Justin Gibbs

Last week my buddy and I didn't just join Google's Agent-Dev-Kit hackathon—we GenHacked our way to a working submission. Between Google Meet and vibe coding in Cursor, every hurdle became a conversation with an LLM: architecture diagrams, Dockerfiles, CI/CD, UX copy—everything. That deliberate, AI-first workflow is what we came to call GoFrame. It's a systematic approach that took our Holiday Inn-level bravado into repeatable results.

Vibe Coding + One Night at a Holiday Inn Express = GenHacking

Andrej Karpathy's "vibe coding" captured the mood of AI-assisted programming—letting the model write 90% of the code while you steer the "vibes." But why stop at code? GenHacking applies the same intuition across every discipline. Unfamilar with Docker but need to deploy to the Cloud? Marketing plan at 2 a.m.? Doesn't matter—because, like those old Holiday Inn commercials, "But I Did Stay at a Holiday Inn Express Hotel Last Night"

The Holiday Inn Confidence—Justified (Usually)

AI gives you a booster shot of audacity: you know nothing about Docker, yet you're spinning containers in prod. That confidence works because LLMs compress months of Googling into minutes of dialog—but they also hallucinate and guide you into loops. The trick is pairing swagger with structure.

GoFrame—Holiday Inn Confidence, OODA Discipline

GoFrame borrows John Boyd's OODA loop, doubling down on Orientation—the part most people skip. Every loop forces five checkpoints:

1. Define a Hypothesis
   You start with a clear belief or assumption you're testing.

2. Define a Target State That Proves/Disproves It Quickly (GenHacking is about loops of minutes, hours, vs months and years)
   You create a tight feedback loop, reducing the cost of being wrong.

3. Constrain the Experiment (What Must Stay Unchanged / What Might Shift)
   Instead of hacking blindly, you define the sandbox—what's fair game and what's off-limits.

4. "Vibe Code" with Continuous Cross-Checking Against Constraints
   This is intuitive engineering, but bounded by logic—think of it as jazz improvisation with sheet music nearby.

5. Run the Test. Did It Fail?
   You're not optimizing for success—you're seeking truth.

GoFrame in Motion

Cloud Deploy

• Hypothesis: We can host the app on Cloud Run and the whole signup journey works.
• Target State: New user signs up, data saves, secrets stay sealed—within 30 min.
• Constraints: Leave business logic untouched; adjust only infra code.
• Quick-&-Dirty Flow: Containerize → Terraform → GitHub Actions.
• Pass/Fail Test: Hit prod URL, walk through signup.

Framework Flip

• Hypothesis: Phoenix LiveView can replicate current UX under 500 ms.
• Target State: All forms and nav run as smooth SPA.
• Constraints: DB schema immutable; APIs stay online as fallback.
• Quick-&-Dirty Flow: Scaffold LiveView → Incremental port → Dual-run FastAPI.
• Pass/Fail Test: Full click-through session.

Why It Works

• Speed: Learning happens inside the build loop, not before it.
• Focus: Constraints turn unknowns into sharp edges the AI can't sand away.
• Leverage: You orchestrate AI like a conductor—Logic Designer is the new job title.

Chip Huyen nailed it: "AI doesn't create new thinking; it reveals what actually requires thinking."

Ready to GenHack?

1. Pick a problem one ring outside your comfort zone.
2. Draft a GoFrame loop—write the hypothesis, target, constraints.
3. Pair with the LLM to build out the experiment, but cross-examine its answers.
4. Iterate and test until the hypothesis holds—or collapses

The modern world wasn't built by experts—it was built by curiosity and curious tinkerers like Claude Shannon. And GenAI just gave all tinkerers superpowers.