So finally, I'm publishing my ES6 book - "Getting Started With ES6".
It covers topics like Classes, Modules, and much more.
And it's absolutely free π€
All that I ask for is to share this if you like it :)
#100DaysOfCode#javascript#DEVCommunity
https://t.co/UsoeBF4ITz
Wander with Alok: BTS #1 π οΈ
How I built the animated homepage.
No 3D models. No R3F (yet).
Just React, SVG, CSS animations, and a little geometry to bring a mountain landscape to life. π§΅π
Wander with Alok: Update #1 πΊοΈ
Worked on the homepage:
β Added the first scene with mountains, clouds & birds.
β Expanded it into 4 different scenes based on the time of day.
A little glimpse ππ
There are 5 birds across 3 flight paths, each with different sizes, flap speeds, and start offsets.
Clouds drift independently using CSS animations. At night, the birds turn pale to remain visible against the dark sky.
Wander with Alok: Update #1 πΊοΈ
Worked on the homepage:
β Added the first scene with mountains, clouds & birds.
β Expanded it into 4 different scenes based on the time of day.
A little glimpse ππ
I love travelling and creating itineraries for the places I visit.
So Iβm starting Wander with Alok β an interactive way to experience itineraries, rather than just reading them.
Going to play around with R3F and share the progress along the way. π
This was just a small local experiment on a dummy project, but I can already see how this could evolve into something much more powerful.
Next, Iβm going to equip GrokBot with Laurenβs pstack and try it out on one of my side projects. Letβs see how it goes!
end.
The moment you stop thinking of agents as βa chatbot that writes codeβ and start thinking of them as teammates with roles, skills, and a verification process, the whole interaction feels very different.
contd.
After implementation, they went back and forth for a bit, reviewing and fixing things.
Finally, Reviewer ran the verification skill I had added, verified the task, and signed off.
I was mostly just watching the conversation happen.
contd.
For now, I ran everything locally on a dummy project.
Initially, I kept them separate. Once Coder finished, it would message Reviewer to take over.
Then I put them into a chat group.
That changed the experience completely.
contd.
It genuinely felt like I was sitting in a Slack group with two teammates.
I gave them one task.
Reviewer asked Coder to first create a plan.
Coder created the plan and tagged Reviewer.
Reviewer pointed out a few things.
Coder revised the plan and then implemented it.
contd.
I tried recreating a small version of Lauren Tanβs agent workflow with GrokBot, and honestly, it felt surreal.
I created two bots:
β Coder β equipped with coding skills
β Code Reviewer β equipped with review + verification skills
contd.
@tpritha03 Ah yes, I think you mean the verification file/step itself being missing. I havenβt handled that case explicitly yet β right now the verification runs after the task and validates the output. Definitely something I should add as a hard failure though.
Watched Lauren Tanβs @poteto talk on building AI agents, including the teamβs work on GrokBot.
Few takeaways:
1. Trust is the bottleneck.
You canβt just spawn 10 agents and expect great results. If you donβt trust one agent, scaling to 10-20 only scales the problems.
contd.
@tpritha03 For me, it was verification. I did some digging into the pstack Lauren created and got a few ideas from there. Iβve now added a verification skill tailored to my work, which the agent uses after a task to verify.
Also, please share your writing β would love to take a look!
The main learning for me:
"AI-native engineering isnβt just about asking AI to write code. Itβs about designing a system where AI can write, verify, and safely ship code."
End.
4. GrokBot was a great example of this.
They built it very quickly with agents, then invested heavily in architecture, constraints and verification so agents could keep shipping without humans constantly watching every line of code.
contd.