DEV Community: Harsha Kumar

The Habits I Was Sure Would Stick, And What Actually Replaced Them

Harsha Kumar — Mon, 24 Aug 2026 04:48:31 +0000

Somewhere around the second week of building XEdge, I made a content calendar. Real one, days mapped out, topics assigned in advance, the whole disciplined founder thing everyone tells you to do. I was convinced this was the system that would carry the whole thing.

It lasted maybe three weeks before I quietly stopped opening it.

What replaced it wasn't a better calendar. It was just checking the previous day's actual number and reacting honestly to it. Some days that meant a stack post. Some days it meant admitting the number dropped and asking why plainly instead of dressing it up. No plan more than a day ahead, most of the time.

I've noticed this pattern show up more than once since then, and it's worth naming honestly rather than pretending every system that got tried actually worked.

There was a badge outreach idea, emailing tool founders whose products I'd used, offering a small badge in exchange for a link back. Genuinely good idea, still is. Got planned in detail, then sat untouched for weeks behind things that mattered more that day. Eventually got deprioritized entirely rather than half-finished, which in hindsight was the right call, a half-built outreach system helps nobody.

There was a homepage rebuild spec, fully written out, every section planned. Some of it shipped. Most of it didn't, because the actual bottleneck turned out to be a single missing checkout button, not the whole page needing a rebuild. The elaborate plan got replaced by fixing one small, boring, unglamorous thing that had been sitting broken for over a month.

There was an instinct to chase every new platform and every new skill the moment something interesting showed up nearby, a friend's project, a new tool, a new opportunity. Some of that got pursued. A lot of it got named honestly as a distraction and set aside, not because the interest wasn't real, but because it was competing with something already open and unfinished.

The pattern underneath all of these, looking back, is that the systems built in a moment of confidence rarely survive contact with an actual bad day. What survives is whatever's simple enough to still do when the number is low and the motivation isn't there. A content calendar assumes you'll feel like following a plan. Reacting honestly to yesterday's real number doesn't require feeling like anything, it just requires looking at what actually happened and doing the next obvious thing.

None of this is a claim that planning is useless. It's more that the plans worth keeping tend to be the smallest, most boring ones, the ones that don't need motivation to execute, just repetition.

Still figuring most of this out as it happens. That part hasn't changed since week one.

The Stack I'd Build a Personal Brand With, If I Had to Start Over From Zero

Harsha Kumar — Fri, 14 Aug 2026 05:55:43 +0000

There's a version of me from five months ago who had one follower on X, fifteen on Threads, and a half finished idea called XEdge sitting in a browser tab. No audience. No proof anyone would care. Just a stubborn belief that the problem was real, even if nobody knew my name yet.

If that version of me showed up today asking how to build something from absolute zero, here's exactly what I'd hand him, no theory, just the stack that actually got used.

Start with the angle, not the platform

The instinct when you have nothing is to ask what's trending and try to ride it. That's backwards. The better question is what genuinely annoys you about how your space gets talked about, what's the thing you'd argue about at 11pm if someone said something wrong about it in a group chat. That overlap, between what you actually know and what quietly bothers you, is worth more than any trend, because it's the only angle nobody else can copy convincingly.

I used Claude for this specifically, fed it my actual background, what I was building, and asked it to help me articulate what I kept noticing that other AI tool content wasn't saying. The answer wasn't obvious until I saw it written back to me.

Stop writing live, every single day

For the first month I sat down daily with a blank page, which is the fastest way to burn out on content before anyone's even watching. The fix was embarrassingly simple. Ten raw, half formed thoughts, dumped into Claude in one sitting, structured into ten actual post drafts. One hour of work, a week or two of content, ready whenever the day called for it. The posts still needed editing to sound human, but the blank page problem disappeared entirely.

Design it once, then stop touching it

One afternoon in Canva, one banner, one consistent post template, one color palette that matched the actual site. That was it. The instinct to redesign constantly is a form of procrastination wearing a productive costume. Consistency in how things look matters more than any individual design being perfect.

Growth came from replies, not posts

This is the part almost nobody actually does, because it's slower and less satisfying than posting your own thing. Five bigger accounts a day, one genuine comment each, not agreement, an actual take that adds something real to what they said. Their audience sees you. That's free distribution that no algorithm gatekeeps.

The metric that actually mattered

Likes felt good and meant almost nothing. Saves and shares were the real signal, because a save means someone thinks they'll need this again, and a share means they're willing to put their own name next to it. A post with quiet likes and real saves was worth more than a loud one that vanished by the next scroll.

What it actually cost

Zero dollars. The tools were all free tier. The only real cost was time and the discomfort of posting before it felt polished enough, before there was any proof it would work.

What it actually took

Roughly sixty to ninety days of doing this daily before the numbers started meaning anything. There were entire weeks that looked like nothing was happening. A few days with one or two people total. The temptation every single one of those days was to assume the strategy was wrong and go chase something new. It rarely was the strategy. It was usually just consistency that hadn't compounded yet.

If there's one thing worth taking from all of this, it's that the stack was never really the hard part. Claude, Canva, replying to strangers, none of it is complicated. The actual hard part was doing it on the days it felt like nothing was working, especially the days that looked exactly like every other quiet day before the one that finally wasn't.

That's the version of the stack I'd hand my past self. Not because it's clever. Because it's the one that actually happened.

From 0 to a real floor, what actually changed over the last month or so

Harsha Kumar — Tue, 04 Aug 2026 05:30:21 +0000

For a long stretch, XEdge sat around 2 to 3 active users a day, sometimes lower. Real product, real traffic sources, but the daily number stayed stubbornly low no matter what got posted.

Somewhere in the last few weeks that shifted. The floor now sits closer to 6 to 10 on a normal day, sometimes higher, without needing a big spike or a lucky post to get there. Worth being honest about what actually caused that, because it wasn't any single thing, and it definitely wasn't what I expected going in.

The first real unlock was fixing something boring that had been sitting broken for over a month. A checkout path from the product straight to the paid playbooks. Not complicated, just kept losing to whatever felt more urgent that week. Once it finally shipped, product page views picked up almost immediately, not because anything got more exciting, because something that was quietly broken stopped being broken.

The second thing was rotating content formats deliberately instead of chasing whatever spiked last. Stack posts, honest numbers updates, open interactive questions, each one works well on its own, and each one flattens fast if repeated too many days in a row. Once that became a conscious rotation instead of a reaction to yesterday's number, the whole week got steadier instead of swinging between big days and dead ones.

The third thing, and probably the one that mattered most, was replying fast and specifically to anyone who actually engaged. A single comment thread on Peerlist a few days ago, someone pointed out I'd been assuming features would announce themselves once shipped, they don't, and that one exchange reshaped how the whole week's content got planned.

None of this was a growth hack. No new platform, no automation, no clever trick. Just closing a few real gaps that had been sitting open too long, and paying attention to what the daily numbers were actually saying instead of guessing.

Still early, still a small floor in absolute terms, but the shape of the graph is different now than it was a month ago, and that's the part worth writing down before it's forgotten how it happened.

The no-code SaaS stack that actually gets to a paying user, not just a working demo

Harsha Kumar — Fri, 31 Jul 2026 05:29:38 +0000

Most no-code guides stop at "you can build this without code," which is true but incomplete. The harder part isn't building, it's getting someone to pay before the excitement of building it wears off.

Here's the stack in the order that actually matters.

Validation first, using Claude. Describe the idea plainly and ask it to argue against it, five real reasons it could fail. This costs nothing and catches the obvious problems before any time gets spent building around them.

Building, Bubble, free tier. Real database, real authentication, real workflows, genuinely no code required. The free tier is enough to validate demand, no reason to pay anything until real usage shows up.

Adding AI, through Bubble's OpenAI plugin. Roughly twenty minutes of setup, and it's the difference between a simple tool and something that feels like it has real intelligence behind it.

Charging, immediately, through Stripe connected to Bubble. Even a small amount. This is the step most people quietly avoid, waiting until the product feels more finished, more polished, more ready. It almost never actually needs to wait. Paying users behave differently from free ones from day one, and that behavior is the real signal, not analytics.

Finding users, inside communities that already exist around the problem being solved. Solve something real in a real thread first, mention the product only when it's genuinely relevant to what's being discussed.

Total cost across all of this, zero. A live product in a weekend is realistic. A first paying user in two to four weeks is realistic too, but only with daily, unglamorous outreach behind it, not a single launch post and silence after.

Full playbook with every prompt and every tool link at xedge.tech.

What actually shipped this week, and the retention question that started it

Harsha Kumar — Thu, 23 Jul 2026 05:06:34 +0000

A few weeks back I asked a Discord community a direct question. XEdge hit 750+ users organically, then flattened. 61 returning users out of that total. I wanted to know if that was a positioning problem, a UI problem, or something else.

Someone answered with a single question that reframed the whole thing. Is there an incentive to keep using your tool.

That's the actual gap. XEdge answers one question well, what AI stack fits your goal, and once someone gets that answer, there's no reason to open the tab again until they have a new goal. That's not a bug, it's just the nature of what a reference tool does. But it means nothing in the product was giving anyone a reason to return.

A friend who's been in these conversations for a while had already suggested the fix independently, before I even asked the retention question publicly. Real builders, their actual stack, a short honest quote, featured on a page that grows over time. Not a login system, not a notification engine, nothing structurally heavy. Just people and their real answers.

This week that shipped. It's called Builder Stories. First person featured is the friend who suggested it, which felt like the only honest way to open it, since the idea was his.

Two things stood out building it. First, the honest gap between what people say they use and what they actually open daily is bigger than expected. Second, the ask itself has to stay small, one sentence, no pitch, or the yes rate drops fast. Learned that the hard way testing different framings across a few days, an announcement post barely moved the number, a direct challenge asking people to drop their actual stack in the replies did.

Daily numbers this week bounced around more than usual, 7, then 3, then 2, then 8. Tracked cleanly against which angle was used that day, not luck. Repetition of the same framing three days running flattened it. A genuinely different hook, even on the same underlying feature, brought it back.

The honest lesson from the week isn't really about the feature. It's that consistency and honest specificity keep working, every time they're tried, and every time they're abandoned for something flashier, the numbers show it within a day or two.

xedge.tech/builders if you want to see where it landed.

This is what i planned in order of trying to get a spike in my traffic

Harsha Kumar — Sat, 18 Jul 2026 10:02:42 +0000

i've added a new feature section in my site xedge.tech where i show case builders and their opinion about tools like what tools they , why only that tool and many questions that begginers wanted answers for and we already got 2 builders in the showcase and they are highly experienced

this can help people to decide which tool to use for their project and also it clears up the confusion . the 2 guys i asked questions about they are experienced they are programming before the ai enhancements were added so their advice might help a lot and I'm gonna find more builders and tech enthusiasts to get more info about ai tools in their own perspective and including why they pay for that specic tool and a lot . so i want you guys to be stay tuned and check xedge daily about new builder stories and support it .

Why I built a page with no immediate growth upside, and shipped it anyway

Harsha Kumar — Thu, 16 Jul 2026 10:05:42 +0000

XEdge has a discovery problem that's already solved. Describe a goal, get a stack, done. What it didn't have was a reason for anyone to come back after that first answer.
A friend pointed this out plainly a few weeks ago, and independently, someone in a community I'm active in asked almost the same question from a different angle, is there any incentive to keep using this. Two people, no coordination, same gap named.
The fix didn't need to be complicated. Real builders, their actual AI stack, one short quote, a photo, a permanent page that grows over time. No login system, no notification engine, no new infrastructure.
It's live now. First person featured is the friend who suggested it, which felt like the only honest way to open something built on someone else's idea.
xedge.tech/builders, and if you want to be one of the next few featured, this post is an open invitation.

The newsletter stack that gets you to your first $50 — no audience required

Harsha Kumar — Wed, 15 Jul 2026 08:47:06 +0000

Most "start a newsletter" advice skips the order of operations. Here's the exact one, five tools, in sequence:
Perplexity for niche validation — search actual complaints in your space, not trend reports. Claude for the first five issues, written in one batch sitting, not live day-by-day. Beehiiv free tier for the platform — zero cost to start. One insight post daily for growth, subscribe link in the first comment so the post itself reads as value, not a pitch. First $50 from one manually-sold sponsored slot, before any subscriber threshold, because the first sale validates the model faster than waiting for scale.
$0 total cost. First subscriber same day you start.
Full playbook with every prompt used at xedge.tech.

What actually matters when someone asks you to "build a chatbot" for their business

Harsha Kumar — Sat, 11 Jul 2026 07:52:00 +0000

Got a real lead this week — small business owner wanting an AI chatbot for his site. Here's the honest breakdown of what that actually requires, since "chatbot" means five different things depending on who's asking.
Scope before building. Most small businesses don't need a custom-trained bot. They need their top 10-20 repeated customer questions automated well. Find that out first.
Platform over custom build. Tidio, Crisp, Intercom's lower tiers — pre-built AI chat widgets connect to a knowledge base fast. Custom builds are rarely worth it below a certain scale.
The bot is only as good as its input. Feed it real past customer conversations, not generic FAQ copy. Specificity is what makes it feel useful instead of annoying.
Always build the human handoff. A bot that traps a frustrated customer in a loop does more damage than having no bot. This is non-negotiable.
Test before launch. Twenty real customer-style questions, see what breaks, fix it before it's client-facing.
Full customer support AI stack at xedge.tech.

The no-code SaaS stack that actually gets you to a paying user

Harsha Kumar — Thu, 09 Jul 2026 05:32:28 +0000

No-code SaaS consistently surprises founders because the barrier to a real, working product is genuinely gone now. Here's the stack:
Validation: Claude — before touching a builder, describe the idea and ask Claude to steelman the 5 biggest reasons it fails. Cheapest possible filter.
Building: Bubble.io — real database, real auth, real workflows, no code. Free tier is enough to validate and launch a basic version.
AI features: OpenAI API via Bubble plugin — the unlock most no-code founders skip. Connects in 20 minutes, instantly makes the product feel more capable than it looks.
Payments: Stripe + Bubble — charge from day one, even a token amount. Paying users tell you the truth; free users tell you nothing.
Distribution: show up where your users already are — solve a real problem in a real thread first, link second.
Total cost: $0. Time to live product: one weekend. Time to first paying user: 2-4 weeks with real effort.
Full playbook at xedge.tech.

this is how I started Xedge from 0

Harsha Kumar — Sun, 05 Jul 2026 03:53:32 +0000

I'm Harsha and i'm from india , once i'm in college i was obsessed with this ai tool market when i tried to use them so many questions ,

which tool is better ?

-does hype really matches that output ?

is there any underrated tool that i'm missing

i just become from obsessed to overwhelmed and then i got an idea like what if there is a guide to tell me , but when i clearly did research there are already a bunch of tools directories out there like the futurepedia , there is an ai for that , etc . i noticed that all they do is like dumb a ton of tools and describe about them . that doesn't make any sense to me . i asked myself " is there a better way of organisation rather than dumping tools ?". then i thought about an ai ecosystem , where they'll get ai tools , stacks , professional prompts , automations , templetes , agents , api's , and i even imagined a interface within it where they can build their custom software by using open sources . they are not only going to buy the templetes from others they can build and sell them too . simply a places where people can access everything related to ai at one place where you can buy , sell and create custom for your own .

i think about it like an ai paradise . but at that point i really don't know how to code , i only know the c language . then just did some research got the elite sources to learn html , css and js i got the documentations in 1 tab and the videos in another and opened vs code for practice . in less than a month i've learned css , html , java script , python and MYSQL .

Then I heard about ai open source high end models writing code for free . Im like " damm i have to use this free 1 time usage " i explained almost everything related to xedge i imagined about and it just gave me a single vibe coded home page , i'm very dissapointed at that point because i thought it might build a whole website for me . then my inner voice kicked in and told me " you have the skills , just do it " , i used that same colour pallete generated by ai and created like 4 pages and learned a lot of stuff in that building phase , and it became very easy but time taking process , then i get to know about claude , gpt , and a bunch of other tools .

I used them to write code faster and after it done doing things i'll check all the code based on my knowledge and do corrections if i want , that is how i built the my ideas first prototype .

then i increased my connections mastered marketing certainly the organic marketing and the SEO game , i did a lot of research and learned a lot of stuff , the on i've been grinding on gaining knowledge among tech side and distribution side while building the product , i put all my efforts and now i have over 600+ users in my platform and i spent almost 0$ ( except for the domain ) .

even though i havent made a single penny from my skills or from my product , I'm really proud to myself for doing all this in less than 3 months and starting from 0 and spending 0 . And i believe i could really get my xedge to its final phase and gnerate the real revenue ,

How to validate any startup idea in 48 hours with $0

Harsha Kumar — Thu, 02 Jul 2026 08:18:10 +0000

Most founders skip validation because it feels like delaying the real work.
Validation IS the real work. Building the wrong thing for 6 months is the delay.
Here's the exact process:
Find real complaints: Perplexity
Search for people already expressing the problem you're solving — Reddit, review sites, forums. Copy the exact language they use. That's your marketing copy and your proof of demand in one step.
Size the market honestly: Claude
Describe your idea and ask Claude to calculate TAM/SAM/SOM using three different methods — top-down from industry reports, bottom-up from customer count, and value theory. If the three methods don't roughly converge, you don't understand your market well enough yet.
Map your real competitors: Claude + Google
Not just direct competitors — what do people do instead of your solution right now? That "instead of" answer is your real competition, and it's almost always more honest than a competitor matrix.
Test demand without building: Carrd
Free one-page landing page with a waitlist signup. If 20 strangers won't give you their email for something free, the idea has a messaging problem at minimum and a demand problem at worst.
Talk to 5 real people
Not friends. Not family. Strangers who actually have the problem. One conversation where someone says "I've been looking for exactly this" is worth more than any analytics dashboard.
Total cost: $0
Time investment: 48 hours
Time saved if the idea is wrong: months of your life
Full validation playbook with every prompt and template at xedge.tech
What idea are you sitting on right now that you haven't validated yet?