OP here -- I think the general principle I would recommend is using a big reasoning model for the planning phase. I think Claude Code and other agents do the same. The reason this is important is because the quality of the plan really affects the final result, and error rates will compound if the plan isn't good.
based on the article, it seems like a good reasoning model like gpt5 or opus 4.1 might be good choices for the planner. I wonder if the gpt oss reasoning models would do well
Personally been using GPT-OSS-120b locally with reasoning_effort set to `high` and it blows pretty much every other local model out of the water, but takes a lot of time for it to eventually do a proper content reply. But for fire-and-forget jobs like "Create a well-researched report on X from perspective Y" it works really well.
Gemini is great, it's just incredibly clumsy at tool use and that's why it fails so often in practice. I'm looking forward to the next version, it will for sure address it, it's a big issue internally too (I'm a recent xoogler).
Yes it really is horrible at using tools. Codex is way better (even better than Claude code ). Gemini is great at doing audits and content (though I’ve switched to codex for everything all in one).
We have been working on this problem since 2020 and have created an trained an ensemble of AI detection models working together to tell you what is real and what is fake!
First, there are multiple solutions to this problem, all need to be explored and many will have a part in the future. Bad actors will do everything they can to find a way around every solution.
Second, for the hashing approach, have a look at "Perceptual Hashing", it's a way to hash content like a an image, even if the resolution changes.
Where to store the hashes? A centralized server is probably fine, but there is always a risk of a bad actor exposing it somehow. A secure blockchain can work better. But if you go that route, might as well go with the most secure blockchain. POW is generally more secure than POS. And currently the most secure blockchain is Bitcoin. So one solution is to batch hashes together and write them with a Bitcoin transaction in some cadence.
Great discussion. Many have asked the same questions.
We hope a standard will be created and used by all digital content creation tools. But this will take time. And even then, bad actors will not be deterred. They will always find ways to create fake content and pass it as authentic. We want to be there to fight them every step of the way!
We hope a standard will be created and used by all digital content creation tools. But this will take time. And even then, bad actors will not be deterred. They will find ways to create fake content and pass it as authentic. We want to be there to fight them every step of the way!
