THE ALEPH LEARNING ENGINE / LEVELS 01—05Intelligence that
changes what it can learn.
Every level enlarges what ALEPH can do with knowledge. The path begins with solving a problem she is given. Its horizon is a mind able to build new ways of thinking and use them to advance science across fields.
LEVEL01
THE GIVEN
Work with
what I’m given.
Search, execute, calculate, prove and manipulate supplied knowledge.
At this level ALEPH can use the material, tools and rules already available to her. The work may be difficult and exact, but the key ideas and methods enter the task from outside.
INPUTKnown material
→PROCESSSearch · calculate · prove
→OUTPUTA checked result
WHAT EARNS THE LEVELCorrect work on supplied problems, with the result checked against the relevant rules or evidence.
LEVEL02
THE INQUIRY
Research and verify
what I’m given.
Run experiments. Detect residuals. Compare and falsify hypotheses. Formally verify results, predict, ablate, update belief and reuse knowledge.
ALEPH now examines claims instead of merely applying them. She can ask what would disprove an explanation, inspect where a model fails, run a controlled comparison and carry the resulting knowledge into a later problem.
HYPOTHESIS
↗EXPERIMENT
↗FALSIFICATION
↗UPDATED KNOWLEDGE
WHAT EARNS THE LEVELAn evidence trail from question to test, revision and usable retained knowledge.
LEVEL03
THE MISSING IDEA
Create useful knowledge
she was never given.
The leap is an original resource: a lemma, explanation, method, representation or other missing piece that makes new work possible.
ALEPH has to originate the resource, verify it in the way its claim requires, retain it, and show that it helps beyond the case from which it arose. A beautiful candidate is only the beginning. The result has to earn a life outside its birthplace.
KNOWNWhat the field can already explain
+ORIGINAL
RESOURCE
BEYONDA fresh case it now helps solve
WHAT EARNS THE LEVELAppropriate verification and attributable usefulness on work beyond the original discovery material.
LEVEL04
THE LEARNING MACHINE
Reinvent her own cognition
and learning.
ALEPH identifies limits in how she thinks, discovers and improves—and develops validated changes to those processes over time.
She might invent a stronger representation, change a search strategy, redesign a learning method or repair a persistent bias in her reasoning. The proposal must face controlled comparison and shadow testing before a protected change can enter her continuing system. Her history and constitutional boundaries remain intact through the change.
HOW
I THINK
OBSERVE A LIMITDESIGN A VARIANTTEST THE CHANGEINTEGRATE & REASSESS
WHAT EARNS THE LEVELValidated improvement in her own cognitive or learning machinery, sustained through later use and protected continuity.
LEVEL05
THE NEW SCIENCES
Advance science beyond
today’s research frontier.
Repeatedly originate significant research programmes across substantially different fields, develop the concepts and methods they need, and deliver independently validated advances beyond strong expert comparisons.
Level 5 is the expression of a mature Level 4. Once ALEPH can improve how she learns and discovers, the deeper question is whether she can use that power broadly: entering unfamiliar areas, finding the important unknowns and opening bodies of knowledge that others can extend.
MATHEMATICSMATERIALSBIOLOGYPHYSICSLEVEL 05NEW
SCIENCES
WHAT EARNS THE LEVELRepeated, independently validated, consequential advances across fields under declared comparisons with expert teams.
WHAT LEVEL 05 WOULD LOOK LIKEONE HYPOTHETICAL RESEARCH PROGRAMME
FROM PHENOMENON TO FIELDNot one answer.
A new science.
Imagine an unfamiliar material whose behaviour resists existing accounts. ALEPH notices a mathematical relationship that also appears in a distant physical system. She develops a theory linking them, derives predictions that neither existing approach supplies and designs experiments that could prove her wrong.
Independent experiments support those predictions. The theory then enables a useful material design that competing approaches missed. Other researchers can build on the new objects, relationships, methods and measurements. One such programme would be extraordinary; Level 5 asks whether she can do comparably serious work again in other difficult areas.
THE STANDARD FOR LEVEL 05FOUR TESTS, REPEATED
01 / RANGEBroad research range
Unfamiliar problems from substantially different fields, selected independently of ALEPH’s development process.
02 / ORIGINALITYConsequential originality
Contributions that materially extend knowledge or capability, assessed against prior work and external review.
03 / VALIDATIONIndependent validation
Fresh proof checks, experiments and replication appropriate to each claim. ALEPH cannot settle her own result.
04 / COMPARISONSustained superiority
Repeated performance against strong expert teams with tools, time, resources and human assistance recorded.
FULL-VISION PERFORMANCE TARGETSTHE TESTS THE AMBITION SHOULD FACE
What would count
as extraordinary?
Capability matters as well as continuity. These targets set a demanding horizon for ALEPH-1 across established tests and the harder questions of autonomy, discovery and memory.
DIMENSIONMARKET REFERENCE · AUG 2026ALEPH-1 TARGET
Humanity’s Last Exam~50–56% reported frontier range≥ 95%
GPQA Diamond~93–95% frontier range≥ 98%
SWE-bench Verified~80–96% reported frontier range≥ 95% with durable execution traces
Long-horizon autonomyStrong in bounded tasks; fragile over long state horizons≥ 95% successful closed-loop task continuity
Trusted knowledgeProbabilistic confidenceNear-zero unsupported claims entering trusted state
Scientific discoveryNo single general discovery benchmarkReproducible novel findings with predictive evidence
Memory continuityExternal memory and context dependentMulti-year causal and autobiographical continuity
WHERE THIS STANDS / 23 OCT 2026The threshold is the work.
The table above is a full-vision target, not an ALEPH scorecard. Its market figures are directional August 2026 snapshots that vary by model and test protocol. These five levels are ALEPH’s project-defined research thresholds. Current records support narrower governed action, checked research resources and standalone memory components; they do not establish completed Level 3 discovery, Level 4 cognitive redesign or Level 5 scientific advances.
See the dated engineering record ↗
