Leaky abstractions

All abstractions are leaky, so code an AI writes for you will have mistakes.

1. Naval Ravikant 19 Feb 2026 Co-founder of AngelList And all abstractions are leaky. So when you have a computer programming for you—when you have Claude Code or equivalent programming for you—it’s going to make mistakes. ↗A Motorcycle for the Mindnav.al · 19 Feb 2026

   Their wordsAnd all abstractions are leaky. So when you have a computer programming for you—when you have Claude Code or equivalent programming for you—it’s going to make mistakes.

   ↗A Motorcycle for the Mindnav.al

   19 Feb 2026

It always helps to have knowledge one layer below your abstraction, because you get closer to reality.

Mental modelFirst principles

1. Naval Ravikant 19 Feb 2026 Co-founder of AngelList You can take this all the way down, but it always helps to have knowledge one layer below because you’re getting closer to reality. ↗A Motorcycle for the Mindnav.al · 19 Feb 2026

   Their wordsYou can take this all the way down, but it always helps to have knowledge one layer below because you’re getting closer to reality.

   ↗A Motorcycle for the Mindnav.al

   19 Feb 2026

Higher-level abstractions do not abolish low-level knowledge; they concentrate it in the smaller group of people who build the abstractions.

Mental modelConservation of complexity

1. Martin Kleppmann 22 Apr 2026 Computer science researcher at Cambridge But, somebody still has to build those lower-level abstractions and from lower-level components. Somebody's got to implement the cloud services. ↗Designing Data-intensive Applications with Martin Kleppmannyoutube.com · 22 Apr 2026 · 0:32:17 into the videoAll korrents from this video

   Their wordsBut, somebody still has to build those lower-level abstractions and from lower-level components. Somebody's got to implement the cloud services.

   Where this was said

   Watch from 0:32:17 plays here

   ↗Designing Data-intensive Applications with Martin Kleppmannyoutube.com

   22 Apr 2026 · video · 1h 25m · spoken · machine transcript

With AI, work is moving up the layers of abstraction, but the more context someone has about how things work under the hood, the better equipped they are to debug and build resilient things.

1. Flora Guo 25 Mar 2026 Founding design engineer at Paradigm as of March 2026 And I think with AI we're also moving up the layers of abstraction. So we're not writing every function by hand, like no one's doing that. But the more context that you do have about how it works under the hood, so how React works, you know, how our props being passed through, we're much better equipped to actually debug and build things that are resilient. ↗Flora Guo - How to accelerate your design career with AIyoutube.com · 25 Mar 2026 · 24:15 into the videoAll korrents from this video

   Their wordsAnd I think with AI we're also moving up the layers of abstraction. So we're not writing every function by hand, like no one's doing that. But the more context that you do have about how it works under the hood, so how React works, you know, how our props being passed through, we're much better equipped to actually debug and build things that are resilient.

   Where this was said

   Watch from 24:15 plays here

   ↗Flora Guo - How to accelerate your design career with AIyoutube.com

   25 Mar 2026 · video · 46m · spoken · machine transcript

Every abstraction in agentic programming — RAG, memory, agentic history, structured output — is just a different way of passing tokens into a model, and understanding that beats learning any of them.

1. Dex Horthy 15 Jul 2026 Founder of the agent company HumanLayer And at the end of the day, they're all like different ways to pass tokens into a model and ask it to produce usually some structured output. And understanding that is a lot more powerful than trying to learn memory and trying to pick some agent framework off the shelf and some memory framework off the shelf. ↗Context engineering with Dex Horthyyoutube.com · 15 Jul 2026 · 0:21:28 into the videoAll korrents from this video

   Their wordsAnd at the end of the day, they're all like different ways to pass tokens into a model and ask it to produce usually some structured output. And understanding that is a lot more powerful than trying to learn memory and trying to pick some agent framework off the shelf and some memory framework off the shelf.

   Where this was said

   Watch from 0:21:28 plays here

   ↗Context engineering with Dex Horthyyoutube.com

   15 Jul 2026 · video · 1h 33m · spoken · machine transcript

Because programming requires forgetting, developers should weigh the costs of their abstractions and who those abstractions leave out.

Mental modelTrade-offs

1. Phil Hagelberg 25 Jul 2022 Programmer known online as technomancy Programming, of course, is forgetting, but we need to at least try to be aware of the costs of the abstractions we choose and consider who it is that ends up being forgotten. ↗in which the word column has several meanings, none of them architecturaltechnomancy.us · 25 Jul 2022All korrents from this piece

   Their wordsProgramming, of course, is forgetting, but we need to at least try to be aware of the costs of the abstractions we choose and consider who it is that ends up being forgotten.

   ↗in which the word column has several meanings, none of them architecturaltechnomancy.us

   25 Jul 2022

Cloud abstraction problems cannot be fixed by adding more abstraction layers on top.

1. David Crawshaw 22 Apr 2026 Software engineer; co-founder and former CTO of Tailscale You cannot solve the fundamental problems with cloud abstractions by building new abstractions on top. ↗I am building a cloudcrawshaw.io · 22 Apr 2026All korrents from this piece

   Their wordsYou cannot solve the fundamental problems with cloud abstractions by building new abstractions on top.

   ↗I am building a cloudcrawshaw.io

   22 Apr 2026

Using an AI system to write code is delegating a task to a non-deterministic agent, not using a new layer of abstraction.

1. Rakhim Davletkaliyev 8 Sept 2025 Programmer and teacher; writes on software But we should not lie to ourselves and call it an abstraction in the same way that a programming language is an abstraction. It is a delegation of a task to a non-deterministic agent. ↗AI is not another abstraction because god plays dicerakhim.exotext.com · 8 Sept 2025All korrents from this piece

   Their wordsBut we should not lie to ourselves and call it an abstraction in the same way that a programming language is an abstraction. It is a delegation of a task to a non-deterministic agent.

   ↗AI is not another abstraction because god plays dicerakhim.exotext.com

   8 Sept 2025

The difficulty of software engineering is probably essentially constant across abstraction levels, because human cognition adapts to new tools.

1. François Chollet 25 Sept 2026, 04:19 UTC Creator of the Keras deep-learning library and the ARC-AGI benchmark I think the "difficulty" of software engineering is essentially constant no matter what abstraction level you move to, because human cognition adapts to new tools until it can fully utilize itself. ↗@fchollet on Xx.com · 25 Sept 2026

   Their wordsI think the "difficulty" of software engineering is essentially constant no matter what abstraction level you move to, because human cognition adapts to new tools until it can fully utilize itself.

   ↗@fchollet on Xx.com

   25 Sept 2026