the_confused_teacher

Teacher, futurist, hopelessly clumsy programmer

[the_confused_teacher @AlephOneAI]

@AiroPrompts No.

[the_confused_teacher @AlephOneAI]

@petergyang @pmarca BBBs

[the_confused_teacher @AlephOneAI]

@mark_k Has been a long road since cygwin.

[the_confused_teacher @AlephOneAI]

@ThrillaRilla369 Not just that. I would even call for caligraphy classes.

[the_confused_teacher @AlephOneAI]

But is computer-use (with real clicks) the right approach? I envision more a world where Apps expose more and more its public methods (e. g. via MCP or whatever) and then we could get rid of all or at least large chunks of the UI interactions. AI systems only need to call then the specific functions.

[the_confused_teacher @AlephOneAI]

@realaicoach But for all small problems this takes longer than building your own version. Today it feels like it takes forever with registration etc.

[the_confused_teacher @AlephOneAI]

Just for fun, I built a C# attribute framework on top of MCP so any UI method or event can be invoked directly by an LLM.

Results either get pulled via follow-up calls or — in the Claude Code CLI — passed straight back to the model through a channel, which can then kick off the next task.
Naturally, it also doubles as a regular unit testing toolkit.

[the_confused_teacher @AlephOneAI]

@LolandBeast67 @TrisH0x2A Agreed. But how do I know that? I can only guaratee that when I do the Memory management by myself. Is that the use case?

[the_confused_teacher @AlephOneAI]

But what do we expect from a next level OS? Maybe true birectional voice chat like with a real person. Some BCI support? I.e. colors and wallpapers are selected based on your mood?

Maybe a big change would be when we could get rid of Setting menus. All is just managed by Chats and agents.

[the_confused_teacher @AlephOneAI]

@MatthewBerman @lennysan Yes it is.

[the_confused_teacher @AlephOneAI]

@moonfarm_dev Annual is too risky for a new unknown service ( from a costumer Perspektive)

[the_confused_teacher @AlephOneAI]

@TolentinoTeach True.

[the_confused_teacher @AlephOneAI]

@mark_k Of course. And I like Emad a lot. Seems to be a smart and nice guy.

[the_confused_teacher @AlephOneAI]

@TolentinoTeach I have never heared the complicated argument. But the inefficient argument a lot. People around me want to speed up education but this is not really possible.

[the_confused_teacher @AlephOneAI]

@QuanticASI The only thing your mind sees is now. That means you have no insight into the Reality of time. Only some Math we made up for predictions.

[the_confused_teacher @AlephOneAI]

@moonfarm_dev AFAIK cold outreach in Europe is a regulatory nightmare. At least skip that one.

[the_confused_teacher @AlephOneAI]

@ChShersh @piliponful But recursion is slow. For-Loops and Stacks for the win 😊

[the_confused_teacher @AlephOneAI]

@espeed @GaryMarcus Thats works great for a moderate number of tables. Up to hundreds of tables it works much better then one might think. However we tested also a really large and badly structured DB too with close to 2500 tables and it was a complete disaster. GPT 5.2 at that time.

[the_confused_teacher @AlephOneAI]

@kimmonismus Where did get the Mid-Tier info?