DEV Community: Nikhil Sharma

I Used to Deride AI Assistants. Then I Met a Stack of Business Cards.

Nikhil Sharma — Sun, 19 Jul 2026 00:18:39 +0000

I used to deride the idea of an AI assistant from the moment they entered the picture (after seeing all the different *Claw variants).

Why would regular people like me need an assistant? The best use case I had heard was: "Oh! It helps us decide whether I or my partner should drive the kids today!" Solving that sounded like a silly problem for an AI assistant to handle. Again, I didn't know any better because I didn't have that problem. I thought I could handle one-off tasks with just an AI subscription. What else was there?

I only found the answer once I had a specific use case for it. I attended a business event, talked to a dozen people, and collected several business cards. I wanted to send each person a personalized email thanking them and continuing our conversation.

If I were to do this manually, the process would look like this: open the email client, manually type in each email address from the business cards, ensure I typed everything correctly, compose my message, and again, make sure I didn't press "send" prematurely. Just thinking about it felt tedious.

That is when the idea of an assistant started making sense. I fire up my coding agent(not a *Claw still), I take a single photo of all the business cards together. I ask the it to extract the names and email addresses. Then, I ask it to loop through the list and ask me what I want to send to each person. It creates drafts(which I still manually review - can't trust them enough), I say send, and then it sends them all automatically.

It feels exactly like talking to a real assistant: you tell them what you want done, and it gets done without you having to press buttons or navigate a UI. That is exactly what I did; I simply gave it instructions using my voice.

This makes me feel that having an AI assistant is indeed helpful. It might have also been useful to jump on this a little sooner, as I could have bought that Mac Mini at the older, lower price.

Stop hiding behind Code (Comfort Zone)

Nikhil Sharma — Fri, 17 Jul 2026 03:06:05 +0000

I missed a chance to take center stage at a networking event yesterday.

Why? Because I chose my comfort zone over actual preparation.

I was heading to a local business meeting and realized I didn't have any physical business cards. As a developer, my immediate instinct was to solve this with technology. I spent my limited prep time building a custom web page and a scannable QR code.

It felt like a smart, modern solution. It wasn't.

When I got to the event, I realized two things very quickly:

The "Tech" is clunky: Asking someone to pull out their phone, unlock it, open the camera, scan a code, and wait for a page to load is a massive amount of friction.

The "Pitch" was missing: Because I spent my time coding, I hadn’t refined my 30-second introduction. When an opportunity arose to take center stage and speak, I didn't take it because I wasn't ready.

The Takeaway: Prioritize the Human Connection

This experience was a harsh reminder that in business, friction is the enemy of connection. By focusing on the "how" (the technology), I completely neglected the "who" (the people) and the "why" (the value I provide).

As developers, we love to optimize processes, but sometimes we optimize for the wrong things. I built a digital handshake, but I forgot that the most effective networking tool isn't a QR code or a business card-it's the clarity of your message and the confidence to deliver it.

Next time, I'm putting the laptop down. I'll be showing up with a clear pitch, a firm handshake, and the courage to take the stage - no code required.

For everyone trying to build something: You are planting seeds. You don't yell at a sprout for growing too slowly. You nurture it and you accept that some plants might not make it. The focus is on the care (fun), not just the harvest (result).

Nikhil Sharma — Tue, 28 Apr 2026 12:09:30 +0000

Software Developers: Redundant or Resilient?

Nikhil Sharma — Sat, 25 Apr 2026 23:23:14 +0000

In the era before AI-assisted coding, my workflow for any feature followed this pattern:

I would analyze the business problem. Even if the Product Manager spent time on it, I would read the documentation and ask clarifying questions. This established a foundational understanding of the problem.
After some back-and-forth discussion, I would begin planning the implementation. This deepened my understanding of the problem.
I would review the existing codebase to identify established patterns and determine what I could reuse. This strengthened my familiarity with the codebase.
If no existing pattern applied, I would research similar scenarios and evaluate design patterns to find the best fit for the problem. This reinforced my coding practices and potentially uncovered new solutions.
Finally, I would start implementing. As I coded, I would continuously consider improvements and alternative approaches. This increased my familiarity with both the problem and the solution.

After completing this process, I could often recall the implementation details and logic from memory during team discussions. If a bug arose, I could usually deduce its cause without immediately inspecting the code, often because I recognized an edge case I had overlooked during implementation.

Overall, this process helped me learn more, retain more knowledge, and perform more of the work myself. These were actually the most fun parts of the process. Today, I spend much of my time reviewing code. However, reviewing is not the same as writing it. As the saying goes in mathematics, you cannot learn simply by reading a textbook; you must engage with the material and put pen to paper.

Maybe times have changed, and I do not even need to know all those details. But then it makes me wonder: am I redundant in this process?

Some people might point out that you bring in taste and judgment. However, what stops a non-developer from showing these skills? They just have to ask AI for alternatives and pick the best solution based on their understanding.

There are still a few places where AI is not as good, especially where there is any integration, whether it involves hardware devices or multiple systems stitched together. However, this mostly covers missing bridges (i.e., AI cannot click buttons on a hardware device or check multiple systems at once). These tasks are limited. Software Engineers working in novel fields might also not feel redundant, but those people are few and far between.

This makes me lean toward "Redundant" as the answer for most dev jobs today. The only way forward seems to be moving to the next level, i.e., truly being an engineer (working with systems that do not exist yet) instead of being a mechanic or developer (working with known systems).

Claude and I aren't vibing at all

Nikhil Sharma — Thu, 23 Apr 2026 02:48:08 +0000

I really enjoyed working with Opus 4.6 in Copilot, especially for tasks that required a bit of creativity (e.g. it was much better at building a landing page compared to GPT-5.3-Codex), but I never got a Claude subscription. I finally did, at last.

After spending 24 hours with Claude Code in VS Code, I've come to the conclusion that it is focused entirely on "vibe coding". It will absolutely do anything to get you to a working solution, even if it isn't best practice.

The code is pretty sloppy. It hardcodes things it should have created configs or constants for. When I asked it to update some sidebar design, along with a change that impacted a common element used globally, I expected it to respect my structure. I have things neatly organized in __sidebar.scss and __global.css, but instead of using each for its own purpose, it stuffed everything into the sidebar file.

It readily fills in the blanks for API calls. For example, while working with the Calendly API, it invented a couple of concepts that looked legit only until I tested them manually and they failed.

Also, the rate limits are insane; you can't just work off a Pro account. Compare this to the fact that I never hit my rate limits on Codex (not even close) or on Copilot, which I was using exclusively just a couple of months ago. People coding entirely on vibes would ask for even the tiniest of things using AI, so it makes sense to extract as much value from them.

Since the code is sloppy, I can imagine the state of the code in a few months. Even if you wanted to maintain it, you would need to ask AI to handle the smallest of tasks because it would be too tiring to comprehend such code.

I also understand now why people need so many prompt files (skills) and whatnot with Claude. It needs a lot of prodding and prompting and has difficulty working off clear instructions at once.