Cheaper AI Will Not Fix a Slow Content Workflow: How LPV Uses GPT-6 Sol and Luna

If every post takes hours to plan, write and check, a cheaper AI model will not give you that time back. The work needs a clear order, and someone still needs to check that the message sounds like your business.

Alex has been testing GPT-6 Sol and Luna in LPV Agency’s ClientWorkSystem. His approach puts the hardest thinking first, then gives smaller production tasks to a faster, cheaper model.

TL;DR

• GPT-6 Sol and Luna launched with API prices 50% below the GPT-5.6 promotional prices cited in the video’s caption.
• Alex uses Astra for planning, Sol to guide and monitor work, and Luna for focused production tasks.
• After 24 hours of heavy testing, Alex said his usage allowance was only 25% down.
• LPV has added the models to ClientWorkSystem and is seeing faster hooks and better scripts and text in its early testing.
• Before publishing, check whether the content says what you do, who you help and what result you create.

What Is This? (Short Answer)

It is a way to split AI content work between three models according to the task. Astra handles high-level planning, Sol gives instructions and checks progress, and Luna works on the focused production tasks.

For LPV, that work feeds ClientWorkSystem, the app the agency uses to make client content. The aim is to spend more of the available budget on thinking and checking while keeping repeatable tasks efficient.

How does this work?

Start with a plan, guide the work, then check the result against the business. Alex’s suggested sequence is Astra for the plan, Sol for directing worker agents and monitoring them, and Luna for work on files and other defined tasks.

He says planning is the most expensive part, but it takes less time than production. That is why he reserves Astra for high-level decisions instead of asking it to perform every small step.

In LPV’s content workflow, the starting point is the client’s business, website and reviews. Those details help the system turn the client’s input into ideas, hooks, scripts, videos and graphics that reflect what the business actually does.

Alex’s first impression is that the newer models follow his instructions better. He reports faster hook generation and improvements in scripts and text, though those are observations from his own early use.

Who is this for?

This approach is useful for businesses and builders who repeat content or app-building tasks but still need careful planning. Alex describes using the same model roles when instructing coding agents and working on client content.

For a UK business using LPV’s social media service, the client records two minutes of video a week. ClientWorkSystem then supports the ideas, videos, graphics and posting around that input, with websites and ads able to carry the same message.

The point is recognisable content. A polished draft is of little use if it misses who the business helps or gives people no reason to trust it.

What does it cost?

The cited GPT-6 Sol and Luna API prices are 50% below the GPT-5.6 promotional prices. That gives Alex more room to refine content without making every small task as costly as high-level planning.

The video does not give a price for LPV’s service or a fixed cost for this workflow. Actual model spending depends on usage and on how often each model is called.

Alex gives one concrete example from his own testing: after pushing Sol and Luna hard for 24 hours, with some Astra used for planning, he said his usage allowance was only 25% down. That is his experience, not a spending forecast for another business.

What are the risks?

The main risk is publishing content that sounds good but gets the business wrong. A quick draft can miss the service, the audience or the result, even when the wording is smooth.

Alex believes Sol and Luna are less prone to making things up, but he still assigns Sol a monitoring role. Better model behaviour does not remove the need to check claims, instructions and the finished message.

There is also a cost decision. Astra is the expensive option in Alex’s account, so giving it every production task would work against the efficiency he is testing.

Key Takeaways

1. Give each model a clear job. Use Astra for the hardest planning, Sol for guidance and monitoring, and Luna for focused execution.
2. Feed the workflow real business context. The website, reviews and the client’s own input give content something specific to say.
3. Judge the output by clarity. Faster hooks and stronger scripts matter only when people can recognise the business behind them.
4. Treat early results as early results. Alex has implemented the models in ClientWorkSystem and reports improvements from his testing so far.

Implementation Checklist

• Write down what your business does, who it helps and the result it creates.
• Give the planning step your website, reviews and direct input before asking for content ideas.
• Assign high-level planning, work instructions and focused production as separate tasks.
• Review each hook and script against your business details before publishing.
• Check that videos, graphics, posts, websites and ads tell a consistent story.

Common Mistakes to Avoid

• Asking one model to plan, produce and approve the entire job in a single pass.
• Publishing a polished draft that could belong to any business.
• Using a costly planning model for every small production task.
• Assuming lower model prices automatically save time without a clear workflow.

Alex’s final check is simple: does this say what you do, who you help and what result you create? If you are building a workflow like this or have questions about his setup, he invites you to ask for support.