AI Coding: Cognition Prioritizes Speed Over Intelligence

For Kaplan, president and co-founder of Cognition, many software engineering tasks have reached a point of "intelligence saturation." He argues that the criteria for selection are shifting towards speed and cost, based on internal data. According to him, a 10-point advantage in intelligence would now only yield a 20% increase in traffic, compared to nearly 69% at the end of 2025.
A Strategic Shift Announced for Coding AI Providers
Kaplan believes that the current evolution could lead to a significant strategic change for AI companies looking to attract software developers and secure enterprise contracts. Players like OpenAI, Anthropic, Meta, and Google have so far emphasized the performance of their coding tools to promote their models. According to Kaplan, as AI agents become more proficient at routine tasks, the intelligence superiority of a model may weigh less in user choice than lower cost or faster performance. He asserts that this trend is confirmed by Cognition's internal data.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Software Engineering Tasks Said to be "Intelligence Saturated"
Kaplan describes that, for many software engineering tasks, AI models have reached a level he refers to as "intelligence saturation." He explains that once a model performs a task correctly, the determining factors become cost and speed. In this context, he believes that attention may shift towards these two aspects, and states that competition among AI models for coding is now more focused on speed than on intelligence.
Advanced Internal Indicators to Support the Shift
To support his claims, Kaplan cites internal data: a 10-point advantage in model intelligence now generates only a 20% increase in traffic for developers, whereas at the end of 2025, this figure was close to 69%. He links this evolution to the improvement of AI model benchmarks and observes that smarter and more capable models are no longer attracting as many users as before. He articulated these observations in a post published on Monday, reminding that the race for AI has long favored the smartest models. Kaplan speaks as president and co-founder of Cognition, the company behind Devin.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.