Bryan Collins

Watch my tutorials on and builds YouTube youtube.com/@bryanjcollins

My AI trading agent ain't afraid of bad news.

I coded some easter eggs into my ai trading agent

My AI trading agent finds profitable rolls

Once a day, my AI trading agent shows me this

My AI trading agent is fun to use

Ended up moving from Kit to Resend. End of an era

"I worry about many things with AI, copyright issues, having my work stolen...." Join the queue. The short answer? If you're worried about copyright, don't put anything into an AI model. In 𝘌𝘮𝘱𝘪𝘳𝘦 𝘰𝘧 𝘈𝘐, Karen Hao explains AI developers see a model training on content as NO different to an artist getting inspired by their peers. (I've yet to meet the creative who consumed their entire genre and went on to build a multi-billion-dollar company.) Still, playing ostrich won't help. Some workarounds exist if you run a website: ↳ Add "User-agent: GPTBot" and "Disallow: /" to your robots.txt file to block OpenAI's crawler ↳ Add <meta name="robots" content="noai, noimageai"> to the header of every page (Raptive, the ad network I use, recommends this) ↳ Use Yoast or Rank Math to add both to WordPress Or just ask ChatGPT how to do it! But blocking AI bots only protects new content. If your posts are more than a year or two old, OpenAI has probably trained on them. Anything on social media is gobbled up too. And OpenAI says nothing about content on the Wayback Machine.

"I want an AI prompt that can help me develop lecture notes." A newsletter reader asked me that exact question. Academia is wary of AI. I'd wager some lecturers struggle to tell AI essays from human ones. Still, warning students off AI is like an early 2000s lecturer telling their class to avoid the web. Well-meaning, but impractical. Try this workflow instead: 1. Open Google NotebookLM 2. Create a new notebook 3. Upload your lecture notes, PDFs and links 4. Ask it for an Audio Overview 5. Download the file Boom. Instant personal podcast. Two hosts go back and forth on the key topics of your course. The same notebook will build a mind map and a written overview of your lectures too.

The accents and intonations inside of Eleven Labs are alarmingly good. A few years ago, I spent 2k on a narrator for one of my audiobooks. It didn't quite work. He had a few verbal tics that didn't fit the book's tone. So for the next one, I booked out a studio and narrated it myself. It took me a week or two. And that was before editing and re-recordings. These days, I'm wondering if an AI voice could do just as good a job. The accents and intonation in ElevenLabs are alarmingly good. I'm not sure I'd book the studio again.

I built a massive library of biz ideas with Claude Code and Argon. The whole thing updates on autopilot a few times a week. Now, ideas are easy. Execution is still hard, though.

Argon is crazy good but also takes a long time vs opus 5.5

I redesigned the sign up funnel on tenderwatch.ie with Gemini Argon and the Mobbin MCP. Mobbin has around 600,000 real app screens. Argon used them to give me three design concepts for every stage of the funnel. Now a visitor signs in with a social account to unblur the rest of a page. Full build: youtu.be/jRZ51EWbLrU

I love it when a new AI model takes apart an older AI model's work. It reminds me of the days when I managed agencies. They'd swoop in and tell us how the last team got it all wrong... and they have a MUCH BETTER idea.

There is no moat or vendor lock-in with these AI models. Sure, they try, but today's frontier is tomorrow's legacy model. I was easily able to port my AI memories from one agent to the next. No wonder Claude and Gemini offer such steep discounts on the latest models.

I asked Claude Fable for 101 business ideas I could build with AI coding. It scored every one against the VRIO framework and put them in an interactive dashboard. The surprise: it told me to ship the best ones on a site I already own, not start new ones. It cost half my weekly Fable tokens. youtu.be/ZZYqWTwVp1w

I built an options trading agent a while ago. It's my own Bloomberg terminal (Yes, it runs in terminal!) Any questions for a YouTube deep dive?

I usually tell AI to ask me questions in a picker so I can steer agents before they build I created a skill called /picker for that

I built Current to help consultants track Irish solar and battery projects. It uses real data, parses official planning records and provides watch lists to consultants. Now, should I launch it?

I built a roadmap engine with Opus 5.5 and Agon. My AI agents get 3 inputs: → DataForSEO MCP for keyword demand → Mobbin MCP for real UX patterns → My GitHub for what’s already built Once a week, it fills a dashboard with build ideas backed by data and working examples, not just AI “insights.” No more staring at a blinking terminal cursor.

I one-shotted this build with GPT-6 Astra It took about an hour. The project uses APIs to help Irish energy and planning consultants find Irish solar and battery projects I dunno if I'll actually build it out, but Astra is lit.