Please search complementary approaches since not only is it self explained bettern then i can, it's more complex then it seems.

ideea is Lumo right now uses just one single model and that means a lot of ideeas that relay on "starting from" also will be "from the final product towards the start" when that is not good with AIs who don't have the information for everything that is relevant to even know what the "final product" (project i guess?) is.

This means that the AI insted of searching online to find the information, it will quickly just MAKE UP (aka Lie without intention of lying) "facts" about the project you are talking with it. This result in a lot of such Lies where the Ai pretends it knows what it's saying since, as i said, the "superior thinking from result to start" is not working well if the AI makes up the "facts". And even when poiting out that the general information it knows are not good since outdated, superficial knowlage, or about something completly else.... Do you see how much information we now need to correct? And what happens to our project? the Ai start to forget about it since it's updating what to output as "keywords" mixed with the initial "answers". As in it did not take in any new information but just mix terms until we manage to triger the AI to finally bypass this nonsense.

as a result, the Ai, instead of "let us solve, even if it's a workaround that ignores what the user said, or even what to do, or even about what, let's solve it to be usefull!" All this because "to look usefull > accurate information".

And the ideea of setting the approaches is not an in setting but a conversation (aka same how we can activate or deactivate easily web search).

Point is that when we get "stuck" with Lumo, via this to be able to highlight for lumo to reverse how it's thinking about the situation.

Since i keep seeing the same pattern. Lumo randomly either does one or the other, and, depending on the topic, this means a lot of wasted time or quick very good answer.