Go Back to the Source Before You Build a Solution
Most problems disappear when you return to the source. Stop solving around the problem. Fix the origin and the output fixes itself.
Video and audio in Dutch
Key takeaways
- Return to the source before building a solution. Most loops and bugs are not code problems. They are clarity problems about where the work actually starts.
- Identify whether your source is actually the source. The DJI gimbal has calibration built in. The real source was the settings, not the image. Check your assumptions before you build.
- Apply your own principles everywhere, not just in one context. Paul had a "return to source" rule for AI work and ignored it completely in his physical setup. Rules you don't apply universally are rules that cost you time.
- Content generated from content degrades. Identity First Media's pipeline always connects back to the original upload. Every derivative piece is mapped against the source, not against the last output.
- Some problems are not problems. The real problem is treating them like they are and spending hours building around something a single setting would have fixed.
Timestamps
Show notes
The Core Insight
Most problems are not problems. They become problems when you build solutions around them instead of going back to where they started. Paul Veth, building Identity First Media and multiple SaaS products with Claude Code, kept getting stuck in loops. The fix was never more code. It was always the same: return to the source.
What This Episode Covers
- Why Claude Code's context window (now 1 million tokens, up from 200,000) still causes quality degradation and how Paul handles it
- How a crooked camera frame led to building an entire correction tool, when a single calibration tap would have solved it
- Why the "return to source" principle is the architectural foundation of Identity First Media's content pipeline
- The telephone game problem: why content generated from content generated from content always loses the original signal
Key Frameworks
Return to Source in AI Workflows
When Claude Code loops, Paul stops and identifies the actual source of the problem. He feeds that source directly into a new conversation. The loop breaks because the origin is clear, not because more code was written.
Return to Source in Content
Identity First Media's pipeline works like this: the creator uploads source material, that material is transcribed and mapped against the original, and all derivative content is generated from that source, not from previous outputs. The source is always the highest quality version. Everything else is downstream.
The Telephone Game Problem
Content that generates content that generates content behaves like the childhood whisper game. "Strawberry" goes in one ear and "chair leg" comes out the other. Data gets lost. Errors compound. The only fix is anchoring every output back to the original source.
Mentioned in This Episode
- Claude Code by Anthropic, used by Paul for approximately 95-99% of code generation across his SaaS products
- DJI gimbal stabilizer, with built-in auto-correction and calibration settings
- Identity First Media, Paul's digital infrastructure platform built on the Identity First philosophy
Topics
Full transcript
View full transcript
Frequently asked questions
What does 'return to source' mean in practice?
Return to source means stopping before you build a workaround and asking: what is the actual origin of this problem? In Paul's AI workflow, it means starting a fresh Claude conversation anchored to the original goal. In content, it means every output traces back to the creator's original upload, not to a previous output.
How does Paul Veth use Claude Code to build SaaS products?
Paul uses Claude Code for approximately 95 to 99 percent of code generation across his SaaS products. He manages context by moving problems into new conversations before the context window degrades. The 1 million token window helps but does not eliminate quality loss toward the end of long sessions.
Why does Identity First Media avoid generating content from content?
Generating content from content compounds errors, exactly like the childhood telephone game. Identity First Media's pipeline always maps derivative content against the original source material. This preserves signal quality and keeps the creator's actual voice and intent intact across all outputs.
What is the Identity First philosophy?
Identity First is Paul Veth's core framework: start with who you are, not with templates or market analysis. In technology, this means AI amplifies your existing identity and intelligence rather than replacing it. Every product Paul builds, from Identity First Media to Aligned to MentoSprout, starts from this principle.
How do you break out of an AI development loop?
Stop adding code. Identify the actual source of the problem, not the symptom. Start a new AI conversation and feed it the source directly. Paul finds that returning to source often reveals the solution already existed. The loop was happening because the origin was unclear, not because the code was wrong.
Related episodes
Get in touch
Want to learn more or collaborate? Feel free to reach out.
Get in touchDiscussion
This content argues that most solutions fail because they target symptoms instead of the source. Where have you caught yourself building a solution around a problem instead of going back to the origin, and what did it cost you?


