Generate
Competing hypotheses
Ask for alternative explanations, edge cases and questions you have not considered.
AI & Information · Research direction
AI can make you faster. It can also make a wrong model feel complete.
AI increasingly mediates what reaches us as information. Reality Skill uses it as a subordinate critical layer without silently outsourcing judgment. The Founding Reality Lab starts by establishing a human baseline, then measures where AI genuinely improves the process.
AI and the information layer
More fluent output does not remove the need for judgment.
Why this matters
AI changes the path between reality and your decision.
Underlying realityEvents, people, systems, primary data, physical constraints.
Information layerPublishers, search, communities, algorithms and source incentives.
AI-mediated contextSummaries, generated explanations, recommendations, synthetic content and agents.
Your modelWhat you believe happened, why, and what matters.
Your decisionAction, commitment, update trigger and later outcome.
A fluent AI answer is an information event. It is not automatically evidence.
Useful jobs
Use AI where abundant cognition helps without granting it automatic authority.
Challenge
Red-team a human-first view
After you record your own reasoning, ask AI to find unsupported assumptions, missing base rates and disconfirming evidence.
Compress
Reduce search cost
Map a domain quickly, then follow decision-relevant claims back to primary or independent sources.
Research later
Private/local tools
Local transcription, memory retrieval and a live process dashboard remain research modes rather than promises of the founding cohort.
AI Reality Check
Six questions before an AI answer is allowed to materially change an important decision.
01
What is the claim?
Separate fact, inference, recommendation and generated hypothesis.
02
Where did it come from?
Can the important claim be traced to primary data, an official record or credible independent evidence?
03
What incentives shaped the source?
Who created the underlying information, and what do they gain from the framing?
04
What is the strongest alternative?
Generate the best competing explanation, then look for evidence that distinguishes the models.
05
Can I verify laterally?
Leave the answer and check independent sources rather than judging credibility from presentation quality.
06
Would this change my action?
If the claim cannot change the decision, stop spending attention. If it can, raise the verification standard.
Research sequence
AI should earn its place in the protocol.
Now
Human baseline first
Establish the human baseline first. Learn what structured independent judgment and the Scored Journal accomplish before attributing any improvement to AI.
Later
Measure marginal value
Test AI red team or a local process copilot against the human baseline. Log interventions, contest them, and retain them only if they demonstrably add value.
Research, not product promise