Should end users manually choose which AI model to use, or should the system auto-route without exposing the choice?

Harry Stebbings · Mar 3, 2025 · hedged

Within three to five years users will no longer select which model they use, in the way one doesn't select which Google to use.

Model choice is an artifact of current specialization that should disappear as the experience matures.

Scope: invites correction: 'Am I completely wrong or do I completely miss the point?'

18:01 20VC: Anthropic CPO Mike Krieger: Where Will Value Be Created in a World of AI | Have Foundation Models Commoditized | When Do Model Providers Become Application Providers | What Anthropic Learned from Deepseek

Harry Stebbings · Apr 11, 2025

It is preposterous that users are made to choose their model, and in five years' time users will have no choice of which model they use

Model names are ridiculous and users have no idea what any of them are

40:27 20Product: How Scale AI and Harvey Build Product | Why PMs Are Wrong: They are not the CEOs of the Product | How to do Pre and Post Mortems Effectively and How to Nail PRDs | The Future of Product Management in a World of AI with Aatish Nayak

Aatish Nayak · Apr 11, 2025

On average humans are lazy and need to be told what is best for them, so the model-picker dropdown paradigm was a trap the whole industry fell into

The paradigm was created for a Silicon Valley/tech audience that enjoyed playing with different models, and nobody — including Harvey — has figured out which model is best for which job well enough to justify exposing the choice

Scope: on average

40:57 20Product: How Scale AI and Harvey Build Product | Why PMs Are Wrong: They are not the CEOs of the Product | How to do Pre and Post Mortems Effectively and How to Nail PRDs | The Future of Product Management in a World of AI with Aatish Nayak

Harry Stebbings · Apr 18, 2025

Making users pick which model to use for a prompt is archaic; in future the best model for each request will be selected automatically

It is like telling a car mechanic which spanner to use rather than trusting them to pick the right tool

36:04 20VC: Foundation Models: Who Wins & Who Loses | How Economies and Labour Markets Need to Change in a World of AI | China vs the US in an AI Race: What You Need to Know | Rich Socher, Founder @ You.com

Richard Socher · Apr 18, 2025

Model orchestration should be automated away entirely, with manual model choice reserved as an optional interface for power users

Scope: power users retain the ability to double-click in and choose

36:21 20VC: Foundation Models: Who Wins & Who Loses | How Economies and Labour Markets Need to Change in a World of AI | China vs the US in an AI Race: What You Need to Know | Rich Socher, Founder @ You.com

Harry Stebbings · Jul 28, 2025

Requiring users to choose which model to use is ridiculous; the system should figure it out for them

The proliferation of model variants is absurd

43:18 20VC: a16z's Martin Casado on Anthropic vs OpenAI: Where Value Accrues | Cursor vs Replit vs Lovable: Who Wins and Who Loses | The One Sin in AI Investing | Why Open Source is a National Security Risk with China

Aatish Nayak · Apr 11, 2025

Making users choose between an assistant mode and a draft mode was a product mistake; the right UX is for the AI to infer intent, ask clarifying questions, and route the query to the best system itself

Almost every user was confused about when to use which mode, and enablement only reduced complaints because people got tired of complaining

38:29 Auto inferring user intent beats forcing a mode choice

Harry Stebbings · Apr 11, 2025

Giving users choice is often bad product design because humans are lazy and get confused by too many options — they should just be told which option to use

The paradox of choice: users get confused when they have too much choice

40:46 Paradox of choice means users should be told not asked