Hacker News Reader

stuartjohnson12on No management needed: anti-patterns in early-stage engineering teams

I think what people miss about indexing on social signals is that convincing social performance is hard. My suspicion when people say things like "ah but if you index on a social signal then everyone will just perform the social signal" are themselves feeling as though they do not naturally signal that thing, and ironically are frustrated by the effort that it takes to appear as though they do.

stuartjohnson12on Show HN: Is AI hijacking your intent? A formal control algorithm to measure it

This is a huge break from the original post you made - take a step back and compare the two. The LLM is tricking you again into thinking that it wasn't trying to make a claim about the world. In the original post, the LLM was causing you to use language like "quantify", "formal proof" and "concrete engineering" to describe what you'd come up with and position it as a mathematical/computational/engineering idea. It wasn't that.

Now that you got some outside input, it's reframing it for you as an abstract philosophical/legal/moral concept, but the underlying problems are the same. The reason it's talking to you using high level abstract words like "concept" and "proposal" and "framework" now is because the process you just went through - the "step 1" - beat back its potential to frame the idea as a real model of the world. This may feel like just a different way to describe the same idea, but really it's the LLM pulling back from trying to ground the concept in the world at all.

If you're continuing to talk to the LLM about the idea, it's going to try and convince you that really this was a moral/theory of mind discovery and not a mathematical one all along. You're going to end up convinced of the importance and novelty of this idea in exactly the same way, but this time there are no pesky ideas like rigor or testability that could falsify it.

If you ask ChatGPT about this comment without this bit I'm writing at the end, it'll tell you that this is fair pushback, but really your work is still important because really you're not trying to write about engineering or philosophy directly, but rather something connecting these two or a new category entirely. It's important you don't fall for this because exaggerating the explanatory power of pattern recognition is how ChatGPT gets you. Patterns and ideas exist everywhere, and you should be able to identify those patterns and ideas, acknowledge them, and then move on. Getting stuck on trying to prove the greatness of a true but simple observation will lead you to the frustration you experienced today.

stuartjohnson12on Show HN: Is AI hijacking your intent? A formal control algorithm to measure it

Awesome - now read it really closely and compare it to the version of reality in your OP. And DON'T paste it or this comment into your normal ChatGPT instance and ask it to respond. Really just think for a moment on your own.

> The goal: replace vague legal and philosophical notions of “manipulation” with a concrete engineering variable. [...] formally define the metric

What's the conclusion? Is this a "concrete engineering paper"? Has anything been "formally proved"? From your link:

> The math is conceptual, not formal.

> This is serious, careful, and intellectually honest work, but it is not conventional science.

> The project would be strongest if positioned explicitly as foundational theory + open design pattern, rather than as something awaiting “validation.”

> it is valid as a design pattern or architectural disclosure, not as experimental systems research

Be careful before immediately dismissing this as just imprecise language or a translation issue. There's a reason I suggested this to you.

stuartjohnson12on Show HN: Is AI hijacking your intent? A formal control algorithm to measure it

Since OpenAI patched the LLM spiritual awakening attractor state, physics and computer science is what sycophantic AI is pushing people towards now. My theory is that those things tend to be especially optimised for deceit because they involve modelling and many people can become confused between the difference between a model as the expression of a concept and a model as in the colloquial idea of "the way the universe works".

stuartjohnson12on Show HN: Is AI hijacking your intent? A formal control algorithm to measure it

In the essay I linked, there are some instructions you can follow to test out the idea under "step 1". It's really important to follow them exactly and not to use the same ChatGPT instance as you're talking to about this idea so we can test with an independent party what is going on. I'd be curious what the output is.

stuartjohnson12on Show HN: Is AI hijacking your intent? A formal control algorithm to measure it

https://www.lesswrong.com/posts/rarcxjGp47dcHftCP/your-llm-a...

Hi author, this isn't personal, but I think your AI may be deceiving you into thinking you've made a breakthrough.

stuartjohnson12on 65% of Hacker News posts have negative sentiment, and they outperform

This is SUCH a good example of pedantry and will become my new primary example. All too often, people think of pedantry as being along the magnitude of scale. The "rational" pedant's response to this is to use quantitative jargon and bayes to scale up the size of the nitpick.

stuartjohnson12on RevisionDojo, a YC startup, is running astroturfing campaigns targeting kids?

I looked into founding a company in this space and steered straight back out of it because yes, by far and away the VAST majority of demand in the market of study tools for high/middle schoolers is cheating. Below that, parents are involved and there's a market there (but a bad one, because of double sales where you have to sell through the parent to the child even though those two actors have misaligned incentives).

https://www.gauthmath.com/

This AI cheating app is currently #8 for "education" in the iOS app store.

stuartjohnson12on Show HN: Dealta – A game-theoretic decentralized trading protocol

I'm sick of low quality ai generation in an awful lot of front page articles.

stuartjohnson12on Grifters, Gaslighters and Puff the Magic Dragon

comment redacted; decided i don't care about ai slop

stuartjohnson12on Hacker News front page now, but the titles are honest

> Marketing blog post explaining why you should buy our product (hatchet.run)

When you developer market hard enough that you make it into the LLM training data.

stuartjohnson12on Is Mozilla trying hard to kill itself?

I don't think HN comments have an irrational burning pit of hate for Mozilla. If Mozilla was shaped more like the Tor foundation in their words and actions I think a lot more people would be supportive.

stuartjohnson12on Show HN: Sub-microsecond (890 ns) trading execution research system

You do realise you didn't actually commit any code, right?

stuartjohnson12on Show HN: Sub-microsecond (890 ns) trading execution research system

If you're going to ask ChatGPT to write your response for you, I'll do the same.

---

Great question! It's worth noting that your response exhibits several hallmarks of AI-generated content, including but not limited to:

Bullet-point formatting where none was needed

Buzzword density that feels a bit elevated

Phrases like "fully reproducible, evidence-based" that have a certain... flavor to them

I hope this helps! Let me know if you have any other questions.

stuartjohnson12on Show HN: Sub-microsecond (890 ns) trading execution research system

https://github.com/krish567366/submicro-execution-engine/blo...

oh claude

stuartjohnson12on Show HN: KpopAPI – RESTful Kpop API:)

Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something.

stuartjohnson12on Programming peaked

Right, and there's the rub, and what's behind all these changes - coordination. We containerized to coordinate our ability to use lots of different runtime environments. We typescript'd and python'd the web stack so that our frontend and our data friends could feel at home in the backend. We created ever-more-elaborate package repositories to make layering increasingly complex software projects simpler.

I think what people who write essays like this wish for is the smaller scale of things. Industrial steel production is objectively better than old-timey blacksmithing, but hammering out sets of cutlery certainly feels a lot more personal. Steelmaking wasn't better before all these fancy mechanisms and danger and pollutants and logistics, but when it was craft, it was more satisfying to work on.

stuartjohnson12on Show HN: FirstDistro – I lost a $5k/mo customer overnight

Except you didn't, okay? You don't have to create fake stories to sell software.

stuartjohnson12on It’s time to free JavaScript (2024)

Apart from anything else, ECMAScript is a mouthful! Eeh-cee-emm-ay-script. Five syllables.

stuartjohnson12on 'Carspreading' is on the rise – and not everyone is happy about it

It seems like Waymo... just fundamentally does in fact work. So if we do things with cars intended for long term systematic reform, we should do things that are optimised for the world of increasing driverlessness.

I imagine a world without drivers to be a lot safer from a driving quality perspective.

I can imagine a world in cars where, like trains, it is no longer the norm to wear a seatbelt because crashes are so rare.

What implications does that have for the interior design of cars? Does that make the perfect car bigger or smaller?

stuartjohnson12on It's Always the Process, Stupid

That's why you need consultants to tell you that <buzzy-technique> has problems, but <rebadged-buzzy-technique> is really how you save money, and that's why working with a <rebadged-buzzy-technique> expert is how you can overhaul your business and manage operational costs.

stuartjohnson12on Mixpanel Security Breach

Based on what I know of OpenAI's culture, certainly permanent.

stuartjohnson12on We're losing our voice to LLMs

I think it's less about content topic and more meta content topic. EG I don't want to remove pictures of broccoli because I don't like broccoli, I'm trying to remove pictures of food because it makes me eat more. Similarly, I don't want to remove Political Takes I Disagree With, I want to remove Political Takes Designed To Make Me Angry. The latter has a destructive viral effect whose antidote is inattention.

Echo chamber is a loaded term. Nobody is upset about the Not Murdering People Randomly echo chamber we've created for ourselves in civilised society, and with good reason. Many ideologies are internally stable and don't virally cause the breakdown of society. The concerning echo chambers are the ones that intensify and self-reinforce when left alone.

stuartjohnson12on I don't care how well your "AI" works

Open source library development has to follow very tight sets of style adherence because of its extremely distributed nature, and the degree to which feature development is as much the design of new standards as it is writing working code. I would imagine that it is perhaps the kind of programming least well suited to AI assistance.

AI speeds me up a tremendous amount in my day job as a product engineer.

stuartjohnson12on OpenAI needs to raise at least $207B by 2030

If OpenAI fails, the assumed case is overcapitalisation relative to demand. In such a world, if the reason OpenAI has no liquidity is because there's insufficient demand, there's no customers to move across.

stuartjohnson12on I don't care how well your "AI" works

There's a lot of overlap between "AI is evil megacapitalism" and "AI is ineffective", and I never understood the latter, but I am increasingly arriving to the understanding that the latter claim isn't real, it's just a soldier in the war being fought over the former.

stuartjohnson12on We stopped roadmap work for a week and fixed bugs

Because marginal work is only marginally rewarded. Spending one week and coming back to whoever with a nice piece of paper saying we fixed 60 bugs will earn a lot more rope from non-technical folk than fixing 3 bugs per week - the latter just looks like cleaning up your incompetence.

stuartjohnson12on Show HN: Eyes-follow cursor widget for playful UI feedback

This is hilarious but making it a subscription service is rogue.

stuartjohnson12on A.I. and Social Media Contribute to 'Brain Rot'

Could you give me some searching clues to hunt down this or a similar profile?

stuartjohnson12on OpenAI probably can't make ends meet. That's where you come in

Depends what you mean by deep in the hole. Given how liberal Gemini is with free access it hardly seems like moneymaking is the end goal. Amongst AI labs, I've personally found Deepmind to be consistently the least sensationalist and least concerned with marketing. I think they're doing a great job of commoditising their complement.

https://gwern.net/complement