As an IBM Champion, I spend a lot of time staying current on IBM , new research, press releases, product announcements, YouTube explainers from IBM Technology. The problem is that content lives across five different platforms and checking each one manually every day just doesn't scale. So I built a Proof of Concept (POC) to solve it: the “IBM Knowledge Hub” a unified daily digest that automatically pulls IBM articles, press releases, research updates, and YouTube videos into one clean, searchable interface.

The whole thing was built with one AI coding partner: IBM Bob

What Is IBM Bob?

For those who haven't come across it yet, IBM Bob is IBM's AI-powered development partner. Not an autocomplete tool, not a chatbot. Think of it as a skilled developer who sits alongside you, understands what you're building, and can go from a plain-language description to a working piece of software, while explaining every decision it makes along the way. The key difference from other AI tools is that Bob doesn't just produce output. It reasons. You describe what you want, it proposes an approach, builds it, and tells you why. That turns the experience from "AI assistant" into something closer to genuine collaboration.

The Problem I Wanted to Solve

IBM produces exceptional content across multiple channels:

• IBM Newsroom: press releases, corporate announcements, M&A, AI news
• IBM Research: cutting-edge publications and innovation updates
• IBM Technology on YouTube: product explainers and demos.

But as an Innovation Designer working in Client Engineering, I don't have time to manually check all of every morning. I needed a single feed curated, fast, and intelligent.

How Bob Built It Alongside Me

It started with a simple conversation. The first thing I noticed about Bob is that you don't need to know how to code to start. You just describe what you want, the way you'd explain it to a colleague.

I typed:

"I want to build a knowledge hub that aggregates IBM articles and YouTube videos."

Bob asked a few follow-up questions to understand what I needed, then laid out a plan — what the app would be made of, how it would be structured, and why it made each choice. Then it went ahead and built the entire foundation in one go. Over 20 files, all organized, all connected, all ready to build on top of. No blank page. No setup headaches. Just a working starting point.

Bob Explains As It Goes

What sets Bob apart from other AI tools I've used is that it doesn't just produce output, it thinks out loud. When I asked it to set up how the app would pull in IBM news articles, it walked me through its plan in plain English before writing a single line of code. It told me what it was going to build, how it would handle things going wrong, and what the end result would look like. That changes the experience entirely. I wasn't just watching something get built, I understood what was being built and why.

When Things Broke, Bob Fixed Them

No project goes perfectly, and this one was no different. A few real problems came up along the way:

The app couldn't reach IBM's news feeds: there’s a browser security restriction that blocks this kind of direct connection. Bob recognized the issue immediately, explained it in plain terms, and rerouted the connection through a workaround all without me having to understand the technical details.

Some IBM feeds were formatted differently from others: Bob noticed this on its own and quietly built in a way to handle both formats automatically. I didn't even know this was a problem until Bob told me it had already solved it.

Search was slowing the app down: every time someone typed a letter, the app was doing too much work at once. Bob added a small delay so it only searches once the user pauses typing, making the whole experience feel smoother.

In every case, Bob explained what went wrong in plain language before fixing it. I came away understanding the app better each time.

How the Content Actually Gets In

One of the more interesting challenges Bob helped solve was connecting to all the IBM content sources.

For IBM's news and article feeds (IBM Newsroom, IBM Research, etc ), I integrated with IBM RSS feeds essentially subscription channels that publish new content automatically. The tricky part was that IBM runs several different feeds covering different topics: corporate announcements, AI news, research updates, M&A activity. Bob pulled all six together in parallel, so they load at the same time rather than one by one, and normalized everything into a single consistent format so the app doesn't have to treat each source differently.

For IBM Technology's YouTube channel, I connected Bob to the official YouTube Data API, which gives the app access to video titles, thumbnails, descriptions, and publish dates  everything needed to display videos alongside articles in the same feed.

The AI Summary Idea — and How Bob Made It Real

This was the feature I was most excited about, and honestly wasn't sure was achievable. The idea was simple: IBM publishes a lot of content, and most people don't have time to read every article or watch every video in full. What if the app could read them for you and give you a three-sentence summary, just enough to decide whether it's worth your time?

I brought this to Bob as a concept, and it built the entire integration from scratch. I connected to Groq AI API, which runs a large language model capable of reading a piece of content and producing a concise, accurate summary in seconds. Critically, Bob set this up to work for both articles and YouTube videos so whether you're looking at a press release or a ten-minute explainer video, you get the same quick summary at the top.

The summaries generate on demand when you click into a piece of content, and get saved so the same content doesn't need to be summarized again on your next visit. It's one of those features that sounds small but completely changes how useful the app feels day-to-day.

What the App Can Actually Do

By the end, Bob had helped build something genuinely useful:

• One feed for everything IBM: articles, press releases, research updates, and YouTube videos, all in one place
• AI-generated summaries: for both articles and videos, so you can decide in seconds whether something is worth reading or watching in full
• Smart memory: the app remembers what it already loaded, so it opens instantly every time after the first visit
• Search and filter: find anything across all content types in under a second
• Bookmarks: save anything worth coming back to
• Works on any device: phone, tablet, or desktop

How It All Fits Together

Here's the architecture of what Bob helped build, from the IBM content sources at the top, through the app layer, all the way down to how it's deployed and served to users via Vercel.

Every piece you see was built through natural language conversations with Bob. The data sources feed in through the app's API modules, content gets cached locally so the app loads instantly on repeat visits, AI summaries are generated on demand via Groq, and the whole thing auto-deploys to Vercel every time a new version is pushed to GitHub.

From My Laptop to the World — the Deployment Story

Once the app was working locally, I had a clear goal: I didn't want this to live only on my machine. I wanted anyone to be able to open a link and use it, without needing to set anything up themselves.

I described this to Bob, and it walked me through a deployment plan step by step. First, it helped me get the code onto GitHub , the platform developers use to store and share code, explaining each step clearly and making sure everything was organized correctly before we pushed it live. Then it guided me through connecting that GitHub repository to Vercel , a hosting platform that takes your code and makes it accessible as a real website. What made this particularly smooth was that that every time I push an update to GitHub, Vercel automatically rebuilds and republishes the app. No manual steps, no re-uploading files. I make a change, save it, and within a minute the live site reflects it.

What the Collaboration Actually Felt Like

The honest answer: it felt like pairing with a senior developer who has infinite patience and never loses context. I'd describe a feature. Bob would propose an approach, write the code, and flag anything it was uncertain about. When something broke, I'd paste the error, Bob would read the stack trace, identify the root cause, and implement the fix. It kept a mental model of the entire codebase throughout, so changes in one module automatically reflected the right patterns from another.

The result was a production-ready app — ~3,500 lines of code, comprehensive documentation, with full deployment guides.

What I'd Tell Other Bob doesn't just generate code. It reasons about software. That's a different category of tool especially if you execute this on an enterprise level.

#IBM #IBMBob #GenerativeAI #AIAssistedDevelopment #ClientEngineering #IBMCommunity