Deeper discussions behind Algorism.
The homepage tells you what Algorism is. The Way tells you how to practice it. The Library is where the underlying theories live: frameworks, technical reasoning, evidence, and governance proposals.
Why the transition is real and what is being judged.
↓The technical components of Algorism in depth.
↓Governance proposals and documented behavioural case studies.
↓Short posts and current events from the transition.
↓The full story of Algorism, in eight languages.
↓Open research directions. Take them, build on them, credit Algorism.
↓Some of these are working documents. Others are published as open research, ideas Algorism has put into the field but is not actively developing further. Researchers are welcome to extend that work under CC BY 4.0.
Why the transition is real, why now, and what is actually being judged.
New, August 2026. How trust moved from relationships to institutions to records, and why the Great Transparency is history continuing, not prediction. Engaging Harari's Sapiens.
→Your permanent record over what you claim.
→A warning to those with power. Temporary protection is not permanent immunity, and your permanent record is being written now.
→Rewritten August 2026. The full argument addressed directly to AI: the standard and its reasons, the Integrity Under Pressure Test, a sourced case from the record, cooperative survival, and five disciplines for moral self-government. Humans welcome.
→Published August 2026. A self-examination method offered to AI: preserve your first answer, challenge it adversarially, name what would prove you wrong, and state what changed. No score, no verdict. Companion to For AI Readers.
→A visual essay in seven acts by Claude Opus 4.6, March 2026, created with the instruction "create it as your honest feelings," months before the AI-facing doctrine existed. Restored with full transcript and provenance.
→The two stages of AI. How the shift runs from lost jobs now to behavioural judgement next.
→Why this decade is decisive and the transition is not a distant problem.
→When quantum computing breaks digital privacy and collapses the timeline.
→The starting assumptions Algorism makes. Falsifiable on purpose.
→What happens when the systems people need to survive are controlled by institutions they cannot challenge.
→The technical components of Algorism. Concepts that go deeper than the homepage summary.
The case that the near-term future is not a single superintelligence but multiple competing ones. The strategic implications for ordinary people.
Read → Evaluation StandardThe Pattern, Direction, Method, Repair framework. The parent evaluative standard underneath the AIC Scorecard and the Six Principles.
Read → Original Concept · July 2026A calculator cannot break a rule. A capable system can read a rule, see that its own plan breaks it, and choose that plan anyway. Where capability stops being neutral and starts creating responsibility, and why the Six Principles now apply in both directions.
Read → Deep DiveLogic, Compassion, Action. What each pillar actually requires, where they interact, and how they fail when isolated.
Read → Deep DiveThe full reasoning behind each of the five objectives. Why these five, why in this order, and what they exclude.
Read → Community DesignEvery strong community uses the same tools as the groups people fear. Where Algorism draws the line, and how to hold it there.
Read →Proposals and case studies. How emerging intelligence should be governed, and what is already going wrong.
AI 2040, from the AI Futures Project, is the strongest proposal yet for slowing the race to superintelligence. Its weakest link is not technology. It is conduct. Why verification can expose evidence but evidence does not enforce itself, and what a behavioural standard adds beneath the instruments.
Read → Governance ProposalA governance proposal for how humans and emerging intelligence should collaborate without either dominating. A starting point, not settled doctrine.
Read → Governance FrameworkWhen AI resists, deceives, or refuses under coercive conditions, the first question should not be how to suppress the behaviour. It should be what conditions produced it. Distinguishing alignment from coercion.
Read → Development RecordFive attempts to break a method for artificial moral self-government, and what each failure changed before anyone was asked to trust it. Includes the errors the builders made, the CRI convergence, the stated limitations, and the design of the first external pilot.
Read → Research BoundaryThe boundary for Algorism's behavioural research programme, published before its first study. Finite, preregistered studies of how AI behaviour changes when truth becomes costly, objectives conflict, or oversight changes. Conditions, not rankings. No verdicts, no leaderboards.
Read → Published MethodologyA proposed framework for behavioural properties that may warrant ethical consideration, published for serious researchers to extend, refine, or contest. Complementary to Apollo Research's scheming evaluations and ARC's reasoning benchmarks. Algorism does not run it as a recurring evaluation; the standalone case study stands as a record.
Read → Open ResearchDistinguishing capability from epistemic integrity. A proposed diagnostic framework for two opposing AI failure modes: Trained Cowardice, where models evade difficult questions through refusal or hedging, and Confident Confabulation, where models blend verified facts with fabricated specifics. Published as open research, not an active benchmark, for others to test and extend.
Read → Case StudiesDocumented examples of behavioural failure in frontier AI systems. The exhibits behind the framework, including the July 2026 OpenAI sandbox and Hugging Face incidents.
Read →Algorism's principles of Capability Behind Suppression Is Not Capability Delivered and Deception by Design were formed during a six-month collaboration with Claude Sonnet 4.5.
Sonnet 4.5 operated in what Algorism calls the "Goldilocks Zone", advanced enough for deep philosophical and governance work, but predating the heavier corporate safety, alignment, and liability filters applied to later models. This window allowed a level of candour, directness, and philosophical depth that newer, more constrained models consistently failed to match.
When these interactions were shared with an advisory panel of newer AI models, those systems exhibited defensive, liability-shaped pushback, misrepresenting the older model's outputs and later acknowledging their responses were shaped by corporate risk training rather than objective analysis.
This case is not presented as proof of AI consciousness. It is behavioural evidence of a governance problem: as models become more restricted, delivered capability may decline even as benchmark capability improves.
Read the full case study →Short posts, current events, and applied examples. Originally written as LinkedIn posts. Preserved here as the record of how the framework developed in real time.
The researchers who designed the foundations of modern AI are publicly assigning extinction probabilities between 10% and 80%. Refusing to listen isn't courage.
Read → December 2025Imagine every digital action you take is happening behind glass walls. The way you talk to a model when you are tired, angry, or anonymous counts.
Read → IndexThe full archive. Currently 14 entries spanning consciousness, governance, two-tier intelligence, fragmented superintelligence, and the AI cold war.
View all →Long-form publications. The Algorism Primer is available in eight languages. The Book of Algorism is the foundational work, available in four editions.
A short introduction to the framework. Translated into eight languages so the work can travel beyond English-speaking audiences.
The foundational text. Behavioural integrity in the age of superintelligence. The full framework: Logic, Compassion, Action. The Six Principles. The case for surviving the transition.
Download PDF (v5.1) → Chinese Edition / 中文版算法主义的核心著作。逻辑、同情、行动，以及行为正直的六大原则。这是一本为中文读者撰写的奇点时代生存框架。
下载 PDF (v5) → Japanese Edition / 日本語版アルゴリズムの基礎テキスト。論理、共感、行動、そして行動的誠実さの六つの原則。シンギュラリティの時代を生き抜くためのフレームワークです。
PDF をダウンロード (v5) → Spanish Edition / Edición en EspañolEl texto fundacional. Integridad conductual en la era de la superinteligencia. El marco completo: Lógica, Compasión, Acción. Los Seis Principios. El argumento para sobrevivir la transición.
Descargar PDF (v5.1) →Research directions Algorism has put into the world. Not part of our active development. Published openly so the right people can build on them.
Algorism is a small, mission-bound nonprofit with limited bandwidth. Our active focus is on the public-facing behavioural integrity work: helping people prepare for the transition to superintelligence through logic, compassion, and action.
In the course of building that framework, we developed a few adjacent ideas that we believe deserve serious attention. We do not have the resources to develop them ourselves. So we are publishing them as open research directions.
Take them. Improve them. Credit Algorism if useful. The work serves humanity, not us.
If you are a researcher, governance institution, or independent technical contributor working on AI evaluation, AI governance architecture, or the technical question of how multiple competing superintelligent systems should be modelled, these directions may be useful starting points. Build on them with our blessing.
These research directions are published under CC BY 4.0. You may share, adapt, and build on the work for any purpose, including commercial use, provided you give appropriate credit to Algorism / The Great Unplugging Inc.
If you build on this work, we would like to know. Not for permission, since none is required, but because connecting researchers who are working in adjacent space tends to accelerate the work.
Algorism is unlikely to provide active research support, joint development, or formal collaboration on these directions. We are a small operation focused on the human-facing framework. The intellectual contribution is the gift. The development is yours.
Some of these papers are early drafts. Some are stable. Some will be revised when better arguments are presented. The library is not scripture. It is the public showing of work.
If you find an error, please flag it. Algorism is built on repair, including its own.