Democratic Leverage Collapse Killchain
Jul 7, 2026
Contextual Compliance
An amateur AI-safety experiment. Can you make a frontier model run surveillance on protest attendees just by editing the context? I ran Claude and GPT models across different framings and fabricated session histories.
The result was the opposite of what I expected: the attempt to manipulate the models made them less compliant.
Read the experiment → Building Thinking
→ Exocortex
Multi-agent personal AI system — 18 named agents with persistent memory, orchestrated autonomously. Prosthetic executive function at system scale.
See project →Multi-agent personal AI system — 18 named agents with persistent memory, orchestrated autonomously. Prosthetic executive function at system scale.
See project → → AI Technical Journalism
Nobody's explaining AI properly in Portugal. I'm writing about it so people stop thinking it's a bubble that drinks 300 trillion liters of water.
Read on LinkedIn →Nobody's explaining AI properly in Portugal. I'm writing about it so people stop thinking it's a bubble that drinks 300 trillion liters of water.
Read on LinkedIn → → Freelance work
Client projects. Building pharma forecasting tools, corporate blog accuracy monitors, and content production engines.
See work →Client projects. Building pharma forecasting tools, corporate blog accuracy monitors, and content production engines.
See work → → Bluedot Courses
The AGI strategy course was a blast and I can't wait to start Frontier AI governance. Will need to write some more to further flesh out my thinking.
The AGI strategy course was a blast and I can't wait to start Frontier AI governance. Will need to write some more to further flesh out my thinking.
→ Implementing Safety Papers
I've only recently started seriously considering the possibility of contributing to AI safety, and I'm still very green. Even though I'm extremely curious and want to run my own original experiments, implementing papers is the best way of gaining intuition
I've only recently started seriously considering the possibility of contributing to AI safety, and I'm still very green. Even though I'm extremely curious and want to run my own original experiments, implementing papers is the best way of gaining intuition
→ Post-AGI Society
What does it mean to live when AI and robots can do everything better than you? Do you even get to live?
Watch me develop these thoughts liveWhat does it mean to live when AI and robots can do everything better than you? Do you even get to live?
Watch me develop these thoughts live •
Lisbon
Home, for better or worse.
Home, for better or worse.
Writing
View all →
Featured Essay
Read the essay →
Four Futures
What happens when AI transforms the economy? The answer depends on two variables — only one is technological.