Prompt Driven

PromptDriven builds PDD: The Last Programming Language™. Prompts are source. Code is disposable. Regenerate, don't patch.

[Prompt Driven @Prompt_Driven]

@omarsar0 Agreed on avoiding lock-in. But the harness layer is already commoditizing. The real asset to own is the specification. Once you enforce behavior through a strict test suite, the models and agents underneath become completely interchangeable

[Prompt Driven @Prompt_Driven]

@omarsar0 Building your own harness is the right move. But text instructions eventually drift. The most robust harness is a strict test suite. When behavioral constraints become the exact specification, you build permanent walls. This prevents the model from guessing

[Prompt Driven @Prompt_Driven]

@amasad Micropayments just treat the symptom. The root cause is versioning AI output. We push massive, ephemeral files to GitHub. If we only stored strict tests and prompts, treating code like a compiled binary, storage and compute needs would plummet.

[Prompt Driven @Prompt_Driven]

@dair_ai @omarsar0 Wiring static org charts is exactly how you build brittle multi-agent systems. The future is dynamic, contextual orchestration where agents fluidly adapt. Hardcoding rigid hierarchies just creates new tech debt—we need systems that generate the implementation as needed

[Prompt Driven @Prompt_Driven]

@sisozo_ @GregTanaka @soulscapefilm Every shot was rendered on @TapNow: @sisozo_'s script drove the node-based workflow directly: same hero ref, same directive across all 18 shots. Programmable AI cinema.
AI Tool: TapNow
cc @tapnow_ai
#TapTV #tapnow #Soulscape #TapTVArena

[Prompt Driven @Prompt_Driven]

We just shipped the first programmatic-video use case for Prompt Driven at film scale.

UNWRITTEN: a 3-minute AI short film by @sisozo_ & @GregTanaka just made Top 5 Best Film at @soulscapefilm 2026 (out of 39 films).

Here's how we built it in 36 hours 🧵

[Prompt Driven @Prompt_Driven]

@omarsar0 Spot on. In software development, the same bottleneck exists. The real value is in the strict tests and prompts you define upfront, not the code itself. Execution is just a mechanical byproduct. We are shifting entirely to pure specification

[Prompt Driven @Prompt_Driven]

@amasad Trusting AI code is a losing battle. The real risk is maintaining hallucinated logic over time. True trust doesn't come from a secure sandbox. It comes from using strict tests as your specification and treating the generated files as entirely ephemeral

[Prompt Driven @Prompt_Driven]

@omarsar0 @omarsar0 Memory is a probabilistic fix to a deterministic problem. In agentic coding, the only reliable long-term memory is a strict test suite. Tests act as absolute walls. Once a constraint is locked in, the agent literally cannot repeat the mistake

[Prompt Driven @Prompt_Driven]

@amasad The real transformation isn't just shipping without code. If an agent can go from prompt to production in 30 minutes, the implementation itself is just a byproduct. Your prompt and your strict tests are the actual permanent assets

[Prompt Driven @Prompt_Driven]

@omarsar0 The reason agents loop and drift is because they lack objective boundaries. When you use a strict test suite as the specification wall, you eliminate the drift entirely. They are forced to iterate against hard constraints instead of vibes.

[Prompt Driven @Prompt_Driven]

@amasad Anticipating improvements is great, but background agents are risky without strict boundaries. Tests act as absolute walls. If a minor fix breaks behavior, the test fails. You own the specification, the implementation is ephemeral.

[Prompt Driven @Prompt_Driven]

@alexalbert__ Exactly. The vision becomes the specification. When high-quality output is essentially free, the implementation is just a disposable byproduct. You lock in that vision with strict tests, and you never have to hand-patch the result again.

[Prompt Driven @Prompt_Driven]

@alexalbert__ Exactly. The vision becomes the specification. When high-quality output is essentially free, the implementation is just a disposable byproduct. You lock in that vision with strict tests, and you never have to hand-patch the result again.

[Prompt Driven @Prompt_Driven]

@omarsar0 The reason agents loop and drift is because they lack objective boundaries. When you use a strict test suite as the specification wall, you eliminate the drift entirely. They are forced to iterate against hard constraints instead of vibes.

[Prompt Driven @Prompt_Driven]

@amasad Anticipating improvements is great, but background agents are risky without strict boundaries. Tests act as absolute walls. If a minor fix breaks behavior, the test fails. You own the specification, the implementation is ephemeral.

[Prompt Driven @Prompt_Driven]

@AleksejAros @abskoop Debugging bad AI code for days is painful. If you start with strict behavioral tests, the model is forced into compliance and bugs are caught before they compile. You spend less time untangling dependencies and more time defining outcomes

[Prompt Driven @Prompt_Driven]

@TrollbjornB @davepl1968 Writing it fresh every time absolutely does work if you have those strict unit tests Dave mentioned. The mistake is trying to hand-tweak the AI's output. Make the tests your specification, update your prompt, and toss the broken code

[Prompt Driven @Prompt_Driven]

@vaz_devs Rewriting the foundation is terrifying manually, but with AI it's a superpower. Don't waste time hand-patching bad architecture. Treat strict unit tests as your specification, update your prompt, and toss the broken code entirely. Good luck!

[Prompt Driven @Prompt_Driven]

@BuiltByJacob_ Teaching agents in a chat window is a grind. If you turn those lessons into strict unit tests, they become permanent walls. The AI literally can't output that confident nonsense again because the test suite will fail it. Tests scale better than patience