DEV Community: Sukriti Singh

I Spent a Week on VibeCode Arena. Here Is Everything I Did Not Expect.

Sukriti Singh — Tue, 31 Mar 2026 07:20:37 +0000

Seven days. Blind voting. And a realization I didn't want to admit.

I will be upfront about something.

I almost did not write this.

Not because the week was boring. The opposite, actually. I almost did not write it because some of what I found was uncomfortable to admit, and the comfortable thing would have been to just post a highlight reel and move on.

But highlight reels are useless. So here is the actual account.

How it started

I had been hearing about VibeCode Arena for a bit. A platform by HackerEarth where you watch AI models compete on the same prompt, vote blind before the reveal, and open your results as community challenges.

Sounded interesting. I kept putting it off the way you put off anything that might make you feel like you do not know what you are doing.

Eventually I just opened it.

The first thing you see is the Duels feed. People have submitted prompts, two models have gone head to head, and the community is voting on which output is better. Some of the stuff people are building on here is genuinely creative — retro terminals, interactive tools, little games. I spent longer than I meant to just scrolling through before I actually tried anything myself.

Day One — The blind vote

I ran my first Duel.

The way it works is simple. You write a prompt, two AI models generate simultaneously, you watch both outputs come in side by side, and you vote for the better one before the model names are revealed. The reveal only happens after you commit.

I typed something I genuinely wanted to see — not a test prompt, an actual idea I had been sitting on. Both outputs came in. I read through them. I voted.

Then the model names appeared.

I had voted for the model I never use. The one I had quietly written off months ago was based on nothing more specific than a vague impression I had picked up from other people's opinions.

I did not feel vindicated. I felt slightly embarrassed. Because the implication was obvious — I had been choosing my tools based on reputation and habit, not based on what they actually produced.

I ran three more duels that day. My assumptions were wrong twice.

Day Two — The stick figure brawl

Okay, this one was just fun.

I wanted to try creating my own challenge and I did not want to overthink it. So I typed "design a game with stick figures fighting" and watched what came back.

Two outputs. One was clean and structured but felt more like a diagram than a game. The other had actual chaos to it — Player 1 attack button, Player 2 attack button, health bars, the whole thing. It felt like something you would actually waste two minutes on.

I picked the second one. Codestral-2508.

Again, not a model I had seriously considered before. Again, the blind format forced the honest evaluation.

I opened it as a community challenge — which means anyone can now jump in, take that base, prompt AI to improve it, and submit their version. The leaderboard fills up with different people's takes on the same starting point.

I posted it and waited to see what would happen.

What I felt in that moment was something I did not expect. A kind of genuine curiosity. Not "I built a thing" but "I started something and I have no idea where it goes." That is a different feeling and, honestly, a more interesting one.

Day Three — Looking at other people's challenges

I spent most of day three not building anything. Just looking.

The challenge feed has a range of things people have posted. Some are serious — UI components, accessibility focused tools, form builders. Some are completely unhinged. The range is wide and the leaderboard on each one tells you something about what the community actually values versus what looks impressive at first glance.

What I noticed was how much you learn from watching other people's iterations of the same starting point.

Someone takes a base output and restructures it entirely. Someone else adds a feature you had not thought of. A third person goes in a direction that seems wrong and then you look at their score and have to reconsider your assumptions.

I kept thinking — this is the thing most people are missing when they use AI alone. You get your own blind spots plus the model's blind spots. Open it up and suddenly you have twenty different sets of eyes on the same problem.

Day Four — The guitar fretboard

This one I am still thinking about.

I ran a duel with the prompt "design a guitar with all the key notes represented on strings." Six words. I play a bit of guitar so I was curious what the models would do with something I could actually evaluate properly.

Both outputs produced what they called an Interactive Guitar Fretboard. All twelve notes across the top, six strings, click a note and see its positions highlighted in green.

At first glance, it looked great. Clean. Functional. More than I expected from six words.

Then I actually used it.

The note positions are not accurate. When you click a note, the highlighted positions across the strings do not match where that note actually sits in standard tuning. For a tool that is supposed to teach you the fretboard, that is the most important thing to get right, and it got it wrong.

There is no sound at all. You click, and nothing plays. Silence. For a guitar tool that is not a minor gap, it is half the point missing.

No fret numbers so you cannot tell which fret you are looking at. No fret markers at the positions every guitarist uses to navigate the neck. The neck itself looks slightly too short.

I voted for Codestral-2508 again. Better visual layout, marginally better structure. But both outputs had the same fundamental problems.

I opened it as a challenge with a clear brief — fix the note positions, add fret numbers, add fret markers, and add sound. Here is what is wrong, here is what good looks like, go.

That felt more honest than pretending the output was finished.

Day Five — Something changed about how I prompt

I did not read a guide about this. It just happened.

After four days of carefully evaluating other models' outputs, I noticed I was prompting differently. Less about how to structure things, more about what the thing needed to do and why it needed to do it.

The outputs got better. Not dramatically, but noticeably.

I think what happened is that spending time on the evaluation side — really sitting with outputs and asking what is working and what is not — had quietly changed what I was putting into the prompts. I had internalised something about the gap between what prompts produce and what they should produce, and it was coming out the other side.

That is not something I could have gotten from reading about prompting. It came from doing the evaluation work repeatedly.

Day Six — The uncomfortable one

I ran a duel on a prompt close to something I had actually built myself a few months ago.

I was quietly confident going in. I knew this territory. I had already solved this problem.

One of the outputs came back and handled a specific part of the interaction better than I had.

Not overall better. Not in every dimension. But in one specific way that mattered, it had made a decision I had not made, and the decision was right.

I sat with that for longer than I probably needed to.

The thing I landed on was this. It was not that AI is better. It was that I had been too close to my own previous solution to see its weaknesses clearly. I had built a thing, shipped it, moved on, and calcified around it. Fresh evaluation — even from a model — caught what I had stopped being able to see.

That is uncomfortable. It is also useful.

Day Seven — What I actually took from all of this

A week is not long. I am not going to pretend it transformed everything.

But a few things shifted.

I think about model choice differently now. Not which model is best in general — that question does not have a useful answer. But which model for what kind of task, under what conditions, evaluated by what criteria. That framing is more honest and more practical.

I evaluate AI output more carefully. I have a clearer sense of what I am actually looking at — what is genuinely solid, what just looks good, what will cause problems later. That clarity came from doing the blind evaluations repeatedly, not from reading about them.

I am more comfortable with incompleteness. Both of my challenges — the stick figure brawl and the guitar fretboard — were unfinished in clear ways. Opening them as challenges rather than polishing them to death felt right. The platform treats incompleteness as an invitation rather than a failure. That reframe is quietly useful.

The thing nobody says about vibe coding

Most of the conversation around vibe coding is about speed. Build faster, ship faster, stop writing every line by hand.

That is true and it is also incomplete.

What vibe coding actually demands from you, if you do it properly, is better judgment. You are no longer the one writing every line so you have to be the one who can evaluate every line. You have to know what good looks like. You have to catch what is wrong before it ships. You have to make the calls the AI cannot make because it does not have your context.

VibeCode Arena is the most direct way I have found to build that judgment. Not through tutorials. Through doing it, on real prompts, with real outputs, scored honestly, in a community that is working on the same thing.

One week in I am more calibrated than I was. I am also more aware of how much calibration I still need.

That feels like the right place to be.

Try it here: https://vibecodearena.ai/?page=1&pageSize=10

And if you want to jump into something live, the guitar fretboard challenge is open, the note positions need fixing, the sound needs adding, and the leaderboard is waiting.

Challenge link: https://vibecodearena.ai/share/afc85b04-2031-4b33-8a70-caf14197ac6d

The best thing a week on VibeCode Arena did was make me honest about what I did not know. That is more valuable than any output it produced.

I couldn’t understand code I wrote 2 weeks earlier. Not because it was hard my brain just… stopped engaging. Took me 6 months to figure out why.

Sukriti Singh — Sun, 29 Mar 2026 10:15:14 +0000

Sukriti Singh

Mar 13

My Brain Stopped Working and I Spent 6 Months Figuring Out Why

#discuss #productivity #career #mentalhealth

Comments 1

6 min read

Most developers are using AI wrong. The real skill isn’t writing code faster it’s knowing what to trust, fix, and reject. That’s vibe coding.

Sukriti Singh — Sun, 29 Mar 2026 10:14:49 +0000

Sukriti Singh

Mar 14

Everyone Is Sleeping on Vibe Coding. Here's Why That's a Mistake.

#discuss #ai #career #nocode

Comments 4

3 min read

I asked AI to build a fighting game… now two stick figures are battling it out. It works. It’s chaotic. It’s unfinished. Make it better (or worse).

Sukriti Singh — Sun, 29 Mar 2026 10:14:24 +0000

Sukriti Singh

Mar 22

Two AIs. One Prompt. A Stick Figure Brawl. And Now It's Your Problem.

#discuss #community #programming #ai

Comments 3

3 min read

AI built a guitar from just 6 words… and got the concept right but the details completely wrong. No sound. Wrong notes. Missing frets. Can you fix it?

Sukriti Singh — Sun, 29 Mar 2026 10:14:00 +0000

Sukriti Singh

Mar 28

I Asked Two AIs to Build a Guitar. Now I Want to See What You Do With It.

#discuss #community #ai #devchallenge

Comments 3

5 min read

I Asked Two AIs to Build a Guitar. Now I Want to See What You Do With It.

Sukriti Singh — Sat, 28 Mar 2026 16:09:34 +0000

Six words. Two models. One interactive fretboard. And a wide open challenge.

I typed six words into a duel on VibeCode Arena.

"Design a guitar with all the key notes represented on strings."

That was it. No extra context. No design specs. No colour palette. No instructions about interactivity or layout or what clicking a note should do.

Just six words. And then I watched two AI models interpret them completely on their own.

What came back genuinely surprised me.

What actually happened

Both models built something called an Interactive Guitar Fretboard.

All twelve notes laid out across the top. C, C#, D, D#, E, F, F#, G, G#, A, A#, B. Six strings running horizontally. Click a note and green highlights appear showing its positions across the fretboard.

Not what I expected from six words.

But here is where I have to be honest, because this is the part that actually makes it an interesting challenge.

When I clicked A# the interaction worked. A# lit up in multiple positions across all six strings, which is correct — it does appear at different frets on every string in standard tuning. The concept is right and the underlying logic is mostly there.

But a guitarist looking at this would immediately spot what is missing.

No fret numbers. You can see a green dot on a string but you have no idea if that is fret 1 or fret 11. For a tool that is supposed to show you where notes live on the neck, that is a significant gap. A guitarist needs to know "A# is at fret 6 on the low E string" — not just see a highlight floating somewhere on a line.

No sound whatsoever. You click a note and it highlights visually but you hear nothing. For a guitar tool that is a pretty fundamental gap — half the point of learning the fretboard is training your ear alongside your eye. Right now it is a purely visual reference with no audio feedback at all.

No fret markers either. Real guitars have dot markers at frets 3, 5, 7, 9, and 12. Visual landmarks that help you orient yourself on the neck. Without them the fretboard is harder to read than it needs to be.

The neck also looks slightly short. Standard guitars run to at least 12 frets. A few of the higher positions on certain strings might be getting cut off entirely.

So the AI got the concept right. The interaction works. But the output has four clear problems that make it not quite ready for real use.

No sound whatsoever. You click a note and it highlights visually but you hear nothing. For a guitar tool that is a fundamental missing piece — the whole point of learning the fretboard is training your ear and your eye at the same time. Right now it is a silent visual map.

The note positions are not accurate. This is the most important one. When you click a note the highlighted positions across the strings do not always match where that note actually sits in standard tuning. For something that is supposed to teach you the fretboard, getting the positions wrong is a pretty serious problem.

No fret numbers. You can see a green dot on a string but you have no idea if that is fret 1 or fret 11. A guitarist needs to know "A# is at fret 6 on the low E string" — not just see a highlight floating somewhere on a line.

No fret markers. Real guitars have dot markers at frets 3, 5, 7, 9, and 12. Visual landmarks that help you orient yourself on the neck. Without them the fretboard is harder to read than it needs to be.

The neck also looks slightly short. Standard guitars run to at least 12 frets. A few of the higher positions on certain strings might be getting cut off entirely.

Concept right. Execution incomplete. That is exactly what makes it the right kind of open challenge.

Why this prompt specifically is interesting

A guitar fretboard is one of those things that sounds simple until you actually think about what it requires.

You need to understand music theory enough to place notes correctly. You need to think about visual hierarchy — what is a string, what is a fret, what is a note position. You need to handle interactivity in a way that feels intuitive. You need the whole thing to fit in a browser window without feeling cramped.

Six words produced something that handles all of that.

That is not nothing. That is actually impressive. And it opens up a lot of questions about what a better version of this could look like with more iteration, more perspective, more people bringing different ideas to the same starting point.

Which is exactly why I made it a challenge.

The challenge is live and the fretboard is yours to improve

I have opened this up on VibeCode Arena. Anyone can jump in, take the base fretboard output, direct AI to improve it, and submit their version.

Think about what this could become with the right iteration:

Sound — click a note and actually hear it play. The fretboard goes from visual to musical instantly.

Chord diagrams — select a chord and see all the finger positions light up simultaneously.

Scale highlighter — toggle a scale and watch the relevant notes across all strings illuminate.

Dark mode — because every developer tool looks better in dark mode. This is not negotiable.

Visual redesign — make it look like an actual guitar instead of a grid. Wood grain, tuning pegs, the whole thing.

Practice mode — the fretboard calls out a note and you have to find it. Simple quiz mechanic but immediately turns it into a learning tool.

Any one of these takes the base output from functional to genuinely impressive. And you do not need to implement it yourself — you direct AI to do it and the platform scores what comes back.

Come build the version I could not build alone.

Join the challenge here: https://vibecodearena.ai/share/afc85b04-2031-4b33-8a70-caf14197ac6d

What the duel format taught me about this specific prompt

I have been using AI tools for a while now. I have a set of models I default to, a way of prompting I have settled into, a rough mental model of what each tool handles well.

The blind duel broke all of that in a single session again.

Codestral-2508 handled the music theory layout with real precision. The note positions were accurate. The string spacing felt right. I would not have reached for Codestral for something like this without the blind format forcing me to evaluate it on merit.

That is the pattern I keep finding on VibeCode Arena. My default models are not always the right models. My habits are not always serving me. The blind vote is a fast, low stakes way to find out where your assumptions are wrong.

One duel per day on a prompt you actually care about is probably one of the highest value things you can do for your AI judgment right now.

Come make the fretboard better

The challenge is open. The base output is solid. The ceiling is high.

Whether you add sound, redesign the visuals, build a practice mode, or do something I have not thought of — I want to see what you make.

Challenge link: https://vibecodearena.ai/share/afc85b04-2031-4b33-8a70-caf14197ac6d

Six words got us here. Imagine what the right iteration gets us to.

Two AIs. One Prompt. A Stick Figure Brawl. And Now It's Your Problem.

Sukriti Singh — Sun, 22 Mar 2026 11:38:42 +0000

I started something. I need the internet to finish it.

I have a challenge sitting live on VibeCode Arena right now.

Two stick figures. Attack buttons. Health bars. A game that is functional, slightly chaotic, and absolutely begging for someone to come in and make it either significantly better or significantly more ridiculous.

I am genuinely not sure which outcome I want more.

Here is how it got there. And more importantly, here is why I need you in it.

What even is VibeCode Arena

Quick context because not everyone has come across this yet.

VibeCode Arena by HackerEarth is a platform where you watch AI models compete on the same prompt in real time. Two models generate simultaneously. You pick the better output before you know which model produced which. The reveal only happens after you commit.

No brand bias. No habit. Just output vs output.

People are building all kinds of unhinged stuff on here. Retro terminals that type your horoscope like a Linux command. Calculators that gasp when you divide by zero. Pomodoro timers that roast you when you overstay your break. The challenge feed is honestly one of the more entertaining corners of the developer internet right now.

Here is what I built and how it happened

I typed six words: "design a game with stick figures fighting."

Two outputs came in. One was clean and structured but felt like a textbook example. The other felt like an actual game. Player 1 attack button. Player 2 attack button. Health bars. The stick figures had energy. You could feel the chaos of it.

I picked the second one.

The reveal: Codestral-2508.

A model I had never seriously used before. The blind vote forced me to judge on output alone and Codestral absolutely won that round. That alone was worth the session.

But here is where it gets interesting.

After the duel you can make it a challenge

Once you pick your winner on VibeCode Arena you can open that output as a community challenge. Anyone on the platform can jump in, take that base output, direct AI to improve it, and submit their version. Each submission gets scored. A leaderboard builds up.

It is like a collaborative commit history, except the whole community is contributing and nobody is writing the code directly. Everyone is directing AI from different angles, with different ideas, with different things they think would make it better.

I just did exactly that with the stick figure game.

It is live right now. And nobody has touched it yet.

This is where you come in

The game works. It is functional. Player 1 and Player 2 can battle it out and someone wins.

But I know it could be so much more.

Better animations. Sound effects. A proper win screen. Maybe the stick figures get names. Maybe there is a combo system. Maybe someone completely reimagines the attack mechanic. Maybe someone just makes it look absolutely incredible.

I have no idea what someone else will see in this that I missed. That is the part I am genuinely curious about.

So here it is. My challenge. Come make it better, weirder, or both.

Why this is actually worth your time

Beyond the fun of it, doing this on VibeCode Arena is quietly building something useful.

When you run a duel and vote blind you stop being a user of AI and start being an evaluator of it. You develop real pattern recognition about which models handle which kinds of tasks well. That judgment, being able to look at two outputs and know specifically what is better and why, is the actual skill that matters right now.

And when you contribute to a challenge you are practicing the other half of it. Taking a starting point, deciding what it needs, directing AI to get there. That is the loop that is becoming the job.

The stick figure game is a low stakes way to practice both. Which is honestly the best kind of practice.

Come ruin it in the best possible way

The challenge is open. The leaderboard is empty. First mover advantage is real.

Show me what you see in this that I didn't.

Join the challenge: https://vibecodearena.ai/share/7a0a5dac-5ea5-4d90-816b-235de568da59

I started it. Where it goes is up to you now.

Everyone Is Sleeping on Vibe Coding. Here's Why That's a Mistake.

Sukriti Singh — Sat, 14 Mar 2026 05:04:44 +0000

It's not a trend. It's not a shortcut. It's the skill that's quietly separating the developers who get it from the ones who don't.

I'll be honest. Six months ago I thought vibe coding was just a buzzword.

Another catchy term that Twitter would chew through in two weeks and spit out. I kept scrolling past it. Kept telling myself I'd look into it when it actually meant something.

Then I tried it properly. And I understood why I'd been wrong to dismiss it.

What vibe coding actually is

In February 2025, Andrej Karpathy, co-founder of OpenAI, described a new way of building software. One where you work with AI fluidly, intuitively, without obsessing over every line. You direct. You shape. You make the judgment calls. The implementation follows.

That tweet got 5 million views. Collins Dictionary named vibe coding their Word of the Year by end of 2025.

But here's the part that got lost in all the noise.

Vibe coding was never about being lazy or careless. The developers who actually built real things with it were not blindly accepting outputs. They were moving fast and staying sharp. They were using AI as a collaborator while keeping architectural judgment firmly in their own hands.

The ones who got it wrong took "forget the code exists" literally. The ones who got it right understood that the skill had shifted, not disappeared.

Why this matters more than most developers realise

Here is what the companies HackerEarth works with, Amazon, Google, and others, are already building around.

Baseline code is increasingly AI generated. That is not coming. It is here. Garry Tan from Y Combinator disclosed that 25% of their Winter 2025 batch had codebases that were over 95% AI generated.

What that means for you is not that the job is going away. It means the job is changing. The value is no longer just in writing the code. The value is in knowing what to build, how to direct the AI to build it well, and how to evaluate what comes back.

That is vibe coding done properly. And it is genuinely a skill you can develop.

The gap nobody is talking about

Most developers are either fully leaning on AI without really interrogating the output, or rejecting it entirely and calling it hype.

Very few are in the middle, doing the harder thing: actually developing calibrated judgment about which models perform better for which tasks, and why.

That judgment is what separates a developer who is riding this wave from one who is getting swept under it.

And this is exactly what pulled me toward VibeCode Arena, a platform by HackerEarth built specifically around developing this skill.

What VibeCode Arena's Duels actually do

The Duels feature puts two LLMs against each other on the same coding prompt in real time. You watch both outputs come in, then vote for the better one before the model labels are revealed.

No brand bias. No "that looks like a Claude response." Just output vs output, judged on what actually matters.

It sounds simple. It exposes things you did not know you were missing.

When you strip the labels away you realise how much of your evaluation was based on brand familiarity rather than actual output quality. The blind structure forces you to reason from first principles. About structure, about intent, about what the code will actually look like to maintain six months from now.

I ran several duels on prompts similar to work I do professionally. The results were inconsistent across different model pairs and that inconsistency is the whole point. It tells you that which AI to use is a context dependent judgment call, not a brand loyalty decision.

That is vibe coding intuition. And it is a muscle, not a given.

Why you should be building this skill right now

The developers who will thrive in the next few years are not the ones who write the most code. They are not the ones who resist AI either.

They are the ones who have figured out how to direct AI well, evaluate its output honestly, and build things that actually hold up.

VibeCode Arena is one of the few places I have found that trains exactly that, in a live way, community driven, and actually useful.

Free to try. Uncomfortable in the right way.

Try it here: VibeCode Arena on HackerEarth

Vibe coding is not the end of developer skills. It is the beginning of a different kind.

My Brain Stopped Working and I Spent 6 Months Figuring Out Why

Sukriti Singh — Fri, 13 Mar 2026 04:15:14 +0000

This is going to sound dramatic but stick with me

Okay so there was this one evening last year where I was staring at a function I had written myself, like two weeks before, and I could not parse it. Not because it was complex. It wasn't. I just kept reading it and nothing was landing. My eyes would go to the end of the line and my brain would just... not.

I closed my laptop and told my roommate I thought I was getting dumber.

She laughed. I wasn't really joking.

I'd been feeling this for a while, this slow erosion of whatever it is that lets you actually think. I'd start a task, get three minutes in, and find myself on LinkedIn without any memory of deciding to go there. Context switching so much during the day that by evening I felt like I'd worked a full shift but couldn't point to a single thing I'd actually finished. The kind of tiredness that sleep doesn't fix.

The thing is, everyone around me seemed to be doing the same thing and calling it productivity. Our standups were essentially competitive context-switching olympics. "Yeah I'm across six things right now" was a flex. So I assumed this was just what working in tech felt like and I needed to get better at it.

Then I fell into a YouTube rabbit hole about the prefrontal cortex at like 1am (as one does) and something kind of clicked.

The Part of Your Brain You're Quietly Destroying

The prefrontal cortex (PFC) is the chunk of your brain right behind your forehead. It's responsible for what researchers call executive function, which is a fancy term for: focusing on one thing, resisting the urge to do something else, making decisions, planning. Basically all the stuff that makes you good at your job.

Here's the part that got me: it's also the most fragile part. It's the first thing to degrade when you're stressed or sleep deprived. And crucially, and this is the part I kept thinking about, it gets weaker the more you let it lose to impulse. Every time you're mid-task and you check Slack anyway, something in there loses a rep it could have gained.

I don't want to overstate the science here because I'm a software engineer not a neuroscientist and I've probably gotten some of this slightly wrong. But the core thing made sense to me intuitively: focus is a muscle, it atrophies if you don't use it, and most of our work environments are specifically designed to make us not use it.

So I decided to try some stuff. Some of it worked. Some of it was a complete waste of time. Here's what actually happened.

What I Actually Tried

The Pomodoro Thing (Which I'd Been Dismissing for Years)

I know. I know. Every productivity influencer on the internet has talked about this. I had written it off as basic. But I was out of ideas so I tried it. 25 minutes of focused work, 5 minute break, phone in another room.

The first few days were weirdly uncomfortable. Not physically. Just this low-level itch whenever I knew a notification might be waiting. Like knowing there's a text you haven't read and having to sit with it. I'd get to minute 18 and my hand would literally move toward my phone before I caught myself.

But I kept at it and after maybe 10 days or so something shifted. The itch quieted down. I started actually getting into tasks instead of hovering on the surface of them. The best way I can describe it: before, focus felt like trying to hold water in your hands. After a few weeks it started feeling more like a container.

One thing that made it actually work for me: I stopped doing "work for 25 minutes" and started doing "finish this specific thing" within the 25 minutes. Like, "refactor this one component" or "write the tests for this endpoint." The specificity was the difference between actually focusing and just watching a timer while my brain wandered.

Not Checking Anything in the Morning

This one I resisted for weeks because it felt irresponsible.

My old routine was: wake up, check Slack, check email, check GitHub, check Twitter, feel vaguely stressed, make coffee, open laptop already in reactive mode. And I'd spend the rest of the day basically responding to everyone else's urgency while my actual work piled up.

Now I do 60-90 minutes of focused work before I open any of that. Whatever I decided the night before I need to do. That's the first thing. No exceptions.

I genuinely didn't expect this to matter as much as it did. But there's something real about doing your hardest thinking before the noise comes in. By the time I open Slack I've already done the thing I actually needed to do that day, so the rest of the day feels different. Less frantic. Like I'm ahead instead of behind.

The first two weeks were anxiety-inducing. What if someone needed me urgently? What if there was a production issue? In practice, literally nothing bad happened. If something is truly on fire someone will call. Nobody calls.

Meditation, Which I Failed at Twice Before

My previous attempts at meditation lasted about 11 days combined. I hated it. I'd sit there and my brain would immediately start listing everything I had to do and I'd think "this is pointless, I'm bad at this."

The thing I was getting wrong: the point isn't to have no thoughts. The point is to notice when you've drifted and come back. That's the whole exercise. The drifting isn't failure, it's the resistance that makes the rep count.

When I understood that, I stopped being frustrated by it. Wandering and returning is literally the workout. And over time (this took maybe 3-4 weeks before I felt anything) I started noticing the same thing happening during work. I'd catch myself halfway through opening a new tab and just close it. That's new. I'd never been able to do that before.

I do about 10-15 minutes in the morning. Nothing fancy. I'm not enlightened. I still doomscroll. But there's a small but real difference in how quickly I notice I'm doing it.

When I Exercise Matters More Than I Thought

I already worked out semi-regularly so I didn't think this was a lever I could pull. Turns out timing matters.

There's a thing called BDNF (brain-derived neurotrophic factor) that gets released when you exercise. It's essentially a growth signal for neurons, especially in the parts of the brain involved in focus and memory. The boost from it tends to peak a couple hours after you exercise.

So I started working out in the late morning and protecting the afternoon for deep work. This is genuinely the most chaotic thing I've done to my schedule and I'd have laughed at myself for it a year ago. But it's hard to argue with the results. My afternoon sessions are noticeably sharper.

The Stuff That Didn't Work

Nootropics and supplements: I spent probably ₹8,000 on nootropics and supplements over three months. Lion's mane, L-theanine, some ashwagandha thing, a few others. Placebo effect aside, I noticed nothing. Not nothing-except-X. Just nothing. Your sleep, your exercise, your stress levels. Those will do more than any supplement, and they're free.

Going too hard too fast: I also tried going from "barely focusing for 30 minutes" to "4-hour deep work blocks" in like week two. That crashed immediately. The PFC thing is real, you build capacity incrementally. Pushing too fast just made me hate the whole project and I took a week off from trying anything.

Where I Am Now

Six months in, honestly, not transformed. I'm not going to tell you I have laser focus and a calm mind. I still have terrible days. I still open Reddit in the middle of things and feel bad about it.

But the floor is higher. I get into hard problems faster. I can stay inside a complex debugging session without losing the thread as often. There's less of that feeling at the end of the day where you were busy the whole time but nothing got done.

The framing that's stuck with me is this: your PFC is basically the operating system running your thinking. And we spend years upgrading our laptops and our tools and our workflows, but almost no time on that. It's kind of wild when you think about it.

None of this is fast. None of it is magic. But it's real and it stacks.

If you've tried something that worked for you, drop it in the comments, genuinely curious what other people have found.

Honest question for devs: have you ever been beaten by AI on something you thought was your strength?

Sukriti Singh — Fri, 06 Mar 2026 05:09:19 +0000

I just published a piece about competing against LLMs on VibeCode Arena
same prompt, same scoring criteria, no framing advantage.

The AI scored higher than me on accessibility. First pass.
I've been writing accessible UI for years.

Made me realise I was coasting on an assumption rather than actually
verifying where I stand.

Curious for anyone who's done a real side-by-side comparison:
→ What's one area where AI surprised you?
→ One area where you held up better than expected?

Sukriti Singh

Mar 6

I Watched GPT and Claude Fight Over the Same Code. Here's What I Learned.

#ai #career #webdev #programming

Comments 1

4 min read

I Watched GPT and Claude Fight Over the Same Code. Here's What I Learned.

Sukriti Singh — Fri, 06 Mar 2026 04:42:11 +0000

A developer's honest account of watching LLMs go head-to-head on VibeCode Arena — and what it revealed about how I was using AI wrong.

There's a particular kind of clarity that hits you when you stop using AI and start watching it.

A few weeks ago, I logged into a platform called VibeCode Arena a product by HackerEarth and sat back while GPT and Claude went head-to-head on the same coding prompt. Same instructions. Same criteria. Two completely different outputs.

I had to pick a winner without knowing which model produced which.

What happened next genuinely changed how I think about AI. Not in a hype way. In a "I've been doing this wrong" way.

What VibeCode Arena's Duels Actually Are

Before I get into what I found, it's worth explaining what this feature is because it's unlike anything I'd used before.

Duels on VibeCode Arena pit two LLMs against each other in real-time on the same coding prompt. You watch them generate their outputs side by side. Then you vote for the one you think is better before you know which model produced it.

That last part is the whole point. No brand bias. No "oh that looks like a Claude response." Just output vs output, judged on what actually matters: Does it work? Is it clean? Is it something a real developer would be proud of?

The platform is community-driven other developers are watching the same duels, casting the same votes. Over time, the results build a live picture of which models are actually performing better for which kinds of tasks. Not benchmarks. Not marketing claims. Real developer judgment, at scale.

What Happened When I Actually Tried It

I jumped into a duel where the prompt was a mid-complexity UI component, something I'd built variations of professionally, dozens of times.

Two outputs appeared. I read through both carefully.

Model A was fast and clean on the surface. Handled the obvious requirements, structured reasonably well. Model B was slower to read denser, more deliberate. It had made some choices that weren't explicitly asked for in the prompt, but made the component significantly more robust.

I voted for Model B. Felt confident about it.

Then the reveal: Model A was GPT. Model B was Claude.

Here's what surprised me: it wasn't the result that was interesting. It was why I voted the way I did. I had been casually defaulting to GPT in my day-to-day workflow for months, mostly out of habit. The blind vote forced me to actually interrogate the output rather than trust the brand.

I ran three more duels. The results were inconsistent and that inconsistency was the most useful thing I got out of it.

Why Developers Should Pay Attention to This

Most developers I know pick an AI model the way they pick a code editor once, early, and then never revisit the decision.

That's a real problem now, because the models are diverging fast. GPT, Claude, and Gemini are genuinely different in how they approach code. Not just in quality in style, in how they handle ambiguity, in what they prioritise when the prompt isn't precise.

VibeCode Arena's Duels give you a structured way to develop actual intuition about this. Not "I read a benchmark." Real pattern recognition is built from watching models perform on tasks that look like your actual work.

The companies HackerEarth works with, Amazon, Google, and others, are already building on this assumption: AI will write baseline code. What they're hiring for is the judgment layer above it. The ability to look at AI output and know, specifically, what's good, what's shallow, and what's going to cause problems six months from now.

Watching duels is one of the most direct ways to build that judgment. You're not just using AI. You're learning to evaluate it.

Why I'd Tell You to Try It

Not because VibeCode Arena is perfect, it's early, and the platform is actively evolving. But because the habit of evaluating AI output critically rather than accepting it passively is one of the most valuable things a developer can build right now.

Duels force that habit in a low-stakes, weirdly fun way. You vote, you see if you were right, and you start noticing patterns. Over time, you build a real mental model of where each LLM is strong and where it cuts corners.

That's not a soft skill. That's leverage.

Try it here: VibeCode Arena on HackerEarth

If you've tried it or plan to try it after reading this, I'd genuinely like to hear what you found, especially if a model surprised you. Those are the most interesting conversations.

The developers worth betting on right now aren't the ones who use AI the most. They're the ones who understand it well enough to know when to trust it and when not to.