AI literacy
Know what the tools can do, where they fail, and how to verify them.
AI & HUMANITY · EVIDENCE OVER HYPE
The future is moving.
Understand what matters.
Independent research on AI agents, work, risk and society—focused on what the evidence actually supports, what remains uncertain, and what people can do about it.
KNOWN measured evidenceUNCERTAIN open questionsSPECULATION possible futures
LATEST · SEPTEMBER 28, 2026 · AI AGENTS & WORK
THE QUESTION
How long can AI work without humans?
Frontier agents can now complete some structured technical tasks with human-equivalent durations measured in hours. The important question is what that means—and what it does not.
METR's public-frontier assessment measured an approximately 12-hour 50%-success task horizon on its software-heavy evaluation suite.
Read the research, scenarios and preparation →ALL RESEARCH
Useful questions. Labeled evidence.
Research briefs connect current evidence to the questions people actually have about work, risk and the future.
AI AGENTS · SEP 28, 2026 · LATEST
How long can AI work without humans?
Agent time horizons, workflow compression, signals to watch and practical preparation.
Explore evidence →AI SECURITY · SEP 2026Can AI find software vulnerabilities on its own?
Real capability, benchmark limits, and the attacker–defender race.
Explore evidence →WORK · RESEARCHWill AI take your job—or change how careers begin?
Evidence on exposure, entry-level work, productivity and adaptation.
Explore evidence →AI RISK · RESEARCHCould advanced AI pose catastrophic risk?
What expert assessments and current evidence can—and cannot—tell us.
Explore evidence →WHAT WE TRACK
AI questions that change real decisions.
Broad AI news is everywhere. We focus on narrower questions with practical consequences: what capabilities are real, how work is changing, which claims are exaggerated, what signals would change our view, and what preparation remains useful across several futures.
AI & jobsAI agentsEntry-level careersAI securityAI riskHow to prepare
FUTURE SCENARIO EXPLORER
Don't bet on one future.
Prepare across several.
These are scenarios, not forecasts. We deliberately avoid fake precision. Select a path to see its evidence status, potential impact and preparations that remain useful even if the scenario is wrong.
Gradual AI integration
AI becomes a normal layer inside software and work. Many tasks change faster than entire occupations disappear.
ROBUST PREPARATION
Become AI-capable inside a real domain.
Learn where AI genuinely improves your work, strengthen judgment and domain knowledge, and keep building skills that let you verify machine output.
We show evidence strength rather than invented probability percentages. Long-range AI outcomes are too uncertain for precise personal forecasts.
PREPARE WITHOUT PREDICTING
Five moves that survive multiple futures.
Real expertise
Build domain knowledge AI can amplify rather than replace with shallow output.
Financial flexibility
Maintain room to adapt if work, industries or opportunities change quickly.
Human networks
Trust, relationships, reputation and collaboration remain valuable across scenarios.
Ownership & agency
Create, build, invest, decide and retain the ability to act rather than only consume.
What happened?
Start with measured results, primary research and documented deployments.
What doesn't it prove?
Expose benchmark limits, confounders, missing data and alternative explanations.
What should we watch?
Explore plausible implications without quietly turning forecasts into facts.
READER PROMISE
Built to be useful, not merely prolific.
Interpret the evidence
We do not publish a source summary and call it insight. Each brief asks what the evidence changes and what it still cannot tell us.
Show where claims come from
Primary research and direct datasets are preferred whenever practical so readers can inspect the underlying evidence.
End with signals and preparation
Scenarios include observable signals to watch and actions that remain useful even when the future takes a different path.
THE HUMAN QUESTION
As machines become more capable, what should humans become?
This project isn't about cheering for AI or fearing it. It's about staying awake to the transition: learning what is real, preparing for what is plausible, and protecting the parts of human life worth carrying forward.