A step by step guide to building an AI-powered side project

Want to build an AI project at your job? Join my one-day AI Engineer Foundations workshop in January (8 spots left): https://parsity.io/ai-engineer-dayFor the last couple of years I’ve told developers to learn AI engineering, switch jobs and chase the title. That advice skipped the thing that actually worked for me: building an AI project at the job I already had.I’ve done it twice. At my last company, I built a semantic search POC in my first week or two, and it became the product they sell to major record labels. At my current company, a hackathon project using an extraction agent to fill out a complex form is now going to production.In this episode:Why an internal POC beats a tutorial project on your resume and in interviews3 high-value project types: semantic search, extraction agents for complex forms, and support assistants (RAG, chat or voice AI)Why an MCP server is becoming table stakes if your company sells API accessHow to pick a project: where’s the data, who’s feeling the pain, and can you demo it in a week?How to pitch it: lead with the problem it solves and what it costs, not the jargonRunning a hackathon to get the halo effectIn the workshop, you’ll build a RAG pipeline and an agent with tools, human approval, evals and observability, then map it to a use case at your own job. You’ll leave with a working app, or I’ll work with you until it works, or I’ll refund you. Use your company's learning budget.Grab a seat: https://parsity.io/ai-engineer-day

Eric Elliott wrote the JavaScript articles a lot of us learned from. He spent 30 years as a developer, wrote books, mentored thousands of people and worked as an engineering manager at Adobe. Now he says he went from making $500,000 a year to living in his car.This isn’t a video to dunk on him. It’s about what his story means for the rest of us, and what I’m actually doing right now to make myself less replaceable.The market makes no sense right now. Seniors are in demand, juniors can’t get hired, and devs stay unemployed for 28 weeks on average. Here’s how I’d play it: be fast with the tools, have your stories ready, work outside of pure engineering, own the AI layer, and do what other engineers won’t.If you want to read Eric’s story or donate, here it is: https://medium.com/effortless-programming/ai-made-me-homeless-a7266d716e59Want to learn to build AI systems that make you hard to replace? https://parsity.io/ai-dev0:00 Eric Elliott: $500K to living in his car1:26 What this means for you2:00 The market makes no sense3:21 Devs stay unemployed 28 weeks4:26 The riskiest move right now4:47 Who gets laid off first5:24 How interviews changed6:33 Middle managers need to get technical7:57 How to become less replaceable9:11 Just be really, really fast12:07 Work outside of engineering14:17 Own the AI layer15:44 Why LinkedIn still matters17:55 Do what other people don’t19:14 A message to Eric

Everybody is trying to cram AI into their business whether it fits or not. Jeff Kuo, founder and CEO of Ragic, has been building software for over 20 years and has some blunt opinions on where AI actually belongs.We get into why most “AI projects” are really just things companies haven’t automated yet, how Ragic swapped an expensive LLM call for a classifier and cut costs 140x while getting more accurate, the lowest-risk first use of AI for any business, and whether new grads can still become senior engineers without writing code by hand.Jeff also shares how Ragic started as his master’s thesis while he was working inside TSMC, and how he bootstrapped it with zero venture funding.Learn more about Ragic: https://www.ragic.comWant to learn to build production AI systems? https://parsity.io/ai-dev

Being the “React guy” or the “Rust guy” used to be a moat. In 2027, it’s not. In this episode I break down the five bets I’m making on what developers should learn and do to stay employable as AI reshapes the job.We cover:Why going deep on one language or framework no longer gives you leverageBecoming the trusted human who cuts through AI hype inside your companyWhy teaching and explaining tech to non-technical people is a career superpowerAI engineering: agents, RAG, vector databases, MCP servers and tool callingCleaning up AI-generated slop for founders whose vibe-coded apps break at 5 usersThe rise of the internal forward-deployed engineer (and how to do it without the title)My take: stop going deep, go wide. Learn how businesses make money, get outside the LinkedIn and X bubble, and be ready for a world where regular people want to hire you to fix the software they built.Want to become an AI engineer, or need help with your app? Head to parsity.io or book a short call with me here: https://calendly.com/brianjenney83/15-minute-meeting-your-web-dev-roadmap-clone

I took over ownership of a coding bootcamp in 2023, just before being laid off as an engineering manager at Clorox. The last few years have been rough. In between working full time and rebuilding after being laid off, I had to restructure Parsity from the ground up.If you're a developer and worried about what AI might do to your own career, perhaps you'll find this useful and if you're interested in becoming an AI engineer, apply for 1 of 15 spots at https://parsity.io/ai-dev

Become an AI engineer in 45 days: https://parsity.io/ai-devLuke Miller, founder of DCNY, joins the Develop Yourself podcast to share how his 40+ person engineering team uses agentic engineering to build and deploy software at remarkable speed and scale for nearly 200 clients.We talk about what agentic engineering actually looks like in practice, how to maintain quality while moving faster, why engineers need to become more product-minded, and what developers should focus on as AI continues to reshape software careers. Luke also shares his perspective on vibe-coded apps, hiring agentic engineers, and why the most valuable skill may simply be learning how to build things that are genuinely useful.If you’re a developer trying to understand where AI is taking software engineering—and how to adapt without losing yourself in predictions—this episode is for you.

Dallas Davenport is founder of Aivara, an AI automation marketing agency. We met in a Skool group when I offered to look at her app, and I fully expected a vibe-coded dumpster fire (I've dunked on plenty of those). Instead it was complex, it already had users, and it was pretty freaking good.The app is dealership software that replies to customer messages, pulls live inventory off the lot, and automates a bunch of other stuff car dealers deal with. She built it with Claude and she's not a software developer (and doesn't want to be).We talk about how she learned Claude well enough to pull that off, why her site barely mentions AI (dealers just want to sell more cars), how she gets B2B customers without social media, and how she'd vet a developer when she can't read the code herself.

Quite simply, I lay out EXACTLY what I'd build if I were trying to set myself up as an AI engineer. The type of project, the concepts and the tools to learn and why.I genuinely believe that in 2-3 years, the people who learn this stuff will have a massive advantage in the market.You can pick up the project here: https://www.parsity.io/ai-eng-side-project

Want to get into the top 10% of your engineering team by leveraging agentic engineering workflows? I'm planning a 2 day workshop here: https://form.typeform.com/to/EDuBEz4UI shipped a feature with AI in record time.Then I actually read the code.60+ files. Weird abstractions. Tests that technically passed but proved almost nothing. Components rebuilt from scratch for no good reason. What looked like a finished feature was probably 70–80% of the way there.And the worst part? This wasn’t my first time doing this.I’ve also sat in a code review as a senior engineer and realized I couldn’t properly explain code that “I” had just written.I use Claude and Codex constantly. I’m not going back to hand-writing everything. But after using AI to build software at startups, inside my own business, and on large production systems, I think we’re learning the wrong lesson from how fast these tools can code.The question isn’t whether AI can write the code.It’s how much you need to understand before you’re willing to ship it.

I'm opening 12 spots for a weekend workshop in October for devs who want to ship quality software with agents.Sign up here: https://form.typeform.com/to/EDuBEz4U"You're absolutely right, I should have caught that"I'm one of the top 10% of AI users in a 1,000-person engineering org this year. I'm not an AI hater, I ship with these tools daily and I think they're genuinely fun to use. But I've also watched "customers don't care about the code" turn into an excuse for engineers to stop thinking entirely, and that's how you end up with 5,000 lines nobody can explain, a catastrophic bug in prod, and Claude cheerfully saying on the way down.Let's explore ways to maintain quality software without becoming a "meat proxy"