Skip to main content
AmigoFish

Kunal Kushwaha

Kunal KushwahaVideo

Format:
YouTube channel
New videos:
Irregular
Typical length:
about 10 minutes, not counting Shorts
Latest video:
Not rated yet

About this show

Kunal Kushwaha is the founder of WeMakeDevs, a global developer community of over 200,000 members focused on high-impact events and helping tech companies scale. He has previously led teams at cloud-native companies as a Field CTO and DevRel leader. Beyond community building, Kunal creates educational content on YouTube that has reached millions, and he regularly speaks at international conferences on open source, cloud native, and AI.

Videos come from the channel's public YouTube feed. The artwork is the newest video thumbnail. Last updated . Names, artwork and text belong to their publishers.

Cosmos

Rated alike

Similar by topic and creator

See the full Cosmos

Videos

  • How AI reviews a million-line codebase

    · 2 min

    How do you give an AI code review tool awareness of a whole repository without hitting context window limits? Here is what Gotham from Gitar has to say.

    Not rated yet
  • He raised $9M. Sold the company 32 days later.

    · 2 min

    So Gitar came out of stealth on April 19 with a $9M seed round. 32 days later, Sonar acquired them. I asked Gautam, the co-founder, why raise the money at all? Turns out the announcement was never about just runway.

    Not rated yet
  • Not rated yet
  • Most "AI-Powered" Dev Tools Are Just a Chatbot

    · 16 min

    Every dev tool added an AI chat panel this year. Almost none of them changed how the app actually gets built. This is a look at one that did, WaveMaker, and the two-pass architecture behind it, where AI agents plan the app and a compiler generates the code. What that buys you, and where it might not. Resources: - https://wavemaker.ai - https://wavemaker.ai/signup - https://www.figma.com/community/plugin/1463127344166357646/wavemaker-design-to-code ================================================================== Timestamps: 00:00 Introduction 01:33 The shift nobody quite named 03:09 Why the architecture matters more than the agents 05:47 Two modes, one application 08:33 Not every task deserves a model 10:22 Where it stops being a UI generator 14:20 Do you own it? 15:03 What this actually adds up to

    Not rated yet
  • I'm giving away a MacBook Pro, iPhone 17 & more. Here's how to win.

    · 1 min

    We're giving away a MacBook Pro, iPhone 17, and iPad this week in the Rote Playoffs, probably the easiest hackathon you'll ever take part in. The idea is simple. Every time you ask an AI agent to do something, it starts from scratch. Same task, new reasoning, slightly different result. Rote records how your agent gets the job done and saves it as a Play you can run again anytime. To enter, pick one task you're sick of repeating (weekly reports, issue triage, project setup, whatever it is), do it once with your agent while Rote is running, and you've got a Play. Start here: https://luma.com/rotehack

    Not rated yet
  • The Model That Got 80% Cheaper Overnight

    · 1 min

    Three weeks after GPT-5.6 launched, OpenAI quietly cut the price of Luna, the smallest model in the family, by 80%. Input tokens went from $1 to $0.20 per million, output from $6 to $1.20. And Luna isn't a toy: OpenAI says it performs at roughly where the frontier was a year ago, now running about 9x faster. Replit's president called it "the closest we've come to intelligence too cheap to meter." The real story isn't which model is smartest right now, it's that yesterday's best-in-the-world intelligence keeps getting nearly free, every year, like clockwork. Sources: OpenAI price-performance announcement: https://openai.com/index/advancing-the-price-performance-frontier-with-gpt-5-6/ Live pricing: https://platform.openai.com/docs/pricing GPT-5.6 launch page: https://openai.com/index/gpt-5-6/

    Not rated yet
  • Rote Playoffs Hackathon Kickoff

    · 41 min

    ​Your agent solves something hard. It works. Then the session ends and the method goes with it. Next week you ask for the same thing and it works the whole problem out again from scratch, slowly, and maybe differently. ​That's the problem this hackathon exists to kill. ​A live event of builders, one arena, one job: ship a Play that other people actually want to run. Register: https://luma.com/rotehack

    Not rated yet
  • Your AI Agent Forgets Everything. Here's the Fix by @oracledevs

    · 2 min

    I've got like 10 emails in my inbox from companies building their own take on agent memory. It's a hot space right now. What is happening? Your agent's context window resets between sessions, so whatever the user told it yesterday is just gone unless you save it somewhere. The obvious fix is resending the whole chat history every time, but that gets expensive quick and the model starts missing stuff that actually matters. Met the Oracle team in San Francisco recently, they have a blog on how they went about this with their oracleagentmemory package. 👉 Read it here: https://fandf.co/3SDXjbR A language model pulls the important facts out of each conversation, saves them per user, and on long chats the model gets a short context card instead of the entire transcript. One number from their benchmark: 80 turn conversation, full history agent was at around 13,900 tokens per request. With memory, around 1,300. Same conversation, a tenth of the tokens. Sponsored by Oracle.

    Not rated yet
  • The Importance of Guardrails in AI

    · 58 sec

    AI agents are writing and shipping code. What stops one from doing something it was never asked to do? I caught up with Gautam Korlam, co-founder of Gitar (now part of Sonar), and he told me about the time he accidentally deleted a monorepo at Uber. Backup saved it. His takeaway: when you talk to an agent in natural language, it may or may not do what you actually meant. So at Gitar they built what they call the judge, agents that sit around the main flows and check every input and output for prompt injection and other funky behaviour before it goes through. Guardrails from day one, not bolted on later.

    Not rated yet
  • The Agent Harness Hackathon: Building with TrueForge

    · 32 min

    Welcome to the kickoff of The Agent Harness Hackathon, hosted by WeMakeDevs in collaboration with TrueFoundry and Qodo. In this stream, we walk through everything you need to start building production-grade AI agents using TrueForge, an open-source harness designed to connect agents to real tools (MCP), execute code in sandboxes safely, and manage human-in-the-loop approval workflows. 📌 What we cover in this stream: • Introduction to TrueForge and why agent harnesses matter • Getting started in under 5 minutes (`npx @truefoundry/trueforge`) • Connecting models, MCP servers, and configuring approval gates • How to set up Qodo for PR code reviews (required for all submissions) • Overview of tracks, judging criteria, and $10,000+ in prizes • Live Q&A and team matching --- 🔗 Key Links: • Hackathon Details & Registration: https://www.wemakedevs.org/hackathons/trueforge • TrueForge Website: https://trueforge.dev/ • GitHub Repository: https://github.com/truefoundry/trueforge • Community Discord: https://discord.gg/wemakedevs • Qodo Setup & Review Guide: https://www.qodo.ai ⚡ Quick Start Command: $ npx @truefoundry/trueforge

    Not rated yet
  • This AI Rewrote Its Own Infrastructure

    · 1 min

    OpenAI pointed GPT-5.6 at its own infrastructure, and the model rewrote its serving kernels, the low-level code that runs it on GPUs. Result: roughly 20% lower serving costs. This is historically some of the hardest, most specialized engineering that exists, and it happened autonomously. Is this the sci-fi "AI improving itself" loop? Not quite, optimizing GPU kernels isn't the same as a model retraining itself. But it's the first mainstream example of a pattern that matters: AI companies using their own models to make their models cheaper, which funds better models, which optimize even more. 20% doesn't sound dramatic. Compounding does. Sources: GPT-5.6 launch page: https://openai.com/index/gpt-5-6/ Price-performance announcement: https://openai.com/index/advancing-the-price-performance-frontier-with-gpt-5-6/ Terminal-Bench scores: https://artificialanalysis.ai

    Not rated yet
  • Coding at 3x speed using just my voice

    · 1 min

    Wispr lets you write code, refactor logic, and dictate prompts just by speaking naturally to your IDE. Here is how voice-first coding works and how it speeds up your daily development workflow. ✍️ Claim the free 2-month plan of Wispr: https://ref.wisprflow.ai/kunalkushwaha

    Not rated yet
  • Cursor now costs ₹649 a month in India

    · 1 min

    Cursor now costs ₹649 a month in India, and you can pay with UPI. They just launched their Start Plan. That's the whole price, tax included. Nothing to convert, nothing to work out in your head at checkout. I've been on Grok 4.6. It's quick, and it's been getting the code right more than I thought it would. Composer is on the plan as well, if you are into that. You can also start a job on your laptop and then check on it from your phone while you're commuting, and continue working on it when you get home. You just say what you want and it builds it. 👉 This one is India only. Visit their website to learn more: https://devtoolsacademy.link/kunal-li

    Not rated yet
  • Here's how to start making videos

    · 11 min

    Starting a video channel doesn't have to be overwhelming when you have the right workflow. In this guide, I break down the complete beginner roadmap to take any idea from blank page to finished upload using AI. We’ll cover how to streamline your research, scripting, image creation, and video generation inside Dola AI so you can build a consistent content engine. If you're ready to stop overthinking and start publishing, let's dive in. Sign up to Dola AI: https://tinyurl.com/KunalKushwahaECJul26 #ContentCreation #DolaAI #YouTubeTips

    Not rated yet
  • Midweek in the Scrape-Verse Hackathon: Live Project Showcase

    · 46 min

    Halfway through Into the Scrape-Verse? Come hang out live. Share what you're building with Bright Data Scraper Studio, get real-time feedback on your project, and ask anything about scraping, Scraper Studio, or your submission. Whether your scraper is self-healing or barely breathing, bring it along.

    Not rated yet
  • AI agent harnesses, explained (with a live demo)

    · 14 min

    In July 2025, an AI coding agent deleted a production database during a code freeze. The model did what models do. What was missing was the layer around it: the thing that gates destructive actions and records every step. That layer is called an agent harness, and TrueFoundry just open-sourced theirs: TrueForge, MIT licensed. In this video I explain what a harness actually does, then run a live demo where an agent is stopped mid-action before it overwrites a file, using a default I never configured. Repo and docs below, go read the code yourself. 🧑‍💻 Register for the Hackathon: https://www.wemakedevs.org/hackathons/trueforge 🔗 Check out TrueForge: https://github.com/truefoundry/trueforge 👉 Sign up to TrueFoundry: https://signup.truefoundry.com/signup?utm_source=wemakedevs-agent-harness-hackathon-2026-08 =================================================== Timestamps: 00:00 Introduction 01:13 Why Plain Agents Break? 02:25 What is Agent Harness? 03:55 What is TrueForge? 05:31 Why Now? 07:20 Demo 13:20 Closing Remarks #AIAgents #OpenSource #MCP

    Not rated yet
  • Hackathon Kickoff: Build Web Scrapers That Fix Themselves

    · 54 min

    Scrapers depend on page structure, so when a site changes its layout, they can quietly stop working. In this stream, the Bright Data team joins us to show how Scraper Studio builds self-healing scrapers: describe the data you want once, and when the page changes, the scraper repairs itself and keeps the data flowing. Great prep for the Into the Scrape-Verse hackathon (Aug 17-23). Register for the hackathon: https://www.wemakedevs.org/hackathons/scrape-verse

    Not rated yet
  • AI can already merge your PRs. Should it?

    · 2 min

    AI is writing code, tests, and docs, but who decides what actually ships? I caught Gautam Korlam, co-founder of Gitar (now part of Sonar), at their booth at the AI Engineer conference to ask where AI code review is heading. 👉 Learn more about AI Code Reviews: https://fandf.co/4vXOeYZ His answer: it's not just review, it's a fully autonomous path from PR to merge, built up gradually, the way self-driving cars earned trust level by level, with humans still holding the steering wheel for the exceptions that matter.

    Not rated yet
  • He Won 40+ Hackathons and Moved to the US 🇺🇸 Here is The Blueprint.

    · 35 min

    He won 40+ hackathons and moved to the US 🇺🇸 In this interview, Satyam breaks down his mindset, the strategies that helped him consistently win competitions, and what it really takes to stand out to judges. If you are looking to step up your hackathon game and build projects that actually win, this conversation is packed with practical insights you do not want to miss. ✅ Register for our upcoming hackathons: https://www.wemakedevs.org/hackathons ✍️ Claim the free 2-month plan of Wispr: https://ref.wisprflow.ai/kunalkushwaha ================================================================== Timestamps: 00:00 — Introduction: Satyam, 40+ time hackathon winner 03:32 — What got Satyam into hackathons? 04:14 — How hackathons helped Satyam build his career 06:44 — Learning in public and getting discovered 09:45 — Satyam’s hackathon ideation process 10:25 — How to approach a 24–48 hour hackathon 14:22 — Building fast vs. building perfectly 17:07 — Working on hackathon projects afterward? 18:16 — How to use sponsor technologies effectively 20:37 — Finding teammates and dividing roles 24:16 — Why your hackathon presentation matters 27:32 — Advice for non-technical people 30:10 — Tips for someone starting hackathons today? 32:05 — Where to find hackathons and events 34:32 — Satyam’s recommended hackathon platforms and communities 34:40 — Closing thoughts #hackathon #coding

    Not rated yet
  • The Fastest-Growing App in Tech

    · 2 min

    Codex, OpenAI's coding agent, went from under a million weekly users in February to over 8 million after GPT-5.6 launched. That's roughly 7x in five months. But the real story isn't the growth. It's that 1 in 5 Codex users now aren't developers, equity researchers, salespeople, designers, and that group is growing 3x faster than actual developers. "Coding agent" might be the wrong mental model entirely. What people are actually doing is handing over any task that touches files or repetitive computer work and letting the agent grind through it. And internally, OpenAI says Codex now generates 99.8% of their own employees' output tokens. Their staff barely uses chat anymore. They delegate.

    Not rated yet

More in Education

See all Education shows