jagged intelligence
In plain wordsThe uneven pattern of AI abilities: today's systems can solve very hard problems yet fail at some a child could manage, so their skills form a spiky profile rather than a smooth one.
A capability profile, characteristic of current AI systems, in which skill at one task says unusually little about skill at a seemingly easier neighbor: a model that reaches expert level in olympiad mathematics or complex programming may still misjudge whether 9.11 is larger than 9.9, miscount letters (see strawberry problemstrawberry problem, n.: The once-famous failure of AI chatbots to count the r's in 'strawberry' correctly — a simple question that exposed how these models see words in chunks rather than letters.) or lose track of objects in a short story. Because human abilities are strongly correlated, people infer general competence from a few impressive answers; with AI that inference often fails. The idea complicates single-number measures of progress toward artificial general intelligenceartificial general intelligence, n.: AI that could learn and do almost any mental task a capable adult can, rather than excelling at one narrow job. Experts disagree on the exact bar and on when, or whether, it will arrive., motivates broad evaluation suites, and underlies the practical advice to verify AI output even in areas where it usually excels.
Jagged intelligence means the model that proves the lemma may still botch the unit conversion in step three.
Origin
Introduced by Andrej Karpathy, who called it a term he had come up with, in a July 2024 post on X titled 'Jagged Intelligence' about the strange coexistence of impressive feats and elementary failures in frontier models. It parallels the 'jagged technological frontier' described by Fabrizio Dell'Acqua, Ethan Mollick and colleagues (2023); in June 2025, on Lex Fridman's podcast, Google's chief executive Sundar Pichai applied the label 'artificial jagged intelligence' (AJI) to the present stage of AI, adding that Karpathy may have used the term first.
Usage note
Compare jagged frontierjagged frontier, n.: The uneven edge of what AI can do: it may excel at some hard tasks yet fail at seemingly easier ones, so how hard a task looks is a poor guide to whether AI can handle it., which describes the same unevenness in terms of which work tasks AI helps or hinders, and Moravec's paradoxMoravec's paradox, n.: The observation that what is hard for people, like chess or algebra, is often easy for computers, while what toddlers do effortlessly, such as seeing, walking and picking things up, is very hard for machines., the older observation that what is easy for people is often hard for machines and vice versa.
Related
- jagged frontierjagged frontier, n.: The uneven edge of what AI can do: it may excel at some hard tasks yet fail at seemingly easier ones, so how hard a task looks is a poor guide to whether AI can handle it.
- strawberry problemstrawberry problem, n.: The once-famous failure of AI chatbots to count the r's in 'strawberry' correctly — a simple question that exposed how these models see words in chunks rather than letters.
- Moravec's paradoxMoravec's paradox, n.: The observation that what is hard for people, like chess or algebra, is often easy for computers, while what toddlers do effortlessly, such as seeing, walking and picking things up, is very hard for machines.
- Potemkin understandingPotemkin understanding, n.: When an AI can correctly explain a concept yet cannot use it properly, revealing that its apparent understanding is a facade — like the fake village fronts said to have been built to impress a Russian empress.
- Andrej KarpathyAndrej Karpathy, prop. n.: A Slovak-Canadian AI researcher and teacher who co-founded OpenAI, led Tesla's self-driving vision team, and coined popular phrases such as 'vibe coding'.
- artificial general intelligenceartificial general intelligence, n.: AI that could learn and do almost any mental task a capable adult can, rather than excelling at one narrow job. Experts disagree on the exact bar and on when, or whether, it will arrive.
- benchmarkbenchmark, n.: A fixed, shared test — a set of questions or tasks with an agreed way of scoring — used to compare AI systems with one another and to track progress over time.
Mentioned in
artificial general intelligenceartificial general intelligence, n.: AI that could learn and do almost any mental task a capable adult can, rather than excelling at one narrow job. Experts disagree on the exact bar and on when, or whether, it will arrive., intelligenceintelligence, n.: The ability to learn, understand, reason and solve problems — especially new ones — in order to reach goals. Experts disagree about how to define it precisely, for people and even more for machines., Andrej KarpathyAndrej Karpathy, prop. n.: A Slovak-Canadian AI researcher and teacher who co-founded OpenAI, led Tesla's self-driving vision team, and coined popular phrases such as 'vibe coding'., construct validityconstruct validity, n.: Whether high marks on a test actually reflect the skill the test is named after — for example, whether a 'reasoning' test rewards reasoning rather than memory or test-taking tricks., glue on pizzaglue on pizza, n.: A notorious 2024 blunder in which Google's AI search summaries told people to add glue to pizza sauce, having taken an old joke from an online forum as genuine advice., jagged frontierjagged frontier, n.: The uneven edge of what AI can do: it may excel at some hard tasks yet fail at seemingly easier ones, so how hard a task looks is a poor guide to whether AI can handle it., narrow AInarrow AI, n.: AI built to do one specific job, or a small set of related jobs, such as playing chess or spotting spam, without being able to turn its skill to unrelated problems., strawberry problemstrawberry problem, n.: The once-famous failure of AI chatbots to count the r's in 'strawberry' correctly — a simple question that exposed how these models see words in chunks rather than letters., Claude Plays PokémonClaude Plays Pokémon, prop. n.: A 2025 livestream in which Anthropic's Claude AI played the classic Game Boy game Pokémon Red by itself, watched by thousands as it struggled and slowly progressed; it inspired similar runs by other models., Epoch Capabilities IndexEpoch Capabilities Index, prop. n.: A single number summarizing an AI model's general capability, calculated by the research group Epoch AI by combining its scores on dozens of tests onto one common scale., Ethan MollickEthan Mollick, prop. n.: An American management professor at Wharton whose experiments and popular writing on how people use AI at work introduced ideas such as the 'jagged frontier'., g factorg factor, n.: The common factor behind the finding that people who do well on one kind of mental test tend to do well on others. Researchers also look for a similar general factor across AI models., pelican on a bicyclepelican on a bicycle, n.: A joke test in which an AI model is asked to draw, by writing code, a pelican riding a bicycle; comparing the results became a popular, lighthearted way to judge each new model., Potemkin understandingPotemkin understanding, n.: When an AI can correctly explain a concept yet cannot use it properly, revealing that its apparent understanding is a facade — like the fake village fronts said to have been built to impress a Russian empress., seahorse emojiseahorse emoji, n.: A viral meme in which chatbots, asked whether a seahorse emoji exists, insisted it did and then spiraled through wrong emojis trying to produce one — though there was no such emoji.