YOUTUBE DESCRIPTION You "use AI" every day. But be honest with yourself — you're mostly just replacing Google searches with slightly longer questions to ChatGPT. You're using a Ferrari to check the mailbox. In this video I break down 3 concrete moves that completely changed...
YOUTUBE DESCRIPTION
You "use AI" every day. But be honest with yourself — you're mostly just replacing Google searches with slightly longer questions to ChatGPT. You're using a Ferrari to check the mailbox.
In this video I break down 3 concrete moves that completely changed how I use these tools — and I'm not talking theory. I've spent the last year deploying real AI systems for real companies, and what separates the people who get actual value from this stuff from everyone else comes down to one thing: how you engage with it.
Here's what's in this video:
→ Move 1: Stop asking questions. Start giving context.
→ Move 2: Use AI to pressure-test your thinking, not just generate stuff
→ Move 3: Give it a role, not just a task
This isn't a tutorial on prompting tricks. It's a posture shift. And once you see it, you can't unsee it.
Drop in the comments: what's the most useful thing you've actually used AI for recently? Not the most impressive. The most useful. I'll be reading.
↓ Subscribe if you want more AI content from someone who actually deploys this stuff for a living.
I got diagnosed with ADHD at 35. By that point I'd barely made it through high school, spent years bouncing between jobs I hated, and thought I was just lazy. Turns out there's a name for it. This video isn't really about the diagnosis though. It's about what happened when AI...
I got diagnosed with ADHD at 35. By that point I'd barely made it through high school, spent years bouncing between jobs I hated, and thought I was just lazy. Turns out there's a name for it.
This video isn't really about the diagnosis though. It's about what happened when AI showed up — and why I think, for people with ADHD specifically, it hits different than it does for everyone else.
What I cover:
Why every productivity system failed me (and why that's not a character flaw)
The blank page problem — and why AI solves it in a way nothing else did
How I use Claude to recover lost context when my brain loses the thread mid-task
Decision paralysis and the 45-minute email problem
The 2am brain dump — and why having an AI that never loses context is actually kind of emotional
The honest part: AI didn't fix my ADHD. Here's what it actually changed.
There's a growing body of research on AI and executive function support that I want to dig into in a follow-up video. If that's something you want to see, let me know in the comments.
I deploy AI systems for real companies — everything from 10-person DTC startups to major international fashion brands. But this video comes from a different place. It comes from four years of figuring out what actually works for a brain that doesn't work the way the world expects it to.
If you have ADHD and you've found a specific way AI helps you — drop it in the comments. That might be the whole next video.
🔔 Subscribe for real-world AI use cases, honest takes on what this technology can and can't do, and occasionally, overly personal stories about executive dysfunction.
Chapters:
00:00 — Diagnosed at 35
01:30 — Why every productivity system failed
02:45 — Problem 1: The blank page
04:00 — Problem 2: Context switching
05:00 — Problem 3: Decision paralysis
06:00 — Problem 4: The 2am brain
06:45 — The honest part: AI didn't fix my ADHD
08:00 — What it actually changed
08:45 — Subscribe / follow-up
AI hallucinated 25% of the time in the largest hallucination study ever conducted — 35 models, 172 billion words of evaluation. But that's not even the most alarming finding. I went through the full study so you don't have to. Here's what it means for anyone building with AI,...
AI hallucinated 25% of the time in the largest hallucination study ever conducted — 35 models, 172 billion words of evaluation. But that's not even the most alarming finding.
I went through the full study so you don't have to. Here's what it means for anyone building with AI, deploying AI, or just trying to understand what's actually happening under the hood.
What we cover:
What "Document Q&A" is and why it's the most important AI use case nobody talks about honestly
The fabrication floor: even the BEST model in the study still makes things up 1.19% of the time — and what that means at enterprise scale
Why bigger models are NOT better at this (a 3B active parameter model beat a 405B model on hallucination)
Why setting temperature to zero might actually be making things worse — including the model that looped 48x more at T=0 vs T=1
The most important finding: grounding accuracy and fabrication resistance are completely different skills, and most benchmarks only test one of them
What to actually do about it: model selection, context length testing, and safeguards
The study: arXiv:2603.08274 — "How Much Do LLMs Hallucinate in Document Q&A Scenarios?" by JV Roig, Kamiwaza AI (March 2026)
I've personally deployed AI chatbots and agentic workflows for companies ranging from 10-person DTC startups to major international fashion brands. The gap between how these tools are marketed and how they actually perform is one of the most important things to understand right now.
If you've deployed AI — what are you seeing? Does prompting help with hallucination? Drop it in the comments.
Want a follow-up where I actually show you how to TEST your own setup for fabrication before you ship it? Let me know below — if enough people want it, that's the next video.
🔔 Subscribe for AI news, real-world deployment stories, and honest takes on what this technology can and can't do.
Chapters:
00:00 — The stat that started this
00:45 — What they were actually testing (Document Q&A)
02:00 — The numbers: best case, top tier, median
03:20 — It gets worse with longer documents
04:00 — Finding #1: Bigger models aren't better at this
04:50 — Finding #2: Temperature zero might be hurting you
06:00 — Finding #3: Grounding and fabrication are different skills
07:00 — What this says about how we're measuring AI
07:45 — The rice story
08:10 — What to actually do: 3 takeaways
09:00 — Subscribe / follow-up video
Is it still worth learning to code in 2026? I started this channel 10 years ago right after finishing a coding bootcamp. 4 weeks later I had a job. For the past decade, "learn to code" has been the best career advice I could give anyone. I'm not sure that's true anymore. AI...
Is it still worth learning to code in 2026?
I started this channel 10 years ago right after finishing
a coding bootcamp. 4 weeks later I had a job. For the
past decade, "learn to code" has been the best career
advice I could give anyone.
I'm not sure that's true anymore.
AI is already writing code. Claude Code wrote 95% of
itself. Entry level dev jobs are down 20% and some
estimates say they'll drop another 50% this year.
So what does that mean for anyone trying to break
into tech right now?
In this video I share my honest take — not the
Bureau of Labor Statistics answer, not the ChatGPT
answer — my actual experience working in tech and
selling AI solutions for the past year.
#LearnToCode #AITools #TechCareers #ClaudeAI
#CodingIn2026 #AIJobs #SoftwareEngineering
#FutureOfWork #CareerAdvice #ArtificialIntelligence
