5 Comments
User's avatar
Aadarshkumar's avatar

Interesting part isn’t who people trust.

It’s why.

Most users don’t evaluate models directly.

They evaluate outcomes.

If outputs feel:

→ inconsistent

→ overconfident

→ hard to verify

trust drops, regardless of the company.

And right now, most people are still using AI in a way that produces exactly those problems.

Single prompt.

Single output.

No validation.

So the experience feels unreliable.

The shift happens when usage changes:

→ structured prompts

→ cross-checking

→ multi-tool workflows

That’s when AI starts feeling trustworthy in practice.

Been working on a system around that shift:

https://shorturl.at/nE0Tw

Heather Baker's avatar

Thanks this is really interesting!

Aidan Dunphy's avatar

Heather, have you looked into GB1 (formerly Locai) by Locai Labs? They're focused on UK-sovereign AI, and have just announced Project Mercury to build the UK's first sovereign frontier models? https://locailabs.com/blog/locai-labs-and-civo-unveil-project-mercury-the-uks-first-family-of-sovereign-frontier-models

I think we all need to plan for disentangling ourselves from OpenAI and maybe even Google.

Heather Baker's avatar

Oh wow, thanks for flagging this to me. I am definitely going to check this out. Looks fascinating!

Phoebe's avatar

I wonder how different this is from consumer to business or company to model (Do people who trust Google also trust Gemini?). I'll be paying attention to this each month, thank you for sharing!