Your AI just got a team

Good morning. A lot happened in AI this week. Here’s what actually matters.

Subscribe | Past essays | Advertise with us

🤝 ^{This edition is brought to you by Lightfield}

Why deals stall. Why customers churn. Why your competitor keeps coming up. The answers are buried in emails and call recordings — but no traditional CRM can parse them.

Lightfield can. It’s a natural language CRM that captures every customer interaction and then reasons over all of it.

Over 2,000 startups have signed up since November.

Launched on Product Hunt today — already #1.

Get 3 months free with code AIODYSSEY23 →

Every AI lab is running into the same issue right now. The best models are powerful, but they’re too slow and too expensive to run everything through.

This release is OpenAI’s way around that.

Instead of relying on one model, GPT-5.4 now splits the work. The main model figures out what needs to happen, then smaller models handle execution in parallel. Different parts of the task get handled at the same time, not one after another.

It’s a small change in how it’s described, but a big change in how it actually runs.

The surprising part is how little performance drops. Mini gets close to the full model on coding and computer-use tasks, while running more than twice as fast. Costs come down sharply too, with nano designed for high-volume, lightweight work like classification and extraction.

Quick snapshot:

It’s already rolling out across ChatGPT, the API, and Codex. For most users, the change will feel simple. Things just respond faster.

Under the surface, though, something more important is happening.

This removes the need to think about models at all. You’re no longer deciding which one to use or how to structure the task. The system breaks it down and handles it for you.

Our take

This isn’t really about mini models.

It’s about moving from a single model to a system that coordinates multiple ones. That’s how you get speed and cost down without losing too much quality.

The catch is that OpenAI still isn’t the cheapest option in this category. Others are pushing lower pricing or better efficiency.

So the win here isn’t price.

It’s control over how the work gets organised.

And that’s where things are heading.

Something changed — fast.

Anthropic is now capturing over 70% of spend from companies buying AI tools for the first time, according to Ramp data. Just a few weeks ago, it was neck and neck with OpenAI. Before that, OpenAI was ahead.

That’s a sharp swing.

This isn’t about model quality anymore. It’s about who’s converting interest into actual revenue, especially with enterprise buyers.

OpenAI still leads on scale. It’s on track for ~$25B in revenue this year, ahead of Anthropic’s ~$19B. But a lot of its consumer usage is subsidised, which makes the economics harder.

Anthropic’s growth looks different. More focused, more enterprise-heavy, and right now, moving faster.

At the same time, OpenAI is reportedly rethinking its direction, pulling focus back toward enterprise after pushing into more consumer products.

Even then, most companies aren’t ready to commit. Many are still using multiple models, waiting to see how things shake out.

Bottom line

The AI race isn’t just about who builds the best model.

It’s about who turns it into a business first.

If one of these stories stuck with you, I’d love to hear which one.

Speak soon,
Lavena

P.S. If you want to get a founder feature about your own story, reply to this email. If you’d like to reach our newsletter audience (founders, creators, and marketers), click the button below.

If you’re new here, I’m over the moon you’ve joined us! To help me craft content that’s actually useful (and not just noise in your inbox), I’d love it if you took 1 minute to answer this quick survey below. Your insights help shape everything I write.

Dive into our other newsletters - where psychology meets the creator economy, e-commerce marketing, and Human resources.