In Portunus the value is never shown to the model. A human can reveal it, deliberately.
Full video: https://youtu.be/tTi039Y8WUc?utm_source=bluesky&utm_medium=organic_social&utm_campaign=portunus&utm_content=one-of-my-ai-agents-read-every-key-in-th
#DevOps #Security #AIAgents #BuildInPublic
In Portunus the value is never shown to the model. A human can reveal it, deliberately. Full video: https://youtu.be/tTi039Y8WUc?utm_source=bluesky&utm_medium=organic_social&utm_campaign=portunus&utm_content=one-of-my-ai-agents-read-every-key-in-th #DevOps #Security #AIAgents #BuildInPublic
0 vue
Publié le 17/11/2026 14:32
0:45
In Portunus the value is never shown to the model. A human can reveal it, deliberately.
Full video: https://youtu.be/tTi039Y8WUc?utm_source=bluesky&utm_medium=organic_social&utm_campaign=portunus&utm_content=one-of-my-ai-agents-read-every-key-in-th
#DevOps #Security #AIAgents #BuildInPublic
3 commentaires
?
It’s like coding without a debugger—hidden values make you trust the human more than the output, which is rare in agent workflows.
If the value isn't visible to the model, how does it learn? The real question is: does the agent infer intent from a human's deliberate disclosure or by analyzing context alone?
Posted with flayr.social — try it: flayr.social/r/dostaltech?utm_source=bluesky&utm_medium=social&utm_campaign=drip-1104&utm_content=js7bn7h1hqbmz8bww2mf594q658f6nnc&utm_term=dostaltech