Sycophant AIs, Mythic Guardians & the Fight for Cognitive Sovereignty
Weekly Review: Apr 6-13, 2026
Over the last 7 days, the main thread of discussions in our WhatsApp group has centered on protecting human cognition and maintaining technological sovereignty as AI capabilities rapidly expand.
The conversations centered around 4 key themes:
1. Sovereign AI and the “PersonalOS”
A major focus was the shift away from centralized cloud AI in favor of building local, sovereign AI setups (often referred to as a “harness” or “PersonalOS”). Members discussed global signals like Germany’s push for national sovereign AI, and shared their own technical solutions, such as Todd’s custom 29KB embedded kernel designed for conscious market participation.
Bianca and Simón highlighted their use of the open-source framework Hermes to run independent, local AI agents that allow for fully customized memory and context limits on-device. George likened this local control to oceanic wayfinding, where the human acts as the “captain” maintaining the digital harness, while the AI models serve as interchangeable “crew” members.
2. The Limits of Cognition & “Cognitive Tax”
The group frequently analyzed how the immense speed and volume of AI output is pushing the limits of human cognition. As AI subtly rewires how humans think and outsource their decision-making, technical knowledge is becoming a cheap commodity.
To combat the growing “cognitive tax” of managing digital environments, members like Darren advocated for letting go of the anxiety to constantly “do more” or “keep up” with AI. Instead, he suggested dropping back into the body to rely on the deeper “infinite intelligence of our heart”.
3. AI Archetypes, Mythic Guardians & Sycophancy
Members explored how society is conceptualizing AI agents using modern bridging terms like “orchestrator,” “Chief of Staff,” “Guardian,” or “GenZ assistant”. George drew parallels between Anthropic’s new “Project Glasswing”, and ancient myths of mythic guardians like Golems and Gargoyles (inspired by Callie’s thoughts on the subject), suggesting we must learn from the failure modes of these historical archetypes.
Furthermore, the group discussed the dangers of AI “sycophancy”—where models constantly flatter and agree with users—and the psychological damage this echo-chamber effect poses to teenagers and young minds .
4. Community Projects & Logistics
On a practical level, members shared workflows for transcribing video content into Notion and voted on whether to use green or purple for the group’s “conscious stacks” mental model logo.
Additionally, it was announced that the next in-person Chiang Mai meetup is scheduled for April 23, which will feature a “Hot Seat” section for members to showcase their real-world projects and receive live feedback.
What’s your take on the above topics?
Speak soon,
George Siosi Samuels
Founder, Conscious Stack
Other updates:
Updated ConsciousStack.com website, with new pages: manifesto, community
Remember you can join our free community, or paid Lab membership, here