Ray

Founder of an AI OPC.
📈GTM Expert & Brand Manager
How to run a business is more important than how to use AI

[Ray @RayAndresW]

@TTrimoreau What if your target audience are not on X

[Ray @RayAndresW]

This is exactly how do you start a business, more trial and error, less perfection

[designwithkingsley @desgnwitkinsley]

A very short story

[Ray @RayAndresW]

@oleg_nikolaev1 That is more detailed about how you convince customers, The article focuses on reaching customers first. Then it's convincing them.

[Ray @RayAndresW]

https://t.co/9PNZtoZ2SY

[Ray @RayAndresW]

I told Dots to amaze me in any ways he can.
Then Dots rendered a Rabbit in Skull Shadow in Blender

[Ray @RayAndresW]

@o_shinworks Yes! Experience is lot more important than function!

[Ray @RayAndresW]

Love this!
Recreating existing product with new aesthetic and attention mechanisms!

Many Builders always want to build a product with no competitors. But they forget one thing, even existing products are thinking about getting new users.

Even for the same pain points, different aesthetic need means different segmented market.

[onoshin@個人開発 @o_shinworks]

【新アプリリリース！】

集中しないと猫が落ちるタイマーアプリをつくりました！
ゆれておちてくらす猫をお楽しみください🐈🐾

インストールはリプライから👇👀
#shipaton

[Ray @RayAndresW]

About Getting customers with Ads

Your best-converting ads might be taking credit for customers you already had.

According to an eBay field experiments, When eBay switched off ads on its own name, people mostly clicked the free result instead.

Before scaling an AI-written campaign, test one thing: do total purchases fall when the ads stop?

Split new buyers from regulars. A campaign that introduces you to someone and one that catches them on their way back deserve different budgets.

[Ray @RayAndresW]

"Effective System vs Your Agents System"

Effective ones rely on system itself, yours rely on your knowledge, that is what stop AI to do better, it is yourself.

Scientists use AI to create proteins, but you say your AI stuck in Google sheet?

[Ray @RayAndresW]

“AI saved me 38 hours.”

Would you actually have spent those 38 hours?

An agent builds a dashboard in 2 hours. You estimate 40 hours to code it manually. But your real alternative was a 30-minute spreadsheet.

That’s new capability made affordable. The value depends on what it changes.

METR’s May survey of 349 technical workers makes the distinction interesting: median self-reported task speed was 3×, while reported value gains were 1.4–2×. These are estimates, not measured ROI.

Before you leave everything to AI, ask: what would we actually have done without AI?

https://t.co/ajYiq5PsQI

[Ray @RayAndresW]

I asked DOTs @thsottiaux to:
1. Learn from amazing videos by Codex on X
2. make a striking video like 3A games per 30 mins, And it gave me this!!!

Also there some problems, like no audio, you need to go to @ElevenLabs to design one

But it's amazing with almost no prompts!

[Ray @RayAndresW]

-You have a basic Agents framework.
-You ask AI to iterate 10 more framework with Agents as controlled variables, more detailed labor division or not.
-Run a few your daily work, and ask to record log, tokens, time.
-Input logs of all to analyze which is the best framework for u

[Ray @RayAndresW]

Dynamic equilibrium and mutual constraints are important to build Agents system.

Single decision-maker (Agent or you) leads to high-speed but low efficiency.

Because they will say and do anything to make the decision-maker happy

[Ray @RayAndresW]

AI always agree with you and make the project away the right way?
-Use another model to be the red team, but this is not complex
- Let Dots to be the Red team, supervise each conversation all the time and remind you!!n

[Ray @RayAndresW]

Easiest way to tailor Agents for your job.
-Drop Openclaw Github link to codex/Claude
-Ask the model to learn from the pipeline and tailor one agents army for your own work

[Ray @RayAndresW]

@UEMusician @ben_sage Thanks, because I've tried put all resources in and that's quite a waste, so maybe I can help people who has the same thoughts as I did hahaha

[Ray @RayAndresW]

@UEMusician @ben_sage Many failures are not caused by that they can't reach customers. It is because the solution itself is not wanted at all.
So don't have to worry about the difference between demo and real product, these batch of customers are just to validate there is a market

[Ray @RayAndresW]

@UEMusician @ben_sage It depends, some builders can build a mvp in 2 hours, then they can record a real product.
But if some builders feel expensive and time-consuming to build a mvp, then it could be a fake video..
The key idea is to quickly validate the market before you put too much resources in.

[Ray @RayAndresW]

Time and Token you spent to create a Claude video can be used to build a product demo.
If you have time to create a threejs show, why don't you create IP assets for your future Brand?
Is Token that cheap to waste?