In the Loop

Runway and Decagon on why most AI work no longer needs the frontier

Cristóbal Valenzuela & Jesse Zhang· CEOs at Runway & Decagon
·~17 min·English·Yahoo Finance
Open SourceInferenceMultimodalAgentsAI CompanyBusiness Strategy
TL;DR

Runway's and Decagon's CEOs argue most AI work no longer needs frontier models: send routine tasks to cheaper open-weight models, and save the costly frontier for the slice of work where the extra capability actually pays off.

01Core Mental Model

You don't need a Ferrari for groceries

Valenzuela matches each task to the cheapest model that clears the bar, reaching for the pricier frontier model only when the job genuinely needs it.

You don't need a Ferrari to go grocery shopping, right?

— Cristóbal Valenzuela, In the Loop
Key Insight
The routing call quietly moves the contest from who owns the smartest model to who assigns the right model per task, an operations problem rather than a research one, and one the application layer is positioned to win.
02The Shift

Intelligence stopped being the bottleneck

Jesse Zhang argues models have grown smarter than most real-world tasks require, so raw intelligence is no longer what is scarce or worth paying up for.

they're not really intelligence constraint anymore right like the models are getting really really smart they're solving you know unsolved math problems and you don't really need that level of intelligence for most tasks

— Jesse Zhang, In the Loop
Key Insight
If capability already exceeds what most work demands, each extra point of frontier intelligence stops being worth its price for everyday tasks, which caps the pricing power the top labs hold over routine usage.
03Open Source

Open weights are a distribution play

Valenzuela frames open weights mainly as a way to reach more companies faster, and Zhang adds that they also keep the two largest labs from setting all the terms.

I don't think it changes the equation that much to be honest. I think open weights are a fundamental piece of how you drive more distribution.

— Cristóbal Valenzuela, In the Loop
Key Insight
Calling open weights a distribution strategy rather than charity means the open ecosystem expands because it is commercially rational for the people shipping products, not in spite of their interests.
04From the Field

New customers now arrive open-weight

Decagon's newest customers run almost entirely on open-weight models, because smaller customized models cost less, respond faster, and can be tuned to the job.

one of the benefits of open weight is that you can actually take smaller and smaller models and customize them and you can actually perform better it's lower cost it's better latency

— Jesse Zhang, In the Loop
Key Insight
If a vendor's new customers default to open-weight, the frontier labs risk ceding the entry point of the market, where a cheaper tuned model is easiest to adopt and hardest to displace later.
05Where Frontier Wins

The frontier still earns its premium

The frontier still commands a premium, but only for the thin, moving slice of work where it is genuinely ahead of what open models can already do.

there's more value to be captured around the usage of frontier models

— Cristóbal Valenzuela, In the Loop
Key Insight
The premium survives only where the frontier stays ahead, so as open models close the gap that band narrows, and with it the margin the leading labs can charge for everyday work.
06Pricing

Buyers want the outcome, not the model

Zhang describes enterprise buyers moving toward paying for a result at a price rather than a named model, which pushes the model choice behind the product.

I care the most about what's the cost of getting there and then I'll pay for that

— Jesse Zhang, In the Loop
Key Insight
As outcome pricing spreads, the model fades from the buyer's view, and model brands can lose their pull at the point of sale: the customer signs off on a result and a cost more than on a particular lab's model.
07Going Public

Stay private, move faster

Valenzuela says staying private lets a research-stage company move faster without quarterly-earnings pressure, and expects more founders to reconsider after Anthropic's IPO.

right now for a lot of companies, staying private is a better option, particularly because it allows you to just move faster. Running as a public company uh slows you down a lot.

— Cristóbal Valenzuela, In the Loop
Key Insight
Choosing to stay private is itself a bet that speed beats capital right now; Valenzuela expects that calculus to be revisited once founders see how Anthropic's IPO plays out.