923 Studio

923 Studio A venture studio that built 14 startups & 50 client apps to $1B, and can do the same for your app.

02/06/2026

We don't sell AI development. We sell operational efficiency.

Sadly, most founders don't want to hear the latter.

Because AI development looks like:

- AI for the sake of AI
- Unclear project goals
- Expensive prototypes
- Timelines that 3x without ROI
- "Just one more extension" contracts that go nowhere

We enter each AI project with a laser focus on operation efficiency.

We identify bottlenecks.

We guarantee data quality before deploying.

We have elite product managers that scope requirements to what's absolutely necessary.

We do all of this before the alternative 6-month MVP, contract extensions, or an expensive prototype.

AI can turn a 6-figure investment into an 8-figure return if you focus on the right things.

Operational efficiency delivers ROI.

AI development delivers an expensive invoice.

Know the difference - and sign the team that's focused on ROI.

OpenAI's success hinges on a massive flywheel. Right now, it's spinning faster than ever.OpenAI recently dropped three m...
02/05/2026

OpenAI's success hinges on a massive flywheel. Right now, it's spinning faster than ever.

OpenAI recently dropped three major updates in a single week: the global rollout of the affordable "ChatGPT Go" plan, the introduction of ads, and revenue data revealing a staggering $20B+ annual run rate.

These might look disconnected.

But if you look closer, they each contribute to a massive flywheel.

1. Compute converts to capability

Models improve because they're investing $ billions into compute.

With revenue and investment, they can laser focus on buying compute and improving their model performance.

More GPUs, better models.

2. Capability drives adoption

Better models mean better products.

And when switching costs are (close to) zero, you'll see broader market adoption if you have the better model.

To accelerate this, OpenAI is using the new $8/month "ChatGPT Go" plan as a potential loss leader.

The goal is to capture the market and get users "locked in" through personalization, making it harder for them to switch platforms later.

3. Usage fuels training

More users mean more human-generated and human-curated data.

This usage data is essential for training future generations of models, creating an exponential advantage.

4. Revenue funds the future

Adoption drives revenue, which funds the next wave of compute.

With revenue hitting $20B (up from $2B just a few years ago), they have the capital to keep this loop spinning.

As Matthew Berman notes, this is a calculated "4D Chess" move: proving profitability and growth while simultaneously expanding access to capture the next billion users.

If they can keep this flywheel going, they'll create a durable enterprise.

But will switching costs stay low?

Will users want to switch if Google has the best model + the best context?

2026 will show us.

What do you think - is OpenAI's flywheel going to keep working? Or will model performance plateauing slow them down?

MIT, Harvard and BCG studied 244 consultants to understand why some get AI superpowers, and others don't. They found AI ...
01/28/2026

MIT, Harvard and BCG studied 244 consultants to understand why some get AI superpowers, and others don't. They found AI access created 3 different species of users:

Centaurs: They maintain control, using AI as a tool for specific tasks. (High Accuracy).

Cyborgs: They spar with the AI, treating it like a teammate in a continuous loop. (High AI Fluency).

Self-Automators: They paste a prompt and go get coffee.

That third group is where you don't want to be.

The study found negative skill development.

They stopped thinking critically. Their skills began to atrophy because they delegated the judgment, not just the ex*****on.

The authors are terrified of this idea - if you stop using the muscle, you lose it.

The actual "useful development" involves a back-and-forth with the AI.

Telling it the problem, working on a solution, questioning the output, understanding why AI made the choices it did.

Treating it like a real coworker.

This should be your goal with AI. Sparring with it, continuously.

That's how you become a Centaur and a Cyborg.

The AI solutions we develop for our customers create Cyborgs and Centaurs.

We want to empower your employees, not take their autonomy away.

But if you're not careful, you'll end up with a company of self-automators.

And by the way, over 90% of the interviews we see from our team, are interviewing Self -Automators.

Let that sink in.

01/26/2026

AI works. We have multiple 6-figure products delivering 7-figure results.

We know it sounds crazy.

But when you start focusing on what's realistic and possible with AI, instead of following AGI...you get results.

01/26/2026

Our $10m+ clients aren't chasing benchmarks when a new AI model comes out.

We set them up so they don't have to because we focus on what the models already can do.

The headlines scream "GPT 6 beats Gemini 4 on this one obscure International Math Olympiad benchmark."

It might be great for the AI labs, but for you?

Benchmarks are not going to magically help your product.

Real Engineering modeled around your business being unique is going to matter.

When we build AI solutions for our clients, we focus on two things:

1. Which ecosystem are you built around (Microsoft, Google Workspace, or OpenAI)
2. Which results are you looking for?

(Notice we don't ask "Which model would you like to use" anymore?)

Yes, our clients can easily switch to different models if one performs better.

But they don't do it when a new model comes out because they are already seeing great results. The model already did its job.

If the new model improves their results, then they switch.

If it doesn't...no need to panic.

Benchmarks aren't everything if you want to generate ROI.

We have multiple 7-figure solutions that prove it.

A 0.5x AI solution might prioritize latency. But a 10x AI solution goes for accuracy.Let us explain with an analogy.At t...
01/23/2026

A 0.5x AI solution might prioritize latency. But a 10x AI solution goes for accuracy.

Let us explain with an analogy.

At the airport, there are two types of people.

Frequent travelers who take the priority line and TSA Precheck, and random vacationers who use the regular line.

Now, what would you do if you saw the Precheck line was longer than the regular line?

Frequent travelers always pick Precheck.

Why?

Because they know the people in that line are going to move efficiently.

They know what they're doing. They're focused on efficiency.

The other line is full of vacationers who forget to take their shoes off, and get their luggage pulled aside.

Even though it might take me less time in that line, it's more painful.

It's the exact same thing with AI.

10 minutes of prompting creates a 10 minute time saving. 5 hours of planning, editing, refining, editing, then prompting - results in 5 hours of your life saved every week.

Hallucinations are not worth the shorter wait.

The fix, of course, is trading off latency. Longer time generally means more accurate results. This is worth it.

We're all getting tired of hallucinations.

Customers hire us because they are, too.

They're seeing the value of LLMs when they can trust the results, even if it takes an extra minute or two.

Accuracy and results > speed.

This is why Gemini with memory inside Google Workspace is so good.

It knows your Gmail, it knows your contact, it knows your memory, it knows the files you're looking for.

You're happy with it taking an extra 25 seconds.

It's already doing the work you took 4 hours to do manually.

Waiting a bit longer gets me a much better experience.

Our clients feel the same.

01/19/2026

When we build AI solutions, we don't pitch "Replacing humans."

We pitch "Give your employees the ability to work more."

There's a 10x difference.

01/16/2026

AI is a bubble. But it's also the biggest opportunity this century.

We're not even close to utilizing AI's full potential.

No matter what happens, it's here to stay - and our customers are positioned for the long run.

01/12/2026

AI can remove bottlenecks, but if you're not careful, it creates new ones.

99% of companies are stuck in the "Valley of AI Despair".They were initially excited about what AI could do for their bu...
12/01/2025

99% of companies are stuck in the "Valley of AI Despair".

They were initially excited about what AI could do for their business, but expectation hasn't been reality.

- Unpredictable Errors
- Countless Hallucinations
- Wasted Time & Capital

Most companies are still in the "pilot" phase while the industry-leaders are already scaling their projects in real-world environments.

Where is your company on the AI Implementation Journey?

AI should be an equalizer, but instead it's creating an even larger gap between SMBs and Enterprise.According to McKinse...
11/26/2025

AI should be an equalizer, but instead it's creating an even larger gap between SMBs and Enterprise.

According to McKinsey & Company's new study on AI, majority of companies making less than $100 million in annual revenue are still "experimenting" with AI.

While majority of companies earning $5 billion or more are "scaling" their AI efforts.

What do you think prevents smaller companies from leaving the experimental phase?

Not everything is an "AI Agent"Let's clear up some confusion. Level 1 - AI Tools These tools require prompts in order fo...
11/19/2025

Not everything is an "AI Agent"

Let's clear up some confusion.

Level 1 - AI Tools
These tools require prompts in order for you to get back specific results or for it to take specific actions.
Usually a 1-to-1 ratio prompt to action.

Level 2 - AI Agents
After your initial prompt with an AI agent, it can continue to perform certain actions.
This creates a 1-to-many ratio of prompts to actions.

Level 3 - Agentic AI
Reinforcement learning, self-directed goals, multi-agent collaboration.
This is where you unlock agents creating other agents.

Address

250 Independence Way
Boston, MA
01923

Website

https://ninetwothree.co/

Alerts

Be the first to know and let us send you an email when 923 Studio posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Share