Drew

Building Financial Super Intelligence @CFOSilvia

[Drew @DCheneler1]

Silvia just beat OpenAI, Claude, and Grok on a tax benchmark. While this is an impressive feat, this is not the important headline. Far from it. https://t.co/1fEs0EN01f

[Drew @DCheneler1]

@IndependentEco Opus 5.5, the only model that matters 😂

[Drew @DCheneler1]

@tibo_maker What are you using for voice overs?

[Drew @DCheneler1]

@awayslice Give @cfosilvia a try!

[Anthony Pompliano 🌪 @APompliano]

People don't want Wall Street telling them what to do with their money.

That is why we built Silvia.

A private, secure AI agent to help you manage your investments and finances.

Try it free: https://t.co/mpzm60uv5O

I explained on Fox Business this morning.

[Drew @DCheneler1]

@IndependentEco All the time! Same with Grok Bot

[Drew @DCheneler1]

The most underrated prompt I know takes 10 seconds:

> Prompt: It's six months from now. This plan failed. Tell me why.

Everyone enjoys a good post-mortem. Why not do a pre-mortem? Super intelligence makes it ridiculously easy.

If you just ask:

> is my plan good?

You usually get polite reassurance. Models tend to lean agreeable. Instead, ask:

> why did it fail?

Now you changed the models job. The new goal is now to find problems, and frontier models are pretty dang good at this.

Pre-mortems are particularly useful for:

> product launches
> budget builds
> marketing campaigns

This is a great exercise to try out if you want a different perspective.

[Drew @DCheneler1]

@emonuxui Exactly! Another great reason I do this all of the time.

[Drew @DCheneler1]

I am becoming convinced that 99% of the World uses AI backwards.

Almost everyone writes a prompt, gets a mediocre answer, and blames the frontier model.

I recommend flipping the process. Have Opus/GPT interview you first.

In all of my prompts, I add one subtle line:

> before you start, ask me 5 questions you need answered to complete this task well

This is great, because I often forget to include important details and context in the original prompt. The interview is a chance for me to provide information I may have missed.

This also forces you to clarify. Half the time, answering the questions is what makes you realize what you actually want.

The best frontier model responses come from real context, not better prompts.

[Drew @DCheneler1]

@IndependentEco I’ll actually be shocked if it is as good as Astra. No way will it compete with Opus 5.5 in my opinion.

[Drew @DCheneler1]

@cfosilvia A trend we will see accelerate across the industry

[Silvia @cfosilvia]

Not your intelligence, not your AI

[Bilal Dhouib @goatlal]

Not your intelligence, not your AI.

Just like the good old saying: “not your keys, not your crypto”

[Shain Noor @shaincodes]

We did the same thing @cfosilvia. Fine-tuned Qwen to do the quant work for our users. It beats Opus 5 on our tests and costs like 93% less to run.

Every company will come to the same realization.

[Drew @DCheneler1]

@opynrijal Grok Bot just too good, it’ll be hard for me to switch. But I know Meta just improved it a lot with new updates

[Drew @DCheneler1]

Meta Muse growth is already beginning to slow down. It’s had a good run, but agent competition is heating up.

I don’t use Muse and can’t think of a personal use case where it makes sense. But that is just me personally. I am sure it’s a great product.

Grok Bot is just too good and it has all of my context. And it becoming model agnostic just locked me in for good. To switch, the ROI would have to be significant. I suspect there are thousands of users in a similar situation to me.

Once you have a personal agent dialed with the right context and workflows, it becomes similar to a credit card. You rarely switch or get a new card more than two times in your life.

[Drew @DCheneler1]

@Payton_Thompson That is actually a good point. I don’t sell anything on Facebook marketplace, so that never crossed my mind.

[Anthony Pompliano 🌪 @APompliano]

I have been banging the drum about companies fine-tuning their own models and now Coinbase is saying they are doing it themselves.

Faster, cheaper, better than the frontier models.

This is exactly how @cfosilvia reached outperformance in important personal finance topics.

[Drew @DCheneler1]

@lv_feng47448 The one arguing for Muse usage?

[Drew @DCheneler1]

@IndependentEco I’ll have to give Fable a shot from an orchestration perspective. I just don’t want to pay that much haha

[Drew @DCheneler1]

When does it make sense to use Fable when Opus 5.5 can handle pretty much everything?

Even Fable on low still underperforms Opus 5.5 Medium/High in my opinion.

[Drew @DCheneler1]

@JustinMeck_ @cfosilvia This is a great use case! Thank you for sharing. I am going to give this one a try as well!

[justin.mecklenburg @JustinMeck_]

I have a complex insurance situation. I recently asked @cfosilvia about it, providing my policy statement and context.

Silvia didn't just answer my questions...it caught risks and policy language that I wasn't even aware of. The answer was far more in depth and nuanced than I originally thought. Silvia presented the risks and outcomes in a clear and concise manner, allowing me to make the best informed decision.

There is no way I am getting this level of expertise and detail from an insurance agent - particularly on a Sunday.

@cfosilvia is freaking awesome.