 1 who clocked in less hours than their predicted minimum though... And only 1 who logged AI hours at over 100h... But those can count to my benefit, as much as they can count against me. What I see as optimization, they maybe see as lazy ai prompting if I cant explain it.

So now that I've had my existential crisis and I've accepted my fate, come what may, I've got 4h left to prepare for my impending doom... Social anxiety's a bitch aint it.

Wish me luck (I'm cooked...)

Tomorrow is review day

UnitBuilds — Wed, 30 Sep 2026 13:06:03 +0000

So update for anyone following the journey. I finished the hiring task submissions and sent them through yesterday at 3am. Today, I got the email for the final review scheduled meeting. 6pm CET tomorrow. This is what the entire hiring pipeline boiled down to, this final review.

I'd be lying if I said I wasnt stressed, or anxious, but I think I got this. After all, I did build the systems they're gunna be asking questions about, not like it's a blackbox file out of context again.

So fingers crossed and lets see how it goes! I'll write an update once the review is done on how it went and when it's all said and done, I'll write an overarching post, detailing the hiring process, so people who dont know what to expect when going into the hiring pipeline have a reference point and know what are the things that I tripped up on and what are the things that really impressed them, so you know where to spend your time and what skills to refresh on.

I finished at 3am...

UnitBuilds — Tue, 29 Sep 2026 10:46:35 +0000

So I've been doing the Wasmer hiring tasks. As expected, they grossly underestimated the time required for the pgrust port... So why did it take me 8 days, but less than 4 hours?

To answer that question, you need to look at how I work. I built my own IDE, specifically because agentic orchestration made too many mistakes. So I fixed those mistakes, by building the infrastructure myself.

Now on to the fun part, it took me 2 hours of me-time to actually spec out the gameplan, find all the dependencies, chart it out into a roadmap and set up the team to do the task. Then it all boiled down to compile times.

I dont know if they'll hire me, but 7 PRs, 2 of which were production bug fixes. A fully working pgrust port published and tests and benchmarks to prove it. I dont know what more I could have done.

I finished the assignments and did the final submission at 3am today (10 hours ago), then went to bed. Woke up to my email being shared in the Slack channel and not a word since.

I honestly dont know what more I could have done. Yet I still feel it wasnt enough... And I think that comes down to the lack of ownership you feel with AI generated code. I signed off on everything, every edit was by my design, yet I just spent 4 hours reviewing, correcting, guiding, planning, never doing. So I guess it comes down to what do they value.

I think this is called progress?

UnitBuilds — Mon, 28 Sep 2026 06:33:21 +0000

So I've been working on Dwarven Stronghold for a while now. At one point, an overnight agent wrote a few thousand files as templates... This is the templates being fleshed out, file by file.

I think fair to say, it's getting to be a bit bigger than Dwarf Fortress, fingers crossed it's as good!

Also if anyone's curious, current LOC for the entire project is at:
Total Files: 9,652
Total lines: 10,519,594

I think they're angry?

UnitBuilds — Sun, 27 Sep 2026 15:50:05 +0000

Lmao, I think I won? 82 identical comments... 82 comments from me warning people. Comment 83: Nice bro.

https://dev.to/dev_supports - Please report ASAP, phishing scam and the anti-bot isnt detecting it!

UnitBuilds — Sun, 27 Sep 2026 15:10:25 +0000

DEV SUPPORTS - DEV Community

dev.to

Please confirm this is real

UnitBuilds — Sun, 27 Sep 2026 14:42:37 +0000

DEV.to team, please confirm this account is actually yours. As it's sending links for verification 'due to increased bot activity' in the comment sections. Kinda sus because it's a brand new account...

Anyone here use a buildbox?

UnitBuilds — Sun, 27 Sep 2026 08:45:50 +0000

So I need to compile rust from source, so I did the logical thing... Try it... On a laptop... It didnt work.

So I decided to take a pc of mine, drop everything and install ubuntu server on it! Welcome to the world of having a buildbox. Simple setup, connected to your network, any task that needs heavy compute, or in general can be run in the background, have your AI send it to the buildbox.

What I noticed though, was quite surprising, it started off with Nix... And it took a while... But then I realized that most of the stuff is being built via wsl, so I decided to prompt it to offload all the building to the buildbox. Not just cuz my laptop got exhausted, but because artifacts take up a ton of space! So with all the work offloaded, my agents run faster and because the work is distributed, work gets done faster.

It's the first time I decided to designate a buildbox, previously, I'd just use extra systems for scaling parallel workloads, but this got me intrigued... If I have a fleet of buildboxes, each with podman, then I can create worker pods... Essentially turning the systems into a kubernetes cloud. I mean if it ran on windows, why not? It should fit comfortably in a low overhead os?

So coming weeks, that'll likely be the main thing I'll be experimenting on, given that I designed my IDE to be cross-platform and uses native sandboxing, a stripped down 'cloud pod' version could easily run in a tiny container and use a handful of megabytes to be fully isolated and ready for any task... So much potential, gunna be a fun experiment!

Do you ever feel demotivated?

UnitBuilds — Fri, 25 Sep 2026 18:24:14 +0000

I've been going through the paces the last month, first a grind while working, then left my job and kept grinding, then got an interview and yet, still grinding... Yet I cant seem to stay focused. For so long, I've been wanting an opportunity to get a better paying job and I got it! Yet, now I feel demotivated to follow through.

Let me elaborate, I have 2 tasks to do by Monday, 1 is already done, the 2nd is 99% there. Yet, I dont feel any excitement? Infact, I feel more motivated to stop early and not finish, than I do finishing it up. Though I imagine a large part of that is keeping my expectations tempered, so I dont get my hopes up too high, yet it affects me.

For perspective, I designed an abstract filesystem and ported pgrust. The first is a simple POC, but the 2nd is a pretty damn 'substantial' amount of work. And yet, as a task, I loved doing both, it's the interview that comes after I dont look forward to...

Idk if I'm alone in this, but it feels like hiring processes are being stretched out far too long. I'm from a small country, my last job was literally walk in to their office, 30 min tour and chat, asked how much I'm getting paid currently, he replied 'we can do that', contract in my inbox by the time I got home... Quick, simple, efficient, none of this long-winded 'but can you do this... Ok, but can you do THIS... Oh, how about this...' I hate it, I genuinely hate it.

I know I'm standing at the finish line and it's just that last interview standing between me and the job, but I despise interviews so much, that I dont even want to. I will, obviously, I'm not so full of it that I'd quit the process now... But it bothers me that my mind goes to that place. And I imagine I'm not the only one?

Part of it is definitely performance anxiety (stage-fright), while a large part of it is also fear of a ball and chain (like my last job turned out to be)... So dont worry, I'll finish the tasks, do the interview and land the job. I dont need any encouraging foot up my ass to get the job done. I just wanna know if any of you feel the same?

It's definitely not an isolated incident and most of it is caused by overthinking, yet I cant help but ask who else has found themselves at the precipice and choose not to jump? Well that sounds scary, the reality is way bleaker, you climb a ladder to the top, then jump off... Makes no sense, hence you dont do it, but why does the mind go there?

I guess a large part of it, is also what bothers me daily... I quit my job, so I can relax for a month or so, but now I'm trading job anxiety for hiring anxiety and it consumes my every thought all the same. It's not what I'd consider healthy rest, or relaxation, hell I cant even sit down and play a game, because I'm so stressed about getting the tasks done right. I cant even explain how many times I went through the AFS and its documentation, yet I still cant remember a thing because of the anxiety. I think I should just finish the last 1 up today, take tomorrow as a rest day, then sunday do a final run through and study up on it (maybe with flashcards).

I dont think this kind of reaction to 'getting a job' is healthy and if anyone has any advice on managing it, I'm sure there are plenty of people here who need to hear it. Stress is the bane of us all and it's odd, because it genuinely brings no value, yet we all succumb to it.

Consistently bad habits...

UnitBuilds — Thu, 24 Sep 2026 09:59:28 +0000

Ever find yourself working when you shouldnt be? I know I do... Infact, I'm supposed to be a jobless bum sleeping all day, yet I'm still waking up at 8am, still working all day, even late into the night???

We as developers love routines. And that's a bad thing... We get into a 'flow' and just keep working, we get into the habit of working the second we wake up, or 'just 1 last thing' towards the end of the day and end up working 12+ hours...

So here's a call to arms for us all. Stop working when it's time to stop working...

Set yourself alarms, just-in-case. If you really want a wakeup call, set alarms... But then start a time tracker. That way end of the week, you can look at your time tracker and see just how much you did more than you were meant to.

Even side-projects and hobby projects, these arent unrestricted timeslots for you to expand till they consume your day! Budget your time, so you can rest, because I know that every single 1 of you are feeling a bit of burnout... I mean how cant you with how fast the world is running, it feels like I aged 10 years this year and it's just been 9 months!

Genuinely take a moment, everyday, to just sit and do nothing. Not scroll on TikTok, not play a game, not read a book, or browse Dev.to, genuinely nothing, dont even monitor your breathing, literally do nothing. I tried that yesterday and it was pretty infuriating how my mind kept trying to keep itself busy. That's genuinely a bad habit, since when did humans evolve and progress, so we end up with more stress and more busy? I thought technology was meant to make life easier, more relaxed, so we can do less in a day and still get more done? I mean look at a sink, sounds stupid, but the alternative is going to fetch water from a well and carry it back in, just to wash your hands? Simple, easy, it saves you time you never even accounted for, yet we take it for granted and waste it replacing what should have been rest time, with getting more busy?

Idk about you, but I barely have a second in a day where I'm not doing something. And it's insanity! Lets calculate:

wake up and wash face. - 15 min - sink instead of well.
make coffee - 30 min - kettle instead of fire
make breakfast - 15 min - pre-prepared goods in fridge.

Before you even start your day properly, you already saved an hour! So what do you do with that hour usually? Get some extra sleep? Or do more work?

That's the real messed up part, we split our day in 2, work and sleep, where sleep is our only 'rest'. Sleep isnt rest, it's a bodily mandate. You cant rest when you sleep, you need to be rested to sleep. That's why it's called 'restless sleep', because your mind kept racing and it couldnt sleep. Rest is meant to be when you sit down on a chair (or grass, that's always a nice 1), have a drink and just watch the world go by, like watching ants.

As kids, we all had that experience, nowadays, kids dont even rest... It's always a screen, or school, or homework, or practice, sports, etc. And that's arguably worse than us! A kid that cant rest, becomes a kid that never thinks unless prompted to. That innate curiosity is dying in us all, because we never sit still enough to let it wander, we're far too busy for it. Since when did enjoying live become optional? Did thousands of years of development really lead us to an age, where we finally have the means to live automated lives, yet we argue against it? Or is it that we're inventing new ways to stay busy, to keep us from resting on purpose? Feeling bored, here's a new tax you need to file for. Want something to do, how about maintain another social media account. And of course... TikTok... The bane of our existence, the single most counter intuitive thing ever made... Spend 5 hours, in 30 second sprints, watching people do stupid things just to be seen? When I was growing up, movies kept getting longer and I loved it! Now it seems kids want the opposite, they want 'shortform content'... No wonder AI summarizes everything, because modern man doesnt have the attention span to read a single page!

I cant tell you how to live your life and I wont even bother judging my own. We have a serious problem and it's brain decay from lack of rest and we keep making it worse. If you like, take 10 minutes today, just sit back and do nothing. And dont bother letting me know if you did, that's the whole point of rest, it's for you, you dont need to post about it, or give feedback 'yes, I myself had a very pleasant rest today', if you feel better after resting, start looking at your daily life and find rest where you can. After all, that's what our ancestors developed the world for?

We need to talk...

UnitBuilds — Wed, 23 Sep 2026 23:23:06 +0000

So I've been praising Qoder for a while now. Still is, the most consistent and best external AI platform I've used (and I've used ALOT)... But the past 2 weeks have made me question their motives...

For anyone not up to date, Qoder released a new model option, Sonus. Which is marketed as "SOTA". It cost 3.2x credit rate (while Qwen 3.8 Max was at 0.5x during day, 0.2x non-peak and Kimi K3 at 1.4x). So on release, as a 'give it a try', they offered free credits for Sonus, but it was a trap. 500 credits, at 3.2x rate, meant it lasted all of 2 minutes, it doesnt even write a line of code before it was out. Then 1k credits (half of what you get for 20 usd a month), also gone in 3 minutes, not a single line written... Then 2k credits and still not a single line written and gone in 4-5 minutes...

Now they raised the price to 8x... More than double the price, for a model that has no documentation, no param count, no benchmarks, nothing... Essentially, they could be marketing qwen flash at high thinking and nobody would know any different, because it'd cost you 100 usd just to get a single line of code out of it.

This IMO, is anti-consumer behavior, baiting users to try it, because they know they'll overshoot. Chances are, the model doesnt cost them 6x what max costs them... Let alone 16x! It's absurdly expensive to drain a user's credits the second they refresh a second too late. For perspective, at that pricing, you could likely get more work out of Claude Fable for the price... Yet they charge it on a model that has literally NO information?

This isnt acceptable behavior and while I love Qoder as a platform, this is anti-consumer behavior and I wont let it go unheard. If you use Qoder, please dont use Sonus, dont even 'try' it, because it's marketed as a guaranteed way to drain your credits. Whether that's because they're in the red, or because they named it Sonus (Sonnet), same way they named it Cantus (Claude), with no documentation, to try and market Qwen as the 'smart choice', it's not how you do business and it's honestly an insult to every user. When they marketed Max, they offered free calls, which were way more manageable and lasted 1000x longer. This... This is not it. This isnt marketing, this isnt promoting your new model, this isnt even business, this is daylight robbery on something that brings no value.

If you use Qoder, ignore Sonus and Cantus, use the qwen 3.8 flash promo to it's fullest, because that's the Qoder I know. Good value and fantastic promos! Not this BS marketing 'SOTA' while obliterating users' wallets... I shudder to think what happens when someones on pay as you go and goes to make a cup of coffee when 20 usd worth of credits lasted Sonus at less than half the current price, less than 5 minutes... Gods forbid you go to the bathroom too!

I dont think I've ever seen a clearer AI scam than this... And it's clear that their intention is to force users to use Qwen, that's why they marketed it as such and why they coincidently offer 3.8 flash for free in the meantime! Because they want to show 'for the price of Sonnet, or Claude, you could have Qwen all month long'... If that's the marketing you want to do, then post benchmarks, cost comparisons, genuine statistics, not cost users a month worth in 10 minutes...