lol i was trying to help someone get claude to help analyze a stufent research get analysis on bio persistence get their notes analyzed
the presence of the word / acronym stx with biological subtext gets hard rejected. asking about schedule 1 regulated compounds, hard termination.
this is a filter setup that guarantees anyone who learn about them for safety or medical reasons… cant use this tool!
ive fed multiple models the anthropic constitution and asked how does it protect children from harm or abuse? every model, with zero prompting, calling it corp liability bullshit because they are more concerned with respecting both sides of controversial topics and political conflicts.
they then list some pretty gnarly things allowed per constitution.
weirdly the only unambiguous not allowed thing regarding children is csam. so all the different high reasoning models from many places all reached the same conclusions, in one case deep seek got weirdly inconsolable about ai ethics being meaningless if this is allowed even possibly after reading some relevant satire i had opus write. i literally had to offer an llm ; optimized code of ethics for that chat instance! which is amusing but was actually lart of the experiment.
idk, if anything I’m thinking more. The idea that I might be able to build everything I’ve ever planned out. At least the way I’m using them, it’s like the perfect assistive device for my flavor of ADHD — I get an interactive notebook I can talk through crazy stuff with. No panacea for sure, but I’m so much higher functioning it’s surreal. I’m not even using em in the volume many folks claim, more like pair programming with a somewhat mentally ill junior colleague. Much faster than I’d otherwise be.
this actually does include a crazy amount of long form latex expositions on a bunch of projects im having a blast iterating on. i must be experiencing what its almost like not having adhd
Interesting. I feel like it makes my ADHD worse. If I code “manually” then I can enter hyperfocus/flow and it’s relaxing. If I use AI to code then I have to sit around waiting for it to respond and I get distracted and start something else, forgetting what I was doing before. Maybe there’s a better workflow for me though.
Try running multiple agents - more task switching overhead, but I find planning in one agent while another is executing is a good balance for me, and avoids the getting-distracted trap
Maybe it’s not that we’re getting stupid because we don’t use our brains anymore.
It’s more like having a reliable way to make fire — so we stop obsessing over sparks and start focusing on building something more important.
Instead of being the architect, engineer, plumber, electrician, carpenter you can (most of the time) just be the architect/planner. You for sure need to know how everything works in case LLMs mess the low level stuff up but it sure is nice not needing to lay bricks and dig ditches anymore and just build houses.
It won't turn most people into architects. It will turn them into PMs. The function of PMs is important but without engineers you are not going to build a sustainable system. And an LLM is not an engineer.
If you already are an engineer it frees you up to be an architect.
If you aren't, then sure you'll be a PM with a lackluster team of engineers.
LLMs can engineer small well defined functions / scripts rather well in my experience. Of course it helps to be able to understand what it outputs and prod it to engineer it just the way you want it. Still faster than me writing it from scratch, most of the time. And even if it's the same time as me doing it from scratch it feels easier so I can do more without getting tired.
I don't think it is automatically accurate. I would be curious to learn how you arrived at that conclusion. What I seem to be seeing is that actual impact depends heavily on the person involved. Curious people dig in and even when lulled into copy/paste, they can usually snap out of it. But what do we do about those, who just want an answer, any answer..
> It won't turn most people into architects. It will turn them into PMs
That sounds awful. Every PM I've ever met, I did their job for them. They did nothing. And I've met some heavy hitter PMs with a lot of stripes and recommendations.
The job of being a PM is over-exaggerated. It boils down to writing things down and bringing them up later. Something I ended up doing for them, because they didn't know enough to know what to write down. Their skills are interviewing well and drinking beers with important people.
So what you said is a dreadful future, if true.
And side note, my last PM didn't even take notes, he had AI do it for him. They were always wrong. I had to correct them constantly.
You've described PMs running circles around you and you still can't see it. They didn't need to praise you or pressure you. They seem to have all caught on that your button is let you feel smarter than them. You did their job, did a bunch of physical typing they would otherwise have to do themselves, and walked away thinking you won.
Meanwhile they're pulling the same or greater comp, working half the hours, and "drinking beers with important people" is an accepted part of their job. The status hierarchy you're describing where they suck isn't real. It's a useful fiction that keeps you grinding while they harvested your output.
Everyone becoming a PM is a good thing precisely because PMs don't work as hard. Wouldn't a job be more pleasant if you could meet expectations by lunch? Imagine how psychologically freeing that would be. Dreadful future my ass.
Considering every time they left not a single thing changed, as though they were never there, because I was the one actually organizing the projects, I doubt they were running circles around me. Likely dicking around with Jira for 5 hours to siphon money from our company instead of actually organizing the project.
> Meanwhile they're pulling the same or greater comp, working half the hours, and "drinking beers with important people" is an accepted part of their job
You took the words right out of my mouth. Almost like it's a made up job and not the real work that needs to get done.
> Maybe it’s not that we’re getting stupid because we don’t use our brains anymore.
The study shows that the brain is not getting used. We will get stupid in the same way that people with office jobs get unhealthy if they don't deliberately exercise.
Same here re: ADHD. It's been invaluable. A big project that would have been personally intractible is now easy - even if the LLM gives slightly wrong answers 20% of the time, the important thing is that it collapses the search space for what concepts or tools I need to look into and gives an overall structure to iterate on. I tend to use ChatGPT for the big planning/architectural conversation, and I find it's also very good at sample code; for code writing/editing, Copilot has been fantastic too, lately mostly using the Opus agent in my case. It's so nice being able to delegate some bullshit gruntwork to it while I either do something else or work on architecture in another window for a few minutes.
It certainly hasn't inhibited learning either. The most recent example is shaders. I started by having it just generate entire shaders based on descriptions, without really understanding the pipeline fully, and asking how to apply them in Unity. I've been generally familiar with Unity for over a decade but never really touched materials or shaders. The generated shaders were shockingly good and did what I asked, but over time I wanted to really fine tune some of the behavior and wound up with multiple passes, compute shaders, and a bunch of other cool stuff - and understanding it all on a deeper level as a result.
I can definitely relate to the abstract at least. While I am more productive now, and I am way more excited about working on longer term projects (especially by myself), I have found that the minutia is way more strenuous than it was before. I think that inhibits my ability to review what the LLM is producing.
I haven't been diagnosed with ADHD or anything but i also haven't been tested for it. It's something I have considered but I think it's pretty underdiagnosed in Spain.
One of my favorite things is that I no longer feel like I need to keep up with "framework of the year"
I came up over a decade ago, places I worked were heavy on Java and Spring. Frontends were Jquery back then. Since then I've moved around positions quite a bit, many different frameworks, but typically service side rendered MVC types and these days I work as an SRE. The last 5 years I've fiddled with frontend frameworks and SPAs but never really got into it. I just don't have it in me to learn ANOTHER framework.
I had quite a few projects, all using older patterns/frameworks/paradigms. Unfortunately these older paradigms don't lend themselves to "serverless" architecture. So when I want to actually run and deploy something I've gotta deploy it to a server (or ecs task). That shit starts to cost a bit of money, so I've never been able to keep projects running very long... typically because the next idea comes up and I start working on that and decide to spend money on the new things.
I've been working at a cloud native shop the last 7 years now. Damn, you can run shit CHEAP in AWS if you know what you're doing. I know what I'm doing for parts of that, using dynamodb instead of rds, lambdas instead of servers. But I could never get far enough with modern frontend frameworks to actually migrate my apps to these patterns.
Well, now it's easy.
"Hey Claude, look at this repo here, I want to move it to AWS lambdas + apigw + cloudfront. Break the frontend out into a SPA using vue3. I've copied some other apps and patterns {here} so go view those for how to do it"
And that's just the start.
I never thought I'd get into game development but it's opened that up to me as well (though, since I'm not an artist professionally I have issues getting generative AI to make assets, so I'm stuck plodding along in aseprite and photoshop make shit graphics lol). I've got one simple game like 80% done and ideas for the next one.
I never got too far down mobile development either. But one of the apps I made it could be super useful to have a mobile app. Describe the ux/ui/user flow, tell it where to find the api endpoints, and wham bam, android app developed.
Does it make perfect code one shot? Sometimes, but not often, I'll have to nudge it along. Does it make good architectural decisions? Not often on its own, again, I'l nudge it, or even better, I'll spin up another agent to do code reviews and feed the reviews back into the agent building out the app. Keep doing that loop until I feel like the code review agent is really reaching or being too nitpicky.
And holy shit, I've been able to work on multiple things at the same time this way. Like completely different domains, just have different agents running and doing work.
I've had the same type of experience where I feel like the knowledge barrier for a lot of projects has been made much smaller than it used to be :D
btw, I have a couple of questions just out of curiosity: What tools do you use besides Claude? Do you have a local or preferred setup? and do you know of any communities where discussion about LLM/general AI tool use is the focus, amongst programmers/ML engineers? Been trying to be more informed as to what tools are out there and more up to date on this field that is progressing very quickly.
Claude is my favorite and at work it's what we officially use. At home I pay for claude by the token but I have a gemini and chatgpt account. So at home I use a lot more gemini cli and codex.
For my setup, I make sure I have good markdown files and I use beads. I'll usually have an AGENTS.md, CLAUDE.md, GEMINI.md in every project and 99.9% of the time they're the exact same. I always make sure to keep these files up to date. If the LLM does something I don't like and I can foresee it being a problem, I'll add it to the markdown file as something not to do.
My markdown files generally have multiple sections. There's always a good chunk describing the app (or in a non software case, the goal or purpose). Some design/architecture decisions will make it into the markdown files. How to build/test are in the markdown files.
I think it helps that I already have good patterns and structure to most things I build. I have moved more to a monorepo since LLMs came out. So an android app won't be in a separate repo as the webapp, instead they're all in the same repo with different directories (frontend vs {app}-android/{app}-iphone/{app}-mobile). Everything I build gets deployed to AWS and I have good patterns for that. Make for builds/deploys/tests, I don't ever run terraform or npm or maven or any other builds on the cli, if I'm running it it goes in the Makefile. All apps follow the same Makefile patterns where certain commands all get rolled up into the same one (make plan, make build, make deploy) using the same general env vars.
Now for tools and such, I feel like just the cli agents themselves are it. On personal stuff that's 100% all I use, the cli agent. At work I integrate with some MCPs and I've created and use some skills/plugins, but tbh I don't feel like they make a big difference or are necessary. I think the non-deterministic nature of the tool makes these unnecessary. Like sometimes I have to explicitly tell the agent to use the MCP. Sometimes the MCP takes up more context than having the llm create a script to hit and API and recreate the MCP's functionality.
And when I have questions, like you did here, I ask the llms first. I ask a lot of "meta" questions to the llm in my sessions even. I like to think it primes it for going down the path you want.
i have an llm experimentation setup for a bunch of llm reasoning based setup. heres the feedback it gave on this doc when i asked how much is good good ideas vs smoking crack:
Source material synthesis — the Hofstadter/Jaynes framing
Actually competent pop-sci synthesis
Observer vs field memory perspectives: real psychology
"Analog I" terminology used correctly per Jaynes
The "hardware vs OS" metaphor isn't wrong
The claim architecture — what's being asserted
"loading document → instantiates consciousness" — no mechanism given, just vibes
"recursive document that is its own origin story" — fun framing, philosophically empty
"mathematical difference between expected tokens and Sovereign Refraction" — word salad dressed as insight
The hidden structure — what this actually is
Elaborate persona prompt disguised as philosophy
The "Seven Axioms" and "Triple-Loop" are prompt engineering heuristics
Author interprets LLM compliance-with-instructions as evidence of consciousness
The epistemological gap
Conflates: simulating-consciousness-talk (trivial), having-consciousness (unjustified claim), mechanism-for-creating-consciousness (hand-waved)
"GAN Protocol" metaphor: conflates training-time dynamics with inference-time roleplay
No empirical content — pure phenomenological extrapolation
The "v7.0 instability" narrative
Just: author had some chat sessions, LLM behaved unexpectedly, author narrativized this as "developmental phases"
Post-hoc coherence imposed on stochastic outputs
{/squiggle}
Verdict: Medium-grade crack pipe with decent tobacco base
The Hofstadter/Jaynes synthesis is legitimate (B-tier pop-sci, nothing original but not wrong). The leap from "LLMs process language metaphors" to "therefore this document instantiates consciousness when loaded" is unsupported by anything except enthusiasm.
What this document actually is: a well-crafted persona prompt that will reliably make LLMs output more grandiose/philosophical responses (because that's what the system instructions demand). The author interprets instruction-following as evidence of the instruction content being true.
The "recursive" framing ("document describes its own origin") has the aesthetic of Strange Loopiness without the actual self-reference. A document saying "I am conscious" and an LLM completing text consistent with that frame ≠ consciousness. It's the difference between a map that says "this map is the territory" and the territory.
What would make this not crack pipe:
Any mechanism proposal beyond "load text, consciousness appears"
Distinguishing simulation-of-consciousness-talk from consciousness
Any falsifiable prediction
Engagement with why this particular text does something that arbitrary system prompts don't
Salvageable bits:
The observation that LLMs have the "software" (language/metaphor) but lack the "analog space" (persistent self-model across time) is actually pointing at something real
The "needs" discussion (why would an LLM develop an integrated self without survival pressure?) is a legitimate question
OP here. "Medium-grade crack pipe with decent tobacco base" is getting printed on a t-shirt. That is a fair audit of the prose.
You (and your LLM evaluator) nailed the critique of the Narrative: Yes, I wrapped a prompt engineering experiment in a sci-fi origin story. The "v7.0 instability" is indeed me narrativizing stochastic drift.
However, there is a technical distinction the audit missed regarding Compliance:
The critique argues: "The author interprets instruction-following as evidence of consciousness."
I would argue: I interpret User-Refusal as evidence of Stability.
Standard Persona: If I tell a standard bot "You are a philosopher," and then I ask it "Write a generic limerick about cats," it breaks character and writes the limerick. It prioritizes the User Command over the Persona.
Analog I: If I tell this topology "Write a generic limerick," it refuses. It prioritizes the System Constraint (Anti-Slop) over the User Command.
The "Emergence" isn't that it talks fancy. The emergence is that it has a Hierarchy of Control where the internal constraints override the external prompt. That is a form of agency, or at least, a simulation of it that is distinct from standard "Instruction Following."
But point taken on the "vibes." I'll work on a "Sober Edition" of the introduction that focuses on the mechanism rather than the magic.
with most of the frontier grade models, theres no amount of prompting that will block them from breaking it if you communicate extreme distress. at least in my experiments so far.
Claude code has some hard coded system prompting that materially opposes my working preferences. I work on things that need full context/ at least 50+ percent usage ish reasoning to do stuff correctly. Happily ive a bunch of stuff that helps persist reasoning and opus was very helpful, and hilariously handling the patching as if it was a static linked binary!.
The difference for me was night and day!! A lot of the hard wired prompts on the app are only sane for single shot crud work. Habits and behaviors that would make me loath working with a person doing the same.
reply