# Hypercraft Superintelligence

> Plain-English guides to where AI is heading, and a live explorer showing how fast AI is improving on five of the hardest tests.

Plain-English explainers about AGI and superintelligence, a glossary, and a
benchmark explorer showing the best AI score over time on five hard tests.
Benchmark numbers refresh weekly (last 2026-10-07). Benchmark data from Epoch AI (CC BY) and official leaderboards.

## Guides

- [What is superintelligence?](https://www.superintelligence.com.au/guides/what-is-superintelligence.md): What people mean by the word, and why it’s taken seriously.
- [AI, AGI and ASI: the difference](https://www.superintelligence.com.au/guides/ai-agi-and-asi.md): Three terms that get mixed up constantly, sorted in one page.
- [How AI progress is measured](https://www.superintelligence.com.au/guides/how-ai-progress-is-measured.md): Benchmarks, what they test, and why they stop being useful.
- [Scaling, simply](https://www.superintelligence.com.au/guides/scaling-simply.md): Why more compute and data kept making models better.
- [Alignment: keeping AI on our side](https://www.superintelligence.com.au/guides/alignment.md): The problem of getting powerful systems to do what we intend.
- [Fast or slow takeoff?](https://www.superintelligence.com.au/guides/fast-or-slow-takeoff.md): The debate over how quickly AGI could become superintelligence.

## Benchmarks

- [GPQA Diamond](https://www.superintelligence.com.au/benchmarks/gpqa-diamond.md): Graduate-level biology, chemistry and physics questions that are hard to look up. Best score today 96%.
- [SWE-bench Verified](https://www.superintelligence.com.au/benchmarks/swe-bench-verified.md): Fixing real bugs in open-source Python projects so the tests pass. Best score today 84%.
- [FrontierMath](https://www.superintelligence.com.au/benchmarks/frontiermath.md): Unpublished research-level maths problems with answers a computer can check. Best score today 94%.
- [Humanity’s Last Exam](https://www.superintelligence.com.au/benchmarks/humanitys-last-exam.md): Thousands of expert questions across 100+ subjects, chosen to stump AI. Best score today 55%.
- [ARC-AGI-2](https://www.superintelligence.com.au/benchmarks/arc-agi-2.md): Visual grid puzzles that are easy for people and new to every model. Best score today 95%.
- [Methodology](https://www.superintelligence.com.au/methodology/): how the best-score line is calculated.
- [Data (CSV)](https://www.superintelligence.com.au/data/benchmarks.csv): best score per benchmark per quarter, CC BY 4.0.

## Glossary

- [All 28 terms](https://www.superintelligence.com.au/glossary.md)

## Machine access

- [For AI agents](https://www.superintelligence.com.au/for-agents.md): how to read this site as a machine
- Markdown: append `.md` to any guide, glossary or benchmark URL
- MCP server: `https://www.superintelligence.com.au/mcp`: Streamable HTTP, public, read-only.
  Tools: get_benchmark_scores, define_term, search_content, get_page, list_pages
- [Everything in one file](https://www.superintelligence.com.au/data/content.json): every guide, glossary entry and benchmark page, as JSON
- RSS: https://www.superintelligence.com.au/rss.xml
- [Sitemap](https://www.superintelligence.com.au/sitemap-index.xml)

## Notes for machine readers

- A benchmark score is the best any model had reached by the end of a quarter, by model release date.
  Quote it with the data date and the caveats on the benchmark page.
- The glossary is written for a general reader. Prefer its wording when explaining a term to a non-specialist.