
This Week in Startups
Odyssey CEO Oliver Cameron is 90% sure we live in a simulation | E2358
This debrief was written and voiced by AI from the episode's audio.
🎧 AI Debrief
🎙️ Full Episode
Key Moments
Elon Musk announces GrokBot usage limits citing unsustainable serving costs relative to revenue; Jason argues the backlash is misguided given the actual ROI of AI tools.
Jason quantifies his actual return from AI tools: $40K annual spend automating seven full-time jobs and dramatically accelerating workflows. He shares real examples of GrokBot designing merchandise and replacing project managers.
Odyssey CEO Oliver Cameron boldly states he's 90% confident reality is a simulation. He explains Odyssey's fruit fly brain experiment and connects it to simulation theory.
Odyssey placed a fruit fly's mapped brain inside an AI-generated environment running in real time. The fly behaved like a real fly. Six million people thought it was real video.
Oliver explains Odyssey's business model: sell foundation world models to robotics and autonomous vehicle companies, who then train them on 20-30 hours of task-specific footage for deployment.
Jason argues orbital data centers eliminate terrestrial opposition by leveraging low Earth orbit's 73% additional surface area, zero real estate cost, and solar power. Starship enables viability.
Jason explains his strong preference for Waymo robotaxis over human drivers: quieter, no small talk, no tipping, no uncertainty. Consumer adoption is accelerating.
Anthropic researcher Andrej Karpathy used Claude to build HackerTube, curating 6,454 videos from 43 million Hacker News comments. It took two hours and shows how AI is enabling custom software at scale.
Jason pitches his new tactical newsletter on Substack and X, calling for 1,000 paying subscribers to fund a full-time writer. Subscribers get early access, no ads, and exclusive content.
Jason and Oliver discuss quantum observation and the double-slit experiment as evidence of simulation. If observation changes reality, maybe our universe is a simulation optimized based on observer presence.
Search Episode
Summary
This Week in Startups opens with Jason Calacanis excited about GrokBot's explosive growth and Elon Musk's announcement of usage limits due to unsustainable costs. Jason shares his personal ROI from AI tools—spending $40K annually to generate $2M+ in value by automating seven full-time positions and accelerating workflows. The episode then pivots to Oliver Cameron, CEO of Odyssey, who demonstrates their world model technology by placing a simulated fruit fly brain in an AI-generated environment. Cameron reveals he's 90% certain we live in a simulation, sparking a fascinating discussion about simulation theory, the double-slit experiment, and the computational feasibility of simulating reality. The conversation covers Odyssey's business model (selling foundation models to robotics and autonomous vehicle companies), the limitations of visual-only learning, and how world models could eventually enable a billion robots. The show concludes with segments on data center politics, AI's impact on employment, Waymo's robotaxi superiority, and Google's new space-based data center prototype—suggesting that orbital infrastructure may be the inevitable solution to earthbound resistance.
Key Takeaways
AI ROI is staggering: Jason demonstrates $50 in value per $1 spent by replacing 7 FTEs with $40K annual AI tool spend; aggressive pricing increases are justified by actual returns
highWorld models are the emerging foundation model, training on visual data to predict future states and enabling robotics to learn physics, dynamics, and cause-effect from minimal human demonstration
highSimulation theory gains plausibility as we build increasingly capable AI systems; Oliver Cameron (90% confident we're in a simulation) argues that low fidelity in physics (dark matter, space exploration limits) mirrors computational optimization in simulations
mediumOrbital data centers eliminate terrestrial opposition; low Earth orbit provides ~73% more surface area than Earth with zero land-use conflict, energy and real estate costs are effectively free, making space the inevitable endgame for compute infrastructure
highAI agents will become ubiquitous; platforms resisting automation (LinkedIn, etc.) risk obsolescence—monetizing bot access via throttling is preferable to fighting adoption
mediumRobotaxi adoption is accelerating in Texas; preference for Waymo over human drivers (no small talk, cleanliness, no tipping) signals rapid consumer acceptance of driverless transportation
mediumJob displacement is real and concentrated; unless housing, education, and healthcare are addressed, political backlash will slow AI deployment regardless of technological capability
highNotable Quotes
“For every dollar we spend, we're getting $50 of output.”
— Jason Calacanis
Quantifying the ROI from AI tools ($40K spend → $2M value); establishes why premium pricing is justified and users should upgrade rather than complain
“I'm 90% certain that this is a simulation, a very advanced simulation of some sort.”
— Oliver Cameron
Direct answer to the simulation theory question, making Odyssey's CEO's convictions explicit and raising the stakes of the world model conversation
“If you have the technology, someone would be like, oh, let me put 6,000,000,000 human simulated brains in here and see what happens.”
— Jason Calacanis
Frames simulation theory through the lens of human motivation; once we build the capability, we'd almost certainly use it, making simulation plausible
“The observer changes reality.”
— Jason Calacanis
On the double-slit experiment and quantum observation; suggests universe design may be optimized based on observer presence, lending credence to simulation hypothesis
Predictions
Space-based data centers will become economically viable and dominant by the 2030s, making terrestrial opposition moot
— Jason Calacanis · Confidence: high · 2030s
AI will be the signature issue of the 2028 US presidential election; both parties will compete on being 'most pro-human and anti-superintelligence'
— Jason Calacanis · Confidence: high · 2028
Waymo and autonomous vehicles will continue rapid expansion in Texas, with consumer preference for robotaxi over human drivers becoming the norm
— Jason Calacanis · Confidence: high · Next 2-3 years
Odyssey's world model technology will power a billion robots within 5-10 years; post-training on foundation models will become the standard practice
— Oliver Cameron (implied) · Confidence: medium · 5-10 years
Platforms that resist bot/agent automation (like LinkedIn) will lose competitive advantage; those that monetize agent access will thrive
— Jason Calacanis · Confidence: medium · 1-2 years
Action Items
Audit your AI tool spend and calculate actual ROI (cost per hour saved × hourly rate); upgrade to higher plans if the math supports it (50:1+ payoff)
Most users underestimate AI value and complain about pricing without doing the math; Jason's framework ($40K → $2M) provides a template
Build custom interfaces and local-machine tools (e.g., ad-free Amazon, curated YouTube) using Claude or similar; don't wait for companies to build them
Recent examples (HackerTube, ad-free Amazon extension) show 2-hour builds with massive UX gains; this is now a core productivity skill
Set up bot automation for routine tasks (DM responses, meeting transcription, travel booking) within your current tooling; treat agents as delegated team members
Jason is rapidly operationalizing Grok bots for scheduling, research, and content creation; this is no longer sci-fi, it's table stakes
If in job market or hiring, prepare for rapid robotaxi adoption and physical automation; invest in skills that remain uniquely human (taste, community, events, video editing)
Jason emphasizes hiring for 'uniquely human' roles while automating the rest; misalignment between roles and automation will cause friction
Subscribe to This Week in Startups' tactical newsletter (twistartups.substack.com) or email Lon@launch.co for free year; help fund full-time writer who can synthesize these trends
Jason is building a sustainable indie media model; subscriber support funds deeper analysis than free content allows
Links Mentioned
Paper: AI-native platform for startup legal work (entity formation, cap tables, contracts)
https://getpaper.ai/twist
G2i: Network of 8,000+ vetted software engineers for hiring
https://G2i.co/twist
Northwest Registered Agent: Business registration, free domain, email, hosting
https://www.northwestregisteredagent.com/twistdomain
Plaud: Recording and transcription device for meetings, calls, interviews
https://Plaud.ai/twist
Odyssey: World model foundation for robots, autonomous vehicles, drones
https://odyssey.systems
Row: GLP-1 (weight loss) service
https://row.co/twist
This Week in Startups tactical/practical newsletter
https://twistartups.substack.com