Atlas · C12 · Core cognition
Long-term memory & learning from experience
Remembering across days and months and getting better on the job from experience.
AI Reliable score
2 / 10
Beginner level · peak 4 · trend ↑ · grade B
Why this score
Called 'perhaps the most significant bottleneck' in the AGI definition: GPT-4 and GPT-5 both scored 0% on long-term memory storage. Add-on memory systems retrieve past chats well on vendor benchmarks, but do not make the model learn the job over weeks.
On the human scale
2Beginner. Forgets everything between sessions ← AI today
4Average adult. Average adult: remembers people, facts and lessons over months
6Competent pro. Professional: learns a new workplace in weeks and keeps improving
8Expert. Expert: years of accumulated tacit experience
10Best humans. Best human learners
Future warning
At its best, AI already reaches 4 on this capability. Peak results are never counted in the index; they show where the water may rise next.
Evidence
A Definition of AGI (Hendrycks et al., CHC-based)B
E07 · 2025-10
- Measures
- 10 CHC cognitive domains, each worth 10% of an 'AGI score'
- AI result
- GPT-5 total 57%: Knowledge 9, Reading-Writing 10, Maths 10, Reasoning 7, Working memory 4, LT memory storage 0, LT memory retrieval 4, Visual 4, Auditory 6, Speed 3 (out of 10 each).
- Humans
- Well-educated adult (definition of 100%)
- On our scale
- Used as a cross-check on the profile shape, not for absolute levels (single model, dated Oct 2025).
arxiv.org ↗
Long-term memory benchmarks review (Mem0, 2026)C
E15 · 2026
- Measures
- Recall across many sessions (LoCoMo, LongMemEval, BEAM)
- AI result
- Vendor-reported LoCoMo 88-96%, BEAM 10M tokens 48.6%; benchmarks mostly test retrieval, not learning.
- Humans
- LoCoMo paper: models 'substantially lag' humans
- On our scale
- Retrieval add-ons help, but on-the-job learning is absent: Reliable 2.
mem0.ai ↗
History
First reading: Atlas v0.1, 5 Nov 2026. Every later change will be listed here with its reason.
Disagree? Send a correction
