Everybody's getting their own AI agent. Who's building it?

I've been deep in Claude Code lately across my work at Flintolabs and beyond. Building automations, testing ideas, shipping things faster than I expected. And the more I use it, the more one thing becomes clear: the bottleneck was never the technology. It's always been the thinking.

The people getting the most out of AI tools right now aren't the ones who know the most about AI. They're the ones who can clearly define a problem, break it into steps, and explain the goal without ambiguity. That's it. That's the skill. Everything else follows.

I came across an article this week by Brad Feld, a well-known venture capitalist, that made me stop and reread it twice. He built what he calls CompanyOS: a system of twelve markdown files that essentially runs his entire company through Claude Code. Email drafting, customer support, meeting prep, content creation, product feedback analysis, all of it. No elaborate software. No expensive platform. Just structured documents that teach Claude how to think about each part of his business.

What struck me wasn't the technical cleverness of it. It was the philosophy behind it. He writes that Claude Code doesn't need another orchestration layer on top of it - it needs domain knowledge. The markdown files aren't code. They're thinking, written down in a format AI can act on.

That's exactly what we're seeing with our students. The ones who build the best projects aren't necessarily the strongest coders. They're the ones who can articulate the problem, describe the user, and explain what a good solution looks like. The AI handles the rest. We've started calling this the real skill, not prompt engineering, but problem framing.

One of the most exciting parts of this shift is what it means for students entering the workforce. AI-native workplaces aren't coming, they're already here. The question every student (and honestly every professional) should be asking is: do I understand my problem well enough to tell AI how to solve it?

That's what we're teaching at Flintolabs. And right now, it's the most important thing we could be teaching.

-Janani

Application Deadline: April 20, 2026

Preiliminary Round: May 16, 2026

Final Round: June 20, 2026

Format: Virtual, Global

What: An online 12-hour business idea competition where high school students from around the world tackle pressing global challenges - environmental sustainability, social welfare, education, and more - and propose science- and technology-based business solutions. On competition day, the challenge is announced at 8:00 AM and teams have 12 hours to develop and submit their idea: a two-page solution document and a three-minute video pitch.

Who: High school students aged 14–18, competing in teams of 3 to 8. No geographic restrictions — you can participate from anywhere with internet access.

Format: Teams are encouraged to combine diverse skills across science, technology, business, marketing, and communication. Winners advance from the Preliminary Round to the Global Final as their country's representative team.

Prizes: Trophy and certificate of achievement for winners; certificate of participation for all teams plus a written feedback sheet on your submission.

What Makes It Special: Unlike most competitions, this one drops the challenge on the day - so it's not about who prepared the "best" idea in advance. It's about how fast and creatively your team can think, collaborate, and communicate under pressure. Exactly the kind of high-stakes problem-solving that colleges and employers notice.

Perfect for: Students who want global exposure, team collaboration experience, and a chance to tackle real-world problems with business and tech thinking. No prior experience required - just curiosity and a team ready to build fast.

The apps you use every day — Google Docs, Slack, Spotify, Duolingo — are all built on a model called SaaS (Software as a Service). You log in, you click around, you get things done. For two decades, this model powered a $300 billion industry. But according to IDC research, AI is rewriting the rules entirely.

The problem with SaaS? It forces people to adapt to the software — learning clunky interfaces, juggling dozens of tabs, copying data between systems that don't talk to each other. AI agents change that equation. Instead of logging into five platforms, you tell an AI what needs to get done and it handles everything in the background. IDC predicts that by 2028, pure seat-based software pricing will be largely obsolete, with 70% of vendors moving toward models built around outcomes and results.

Microsoft's CEO already declared SaaS dead. The question isn't whether the shift is happening — it's whether your student is on the building side or the sideline when it does.

The students who understand how AI agents work, how to design for outcomes instead of users, and how to create tools that solve real problems — those are the ones companies will be desperate to hire or fund. And you don't need to wait until college to start.

Read the full Flintolabs breakdown →

If the first article told you the old software model is going away, this one shows you exactly what's replacing it, and why the timing matters for students right now.

This week, Anthropic launched its most aggressive enterprise push yet: a full plug-in system that lets companies deploy AI agents for finance, HR, legal, and engineering, customized to each organization's specific workflows and data. Pre-built agent templates for tasks like financial modeling, generating offer letters, and drafting job descriptions are ready to deploy, and connectors for Gmail, DocuSign, and other tools mean agents can pull live data directly from the systems companies already use.

Anthropic's head of Americas said plainly: "2025 was meant to be the year agents transformed the enterprise, but the hype turned out to be mostly premature. It wasn't a failure of effort. It was a failure of approach." The new system fixes that by giving corporate IT the controls and customization they need to actually deploy these tools at scale.

The goal, as Anthropic's product officer put it, is "everybody having their own custom agent." But here's the catch - someone has to build those agents, configure them, and manage them. That's not a job for someone who just knows how to prompt an AI. It's a job for someone who understands how these systems work from the inside out. The SaaS era is closing. The agent era is open. The builders who show up early will define it.

Read the full TechCrunch article →

This week's session at Flintolabs tackled one of the most important questions of our time: how do you know if what you're seeing online is actually real?

We devoted a full session to deepfakes; what they are, how they're made, and critically, how to fight back.

The session opened with something that surprises most students: deepfakes aren't magic. They're built using a class of AI called generative adversarial networks, or GANs, where two AI systems compete against each other, one generates fake images or videos, the other tries to detect them. This back-and-forth is exactly what makes deepfakes so convincing. And it's also what makes detection possible.

Students learned the tell-tale signs humans can spot: unnatural blinking, mismatched skin tones around the hairline, inconsistent lighting on the face, and audio that doesn't quite match lip movements. But they went further than just spotting fakes with their eyes — they explored the technical side of how detection actually works at scale.

Real-world deepfake detectors analyze pixel-level inconsistencies, frequency patterns that don't appear in authentic media, and artifacts left behind by the generation process. Tools built on computer vision - the same technology we've explored this semester with OpenCV - can flag manipulated media that the human eye would miss entirely.

The session ended with a question we asked our students directly: if detection technology exists, why aren't platforms using it? The conversation that followed was one of our best all semester. Students aren't just learning to build — they're learning to think about what should be built, and why urgency matters.

Watch student projects from Flintolabs →

Our March cohort is filling fast. If you've been thinking about enrolling, this is the moment.

Here's what you get with just one hour per week for 6 months:

✅ Hands-on AI skills through building real applications, not watching lectures

✅ 3 transferable college credits from University of Colorado Denver

✅ Portfolio of real work that demonstrates capability to colleges and employers

✅ Small class sizes (capped at 20 students) ensuring personalized attention

✅ Advanced concepts like OpenCV, Minimax algorithms, computer vision, and more

✅ The critical thinking and problem-solving skills employers desperately need

While college grads struggle with workplace readiness and entry-level jobs disappear to automation, Flintolabs students are building portfolios of real work that demonstrate genuine capability.

Classes start Saturday, March 7!

Enroll now →

Follow our LinkedIn page for free Q&A sessions where you can ask anything about our program, how you can earn credits, typical lessons students learn and really anything you have in mind before enrolling!

Our program has a 5-star rating with reviews from both students and parents.
Questions? Email us at [email protected]

View all reviews on Google

Found this valuable? Forward this newsletter to other high schoolers and parents who want to be informed about AI trends and what is needed to prepare for an AI-driven future. Every student deserves the chance to build real skills before college.

You're receiving this newsletter because you expressed interest in Flintolabs or crossed paths with our community. If you believe you received this in error, please feel free to unsubscribe using the link in this mail.