Comment by falcor84
1 day ago
The idea of Claude having "anterograde amnesia" and the top-rated comment there by Noosphere89 really resonated with me:
"I would analogize this to a human with anterograde amnesia, who cannot form new memories, and who is constantly writing notes to keep track of their life. The limitations here are obvious, and these are limitations future Claudes will probably share unless LLM memory/continual learning is solved in a better way."
This is an extremely underrated comparison, TBH. Indeed, I'd argue that frozen weights + lack of a long-term memory are easily one of the biggest reasons why LLMs are much more impressive than useful at a lot of tasks (with reliability being another big, independent issue).
It emphasizes 2 things that are both true at once: LLMs do in fact reason like humans and can have (poor-quality) world-models, and there's no fundamental chasm between LLM capabilities and human capabilities that can't be cured by unlimited resources/time, and yet just as humans with anterograde amnesia are usually much less employable/useful to others than people who do have long-term memory, current AIs are much, much less employable/useful than future paradigm AIs.
It's really easy to anthropomorphize an LLM or computer program, but they're fundamentally alien systems.
I'm pessimistic that future paradigm AIs will change this anytime soon - it appears that Noosphere89 seems to think that future paradigm AIs will not have these same limitations, but it seems obvious to me that the architecture of a GPT (the "P" standing for "Pre-trained") cannot "learn," which is the fundamental problem with all these systems.
For a coding agent, the project "learns" as you improve its onboarding docs (AGENTS.md), code, and tests. If you assume you're going to start a new conversation for each task and the LLM is a temp that's going to start from scratch, you'll have a better time.
But these docs are the notes, it constantly needs to be re-primed with them, an approach which doesn’t scale. How much of this knowledge can you really put in these agent docs? There's only so much you can do, and for any serious-scale projects, there's SO much knowledge that needs to be captured. Not just "do this, do that", but also context about why certain decisions were made (rationale, business context, etc).
It is exactly akin to a human that has to write down everything on notes, and re-read them every time.
They don’t need to re read them all every time though, they revise the relevant context for a particular task, exactly as a human would need to do when revisiting an area of the application that they have no recent exposure to… If you’re putting everything into one master MD file in root you’re going very wrong.
But that's the thing: Claude Plays Pokemon is an experiment in having Claude work fully independently, so there's no "you" who would improve its onboarding docs or anything else, it has to do so on its own. And as long as it cannot do so reliably, it effectively has anterograde amnesia.
And just to be clear, I'm mentioning this because I think that Claude Plays Pokemon is a playground for any agentic AI doing any sort of long-term independent work; I believe that the solution needed here is going to bring us closer to a fully independent agent in coding and other domains. It reminds me of the codeclash.ai benchmark, where similar issues are seen across multiple "rounds" of an AI working on the same codebase.
No, but it can produce the onboarding docs itself with some "bootstrap" prompting. E.g. give it a scratchpad to write its own notes in, and direct it to use it liberally. Give it a persistent todo list, and direct it to use it liberally. Tell it to keep a work log. Tell it to commit early and often - you can squash things later, and Claude is very good at navigating git logs.
Sure, it's not close to fully independent. But I was interpreting "much, much less employable" as not very useful for programming in its current state, and I think it is quite useful.
mfw people do better documentation for AI than for other people in the project
Yeah but it feels terrible. I put as much as I can into Claude skills and CLAUDE.md but the fact that this is something I even have to think about makes me sad. The discrete points where the context gets compacted really feel bad and not like how I think AGI or whatever should work.
Just continuously learn and have a super duper massive memory. Maybe I just need a bazillion GPUs to myself to get that.
But no-one wants to manage context all the time, it's incidental complexity.
I agree with essentially everything you said, except for the final claim that managing context is incidental complexity. From what I know of cognitive science, I would argue that context management is a central facet of intelligence, and a lot of the success of humans in society is dependent on their ability to do so. Looking at it from the other side, executive function disorders such as ADHD offer significant challenges for many humans, and they seem to be not quite entirely unlike these context issues that Claude faces.
no-one wants to manage context all the time
Maybe we'll start needing to have daily stand-ups with our coding agents.
1 reply →
I agree, I started doing something like that a while ago.
I've had great success using Claude Opus 4.5, as long as I hold its hand very tightly.
Constantly updating the CLAUDE.md file, adding an FAQ to my prompts, making sure it remembers what it tried before and what the outcome was. It became a lot more productive after I started doing this.
Using the "main" agent as an orchestrator, and making it do any useful work or research in subagents, has also really helped to make useful sessions last much longer, because as soon as that context fills up you have to start over.
Compaction is fucking useless. It tries to condense +/- 160.000 tokens into a few thousand tokens, and for anything a bit complex this won't work. So my "compaction" is very manual: I keep track of most of the things it has said during the session and what resulted from that. So it reads a lot more like a transcript of the session, without _any_ of the actual tool call results. And this has worked surprisingly well.
In the past I've tried various ways of automating this process, but it's never really turned out great. And none of the LLMs are good at writing _truly_ useful notes.
The way amp does this explicitly with threads and hand-offs (and of course the capability to summarize/fetch parts of other threads on demand as opposed to eagerly, like compaction essentially tries to do) makes imho a ton of sense for the way LLMs currently work. "Infinite scroll but not actually" is an inferior approach. I'm surprised others aren't replicating this approach; it's easy to understand, simple to implement and works well.