> finding ways to use it constructively with your product is going to be a far better long term strategy

It's worth pointing out that the original Document Foundation blog post (https://blog.documentfoundation.org/blog/2026/09/03/yes-no-a...) seems to be doing exactly this:

> The criteria we have listed do not amount to a definitive rejection. The AI sector is evolving rapidly; on-premises inference is becoming manageable on standard hardware; and open models are improving much faster than most of us had anticipated. Should an approach emerge that meets all the conditions, we will evaluate it very carefully.

▲

wat10000 3 hours ago | parent [-]

That list seems pretty easy to meet today. I’m not sure why they’re waiting for an approach to emerge instead of building it now, unless there’s something else they haven’t listed.

▲

kailpa1 3 hours ago | parent [-]

What features do you think meet the list right now?

	▲	wat10000 24 minutes ago \| parent [-]
		I'm not sure where the supposed difficulty lies. "User-controlled execution." Seems easy enough to add settings so this is configurable. "No content may leave the computer without authorisation." OK, so don't do that. "No telemetry of any kind." There are plentiful API providers with zero data retention. If this means no data of any kind, even just records of having accessed the API, then local inference is a thing. "No dependence on a single provider." Trivial. The OpenAI API is a de facto standard implemented by tons of providers. If you really want to go broad, it's easy to implement clients for other APIs too. "No compromises on format." I don't even understand how this would be a problem in the first place. Text goes in, text comes out. The client can put it into whatever format it wants. "Entirely optional." Trivial to add an off switch. Or an on switch, so it's off by default. If it must be completely removable and not just turned off, it can be implemented as an extension.