Yeah. I noticed Claud suffers when it reaches context overload - its too opinionated, so it shortens its own context with decisions I would not ever make, yet I see it telling itself that the shortcuts are a good idea because the project is complex...then it gets into a loop where it second guesses its own decisions and forgets the context and then continues to spiral uncontrollably into deeper and deeper failures - often missing the obvious glitch and instead looking into imaginary land for answers - constantly diverting the solution from patching to completely rewriting...
I think it suffers from performance anxiety...
----
The only solution I have found is to - rewrite the prompt from scratch, change the context myself, and then clear any "history or memories" and then try again.
I have even gone so far as to open nested folders in separate windows to "lock in" scope better.
As soon as I see the agent say "Wait, that doesnt make sense, let me review the code again" its cooked
> Yeah. I noticed Claud suffers when it reaches context overload
All LLMs degrade in quality as soon as you go beyond one user message and one assistant response. If you're looking for accuracy and highest possible quality, you need to constantly redo the conversations from scratch, never go beyond one user message.
If the LLM gets it wrong in their first response, instead of saying "No, what I meant was...", you need to edit your first response, and re-generate, otherwise the conversation becomes "poisoned" almost immediately, and every token generated after that will suffer.
Yeah, I used to write some fiction for myself with LLMs as a recreational pasttime, it's funny to see how as the story gets longer, LLMs progressively either get dumber, start repeating themselves, or become unhinged.
I’m keeping Claude’s tasks small and focused, then if I can I clear between.
It’s REAL FUCKING TEMPTING to say ”hey Claude, go do this thing that would take me hours and you seconds” because he will happily, and it’ll kinda work. But one way or another you are going to put those hours in.
I get that. But I think the AI-deriders are a bit nuts sometimes because while I’m not running around crying about AGI… it’s really damn nice to change the arguments of a function and have it just go everywhere and adjust every invocation of that function to work properly. Something that might take me 10-30 minutes is now seconds and it’s not outside of its reliability spectrum.
There’s definitely a certain point I reach when using Claude code where I have to make the specifications so specific that it becomes more work than just writing the code myself
For me, too many compactions throughout the day eventually lead to a decline in Claude's thinking ability. And, during that time, I have given it so much context to help drive the coding interaction. Thus, restarting Claude requires me to remember the small bits of "nuggets" we discovered during the last session so I find myself repeating the same things every day (my server IP is: xxx, my client IP is: yyy, the code should live in directory: a/b/c). Using the resume feature with Claude simply brings back the same decline in thinking that led me to stop it in the first place. I am sure there is a better way to remember these nuggets between sessions but I have not found it yet.
Perhaps, but I already have a CLAUDE.md file for the general coding session. Unique items I stumble upon each day probably should go into another file that can be dynamically updated. Maybe I should create a /slash command for this?
Edit: Shortly after posting this, I asked Claude the same type of question (namely how to persist pieces of data between each coaching session). I just learned about Claude's Memory System - the ability to store these pieces of data between coding sessions. I learn something new every day!
That has been my greatest stumbling block with these AI agents: context. I was trying to have one help vibe code a puzzle game and most of the time I added a new rule it broke 5 existing rules. It also never approached the rules engine with a context of building a reusable abstraction, just Hammer meet Nail.
HR people are usually the least humane people on the Org chart. They often will include the worst offenders of out-of-touch views and often privileged lives whilst championing their own self interests yet gate keeping the ability from anyone else. HR is a disease - the worst of the worst .
I have been doing something minor - not full language - but I am working on a couple DSLs for my game engine. The games themselves are interpreted by my engine console - then transpiled to webgl and JavaScript.
The DSL is for cutscenes and scripting dialogues and movements and transitions.
The idea was to have a screenplay like language that would run in my game engine so I can write the scenes like a script instead of like code.
—
Then I am using a LUA like subset for the core game logic.
I don’t think I would want to have the entire thing written in a custom language though, as that seems unnecessary
This is not surprising to me. It’s clearly directly wired into our emotions and nervous system. The nervous system is the key part which is being healed and so stimulation will likely increase the speed of reconnections
It would be nice if JSX was natively supported in the DOM as some kind of syntactic sugar on components by the browsers and this would fix the cycle of abuse and allow us to eventually get back to less compilation except for typescript
It’s fascinating how society refuses to acknowledge the systemic predators (corporations like MindGeek and Roblox and OnlyFans etc) - but will constantly try to rake the moral high ground on political fads and petty crime.
It seems like the more horrific the crime - the more profitable it is to someone….once a business mode supports the criminals…. Well….at what point is it collusion, willingly turning a blind eye, so the execs maximize that next quarterly bonus —-
How much of a bonus is worth the mutilation and destruction of kids…. ?
—-
Another question…. Why do the porn industry and the game industry both sequester themselves in Montreal primarily?
Is it the mafia in the city? Is it corrupted politicians? Is it taxes? What makes pedophiles and sexually exploitation and child grooming and Wokism…. Reside in Montreal in such abundance?
It’s because everyone else is dumber than them…. So they constantly see avoidable mistakes and misunderstandings that could have been avoided…. Yet they cannot make the other people understand….because they think differently about it, and the people who don’t have that intelligence will not necessarily even be able to reorient their brains for the new information to be absorbed correctly.
I constantly get demoralized by stupid people….. it’s truly horrific. It’s a disability as far as I can see…I am disabled by others stupidity….
You just sound like a misanthrope. If you’re so smart, why are you surrounding yourself with stupid people? Is it possible you’re not as smart as you think, and in fact, just as fallible as the rest of us?
How can I not surround myself with stupid people, though? There are so many. No matter where I go, there will be huge numbers of them. I'm surrounded by smart people at work. I can try to only socialize with smart people, but it's hard when a lot of socialization is determined by family. And there's absolutely nothing to be done about the general public on the street, working retail, etc. And worst of all: working in, running, or leading governments.
To be fair, if you are significantly smarter than the average, surrounding yourself with stupid(er) people is the default, as the majority will be less smart than you
or maybe they have spent time with some really brilliant people and they know they are not brilliant, but they also know they are not as dumb as the person they are talking to? they don't have to be mutually exclusive.
I recently encountered someone who spoke like this and I researched what might be the issue.
I came across narcissism. The idea that you’re smarter than everyone else. Comes from a grandiose sense of self importance. But the truth is most people are smarter than you in some ways and less smart in others, but you’re unable to see it because you’re in this black and white mode where preserving your ego relies on you being the smart guy amongst the idiots.
It’s very common in tech to see this. Maybe because we were all exceptional at maths when we were young and got the idea that meant we were super smart and this compensated for our nerdiness.
I worked with a bunch of physicists and every single one of them was smarter than me at maths and physics, I wasn’t even close. But they sometimes talked about politics and current affairs, which I’m very well read in. I didn’t say anything, but I was shocked at how little they knew and how overconfident they were.
None of those folks were narcissists, thankfully they were lovely people, but for sure it highlighted how poor people were at judging their own expertise in an area.
It’s so easy to dismiss people, criticising is easy, and so hard to see just how stupid you can be yourself.
I don't think a bunch of physicists is the best example if you're trying to make the point that most people are smart but just in different ways. Physics requires and attracts intelligence, of course a bunch of physicists were smarter than you in some ways.
Try it with, say, a parole officer's case list, a group of high school dropouts, or people who have not touched a book in the past ten years. You'll certainly find things they know that you don't, but that's mostly going to be a matter of experience, not them being smarter in that area. No doubt the average petty thief knows more about shoplifting than I do, but I'm pretty sure I could learn quickly and become a much better shoplifter than them if I put my mind to it. And those groups will certainly contain some really smart people who just happen to have ended up in those groups, but that's going to be a small number of them.
Yeah no I don’t think intelligence works like that.
There’s not some core reasoning engine in your brain that is independent of your knowledge. The two are intertwined.
Some people are better at reasoning about politics vs maths, for example, because they have both the knowledge, skills, and experience to understood how such systems work vs a mathematician who does not.
> I came across narcissism. The idea that you’re smarter than everyone else. Comes from a grandiose sense of self importance.
This may be the colloquial description of how narcissism manifests, but it isn't even close to (and possibly completely opposite) clinical narcissism. The grandiosity isn't so much a belief as it is a "false self" put on to garner caretaking from others. It's not "I got all the toys as a kid, so I deserve more stuff!" but a failure to individuate from caregivers. "Mom (as a tool, not a wholly independent person) came when I cried as a kid, so I need you to lavish attention on me and make me feel better now as an adult. I can't see myself without external input; I only see myself as a reflection through you."
Disclaimer: I'm just an...interested layman, but as far as I know we don't really know. It looks like there's both genetic and environmental components to it; your genes have a big impact on your "temperament" and more volatile temperaments correlate more strongly with cluster B disorders. But it seems like neglect, a failure to recognize a child as their own independent person (basically "my needs are your needs"), from parents/caregivers who are often themselves narcissists contributes pretty strongly, too.
The Little Shaman[1] has been one of the most comprehensive-yet-approachable resources I've found for understanding these kinds of high-conflict personalities. In particular [2-4] are pretty relevant to your question.
Perhaps there are some people like you say that are grandiose narcissists and ALSO some people that are genuinely smarter than most others in the room and can see the obvious through the fog.
I think the frustration they're experiencing is more likely to do with a lack of control over their environment (including the lack of ability to control others).
I do know for sure that I am able to share my intelligence with anyone.
Moreover I do enjoy it.
I cannot conceive a lot of things more pleasant than making people smart.
And to me that should be your responsibility to do so, as a gifted person.
You should cease to complain about other people being dumb, and just work on being understandable by anyone.
That's a very complex job, as it may lead to what I'd call "extravagant analysis" (i.e. unfolding abstractions to the point you reach atoms or "implementations", [note, I do remember of a (joke) book titled "How to ride a bike" where the author explained literally everything you needed to know to be able to ride a bike, to the point it became absurd]).
Anyway, you should at least try it. Smart people often are terrible at explaining stuff because they don't need to do the work of diving into the atoms of abstractions, and because sometimes also language is not their primary tool to think about things.
Location: Calgary
Remote: Yes
Willing to Relocate: Yes, possibly with assistance depending on the location.
Technologies: JS/TS, React, PHP8, C, C++, python, WebGL, OpenGL, Bash, SQL, NoSQL
Resume: https://linkedin.com/in/kderbyma (it’s posted for view)
Email: kderbyma at Gmail
I have 10+ years of experience. I have worked in startups from the ground floor up through scale ups with 150 employees. I cofounded my own SaaS out of university, spent years consulting businesses, did freelance work, and contributed to some open source projects with PRs. I am studying game engine development and game design, studying Ancient Greek, and focusing on areas of digital sovereignty and cultural preservation. I like nature, skateboard, hobbyist photographer, and enjoy philosophy and ancient history. Additionally, I have a Bsc in Electrical Engineering.
I would like to work in antiquities, humanities, game design, ETL analytics platforms, and custom business logic, and also electronics and IoT.
I think it suffers from performance anxiety...
----
The only solution I have found is to - rewrite the prompt from scratch, change the context myself, and then clear any "history or memories" and then try again.
I have even gone so far as to open nested folders in separate windows to "lock in" scope better.
As soon as I see the agent say "Wait, that doesnt make sense, let me review the code again" its cooked