Introduction: The Shift from Technical Labor to Creative Strategy
What happens when a veteran systems engineer, software architect, and digital artist who's spent 30+ years in the enterprise (building for IBM, JPMorgan Chase, AT&T, ADP, QVC, etc.) starts integrating generative AI into their day-to-day production workflows?
Throughout my career, I've seen massive technological shifts, but the rise of AI is in a very different category. It's not just about doing things faster in a mechanical sense. It's actually changing the nature of what we do as software engineers and in the creative arts. My role is shifting from getting down into the nitty-gritty of syntax and getting things to work to high-level vision, workflow, and creative strategy.
I will admit, one new paranoia has seeped into my writing. In the past, I've used tools to help with grammar and spelling, etc., but now I have to worry about patterns that may show up as AI. Is this the way I would write it, or is AI saying it for me in a different way? At some point, this gets pretty silly, and I just use the tool. The ideas, concepts, and lessons are mine; AI didn't experience them, I did. The following five takeaways are the most surprising lessons I've learned through actual hands-on experimentation across traditional systems engineering, 3D rendering, animation pipelines, and agentic vibe coding.
Takeaway 1: Humans and AI Learn Thinking in Exact Reverse
Human intelligence is bottom-up. We start as infants by crawling about on the floor, looking up at the giants around us and taking in the world through all our senses (sight, sound, touch, taste). Language comes in much later as an artificial device to describe the sensory experiences and internal mental states we have already accumulated.
AI develops in the exact opposite direction. It learns the fundamental concepts through language first (by processing humanity's descriptions of the universe in text), without the benefit of 24/7 sensory input or direct experience of the physical world.
Because human language is a post-hoc way of describing our internal experience, the AI's fundamental architectural constraint is that it 'sees' and 'understands' the world after the fact.
"Language came after intelligence, but AI is the flip side of that. It's learning from language, which is what we use to describe the world in the aftermath."
Takeaway 2: Real Coding is Shifting from Technical Syntax to “Vibe Painting”
In traditional software development, you'd quite often have the programmer thinking that the actual thinking was down in the realm of the mechanical, i.e., the memorizing of syntax, the configuring of API parameters, the agonizing over the logic of a particular algorithm. AI is absolutely superb at the micro technical stuff. It takes that drudgery off the engineer's back and lets them do the real thinking.
This transition resembles the evolution of software methodologies. In the past, we relied on rigid Waterfall planning with fixed start and end parameters. Then Agile let us be more organic through iterative discovery. Today, agentic programming—or “vibe coding”—takes this evolution further into conversational, continuous discovery.
When I code alongside AI agents, my effort shifts entirely toward overall process design, user experience, and strategic intent. Programming now feels less like rigid structural engineering and more like painting a canvas. By managing intent rather than micro-technical parameters, I encounter Bob Ross-style “happy little accidents” that reveal unexpected problem-solving paths during the creative flow.
“I just didn’t realize how much I had to work on making something actually good, rather than just fixing a particular call or parameter… I don’t miss doing micro-technical research to solve syntax issues.”
Takeaway 3: AI Can Move the Camera Inside Static 10-Year-Old Renderings
In my earlier 3D digital art workflows, rendering complex environment stills required massive computational time. Using legacy tools like Vue—the software famous for environment work on Pirates of the Caribbean a single high-detail forest frame rendered on my PC back in 2011 took most of an entire day.
When I took those static 2011 Vue renders (and complementary character renders from Maya and Poser) and fed them into modern generative video tools like Runway AI, the system didn't simply apply a visual style filter. It extrapolated unrendered spatial geometry, generated missing environmental details, and pushed a virtual camera directly into the scene.
As the camera pushed forward, my static 3D foreground characters dissolved away, revealing a fully realized, deep, dark forest bathed in dynamic, shimmering light. This experiment proved to me that generative AI has transcended pre-production sketching and is entering direct production asset generation.
“The image came to life as a camera moved through this forest that never existed before… It was creating additional details, extrapolating my image into a fully formed world.”
Takeaway 4: Context is the Holy Grail—Moving Beyond Generic Chatbots
Over the past several years, I have tracked AI coding platforms through four distinct evolutionary stages, differentiated by how deeply they understand codebase context:
Basic Chatbots (ChatGPT, Claude): Provide high-level code snippets and general logic analysis, but operate in total isolation without awareness of local repository files or project architecture.
IDE Assistants (Cursor, Copilot, Windsurf): Embedded directly inside development environments like Visual Studio Code, utilizing active file context to suggest inline code completions and function updates.
Agentic Platforms (Claude Code, Devin, n8n, MCP): Autonomous tools capable of task planning, multi-file edits, test execution, and workflow initiation. These are primarily terminal-focused or tailored for traditional computer scientists.
Vibe Coding Environments (Replit, Lovable, Bubble): Conversational platforms that guide users end-to-end from initial prompt ideation through GitHub management, staging, and production deployment.
Full codebase context fundamentally alters developer collaboration. In platforms like Replit—which pairs complete project context with advanced LLMs like Anthropic’s Claude—I can ask strategic questions about ROI, pitch deck research, or 3-year rollout plans grounded in my actual project architecture, not generic boilerplate advice.
However, vibe coding carries distinct technical friction points. Autonomous AI agents regularly suffer from memory loss in complex projects—forgetting prior work, undoing functional code, getting stuck in error loops, or utilizing outdated workflow libraries. To manage this, developers must actively challenge the AI’s architectural analysis, rigorously review production deployment pipelines, and reword prompts whenever analytical gaps appear.
Takeaway 5: Generative AI is Finally Unlocking the Solo “Auteur” Animator
Historically, solo artists could thrive in writing or painting, but feature-length animation remained trapped behind massive studio headcount and capital constraints (e.g., Disney or Pixar). Independent creators attempting solo animation—such as Brian Taylor with his inspiring 2000s project RustBoy, or my own experience pitching my feature animation concept FranknSon in Hollywood in 2010- routinely hit massive walls around render times and production resources.
With generative AI tools, solo creators and small teams can bridge the production gap. In my studio tests, I have achieved compelling results combining custom digital art with AI workflows:
Generating subtle facial performances, lip-syncing character walks, and animation motion based on original concept art, digital paintings, and my brother David Rivera’s historical artwork.
Cloning voice models in ElevenLabs to test a single British accent across both a 1950s live-action-style character and my animated FranknSon character, while also testing American female narration in classroom settings.
Executing seamless multilingual performances, enabling my animated character “Terry, space pioneer” to speak fluently in both English and Spanish with accurate lip synchronization.
Yet, AI agents function like junior assistants. Without strict human domain expertise to guide them, these tools generate empty internet “party tricks” devoid of narrative meaning, emotional resonance, or structural substance.
“You need to use your own expertise and experience to guide these junior AI agents, or else you will simply be creating empty calories, devoid of any true meaning or story…”
Conclusion: Partnering with AI to Reclaim Human Creativity
While technical oversight and domain knowledge are more vital than ever—necessary to catch AI hallucinations, memory loss across large codebases, and visual render artifacts—the net result is an unprecedented leap in human leverage.
By actively catching these technical failure modes early, we free ourselves from low-level syntax research and return to what matters most: crafting elegant, human-centric solutions and telling stories that people truly feel and understand.
If syntax overhead, deployment friction, and technical setup details were no longer standing in your way, what major project or creative vision would you build first?