Hacker News Reader

>Model collapse has not happened yet because of luck.

Oh dear.

How many years will you say this before conceding it's no longer luck?

SamPatton The gap in shared understanding of LLM capability is widening

Well done.

It's incredible to see AI denialists still active in late 2026.

SamPatton The gap in shared understanding of LLM capability is widening

It's insane how confidently you can make a claim like this in a forum where thousands (tens of thousands?) of people are using these tools daily to build useful software for themselves and others.

How can you possibly believe it's all hype? Is everyone just lying? Are we all paying $200 per month because... why?

SamPatton The gap in shared understanding of LLM capability is widening

Being good (or superhuman) at software dev and maths shouldn't be treated as though it's equivalent to other domains.

Unlocking mastery of those domains would lead to incredible gains in all others. Whereas a fundamental advancement in many other domains doesn't spread as readily.

SamPatton The gap in shared understanding of LLM capability is widening

It's not that we're assuming the AI has made no mistakes. It's that we're confident in the general capability of the models such that we're willing to accept it'll make some mistakes, but working together with it, we'll still achieve our end goals.

I presume you also know your own code has mistakes, too? You wouldn't refuse to even look at the code of a colleague who has ever made a coding mistake, would you?

SamPatton The gap in shared understanding of LLM capability is widening

What time scales are you using to make this judgement?

Everything isn't getting worse constantly. The data is quite clearly the opposite. How old are you?

SamPatton Ask HN: What are you working on? (September 2026)

https://pelicans.art/

I was curious if AI agents could go beyond just making SVGs and create entire skits with them.

So I built a platform to do that. Results were meh until Astra, so I finally open sourced it last weekend.

SamPatton Research acceleration: The view inside OpenAI

I don't see how it's different from what we have now? Expensive cancer treatments already exist and are rationed by the companies who create them, then over time become more widely available.

If AI companies create better treatments they'll offer them to people who pay for them. So wealthier people will have better cancer treatments. If they can make them inexpensively, then a broader audience also gets treatment.

Eventually they lose IP rights and they become generics and the world gets better cancer treatments.

In all scenarios having them find better treatments benefits humans with cancer more than if they didn't.

I don't like trusting any organization, corporate or governmental, with control over people's lives, but it's obvious that we're not going to get cutting edge medical treatments without these organizations.

SamPatton DeepSeek v4.1 Flash

I agree that fears are overblown. But we have definitely seen some attacks, especially in the crypto space. Three major ones just in the past month: Coldcard wallet, Liquid, and Trezor email compromised.

They're almost certainly a result of more competent models finding exploits.

SamPatton Research acceleration: The view inside OpenAI

It is a good thing when it didn't exist before and it does exist now and wouldn't have existed without them.

If they profit immensely from curing cancer, good.

SamPatton Apple introduces M6 and M5 Ultra

Thanks for sharing this sentiment and including data. So few people seem aware of the wonderful progress humanity keeps making. Makes me worry that the progress will stall or even reverse because people don't even know it's happening.

SamPatton Why aren't smart people happier? (2022)

That's one of the only real downsides of apparent intelligence that I've seen: people in positions which exist to be supportive tend to assume you don't actually need their help.

Which, comparatively, might be true. But just because someone else needs more help doesn't mean it wouldn't be welcome.

Parents, teachers, advisors, supervisors, they just let you figure it out because they know you can, and they need to focus on those who can't. I had a career coach once tell me I didn't need to worry about finding a job, smart people always find work.

If you've never needed to ask for help before, this can really cause problems when you finally do.

SamPatton Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots

lLM

SamPatton Ask HN: What are you working on? (August 2026)

The Pebbble transcription is part of their SDK and very easy to use programmatically, also it's quite accurate (not perfect but good enough that I turned confirmation off).

The TTS is Piper installed on the same VPS as Hermes.

Happy to answer questions or share code. I might publish the device app eventually. I hope your Pebble arrives soon, it's a great device.

SamPatton Ask HN: What are you working on? (August 2026)

I have a Pebble Time 2, which is an open source watch. I built a voice operated AI assistant for it.

It's hosted on a VPS, connected over Tailscale, and has a realtime TTS server (Piper).

The Pebble has built in transcription, so my app takes the transcript, hits the server endpoint with three possible paths:

1. Deterministic response

The server looks for certain triggers from the transcript like "Add task," "Complete task," "Set a timer," etc that will run a simple function on server and write / retrieve to a database, no LLM call needed. It then responds with the TTS output and the watch speaks. This path is nearly instant.

2. Fast response

If no trigger is found, then it hits a gpt5.6 Luna endpoint, which decides if it can answer the query itself, or if it needs to defer to the big brother model. If it can handle a query itself, it usually responds to simple questions in less than 8 seconds. It's also good about properly deferring, very close to 100% after a few weeks.

3. Deferred response

The VPS hosts a Hermes installation, and for more complicated tasks or for ongoing projects the Luna router sends it to the Hermes agent. This is typically quite slow, as you'd expect, but remarkably capable. I built the voice app to cache a few lines like "this will take some time" which will automatically trigger if the router indicates it's deferring, and then will respond with the full response from Hermes once it's finished (sometimes minutes later).

Pebble has a quiet mode, and my app will only display the text response if it's on, otherwise all responses are run through the TTS. Piper is so fast that this adds almost no time to the response.

In only a few weeks this has become the primary way I set up tasks, review and close them, set timers and reminders, log my weight, check in on the progress of projects I'm running on Hermes, and ask simple day to day AI questions.

Since Pebble is open source, AI tools are very good at building it, and it's trivially easy to set it up so that you can just ask the model to deploy it live. It's an awesome experience to tell it to change something, and two minutes later, that change is on your wrist.

Very fun project!

SamPatton Inflect-Micro-v2: complete voice in 9.36M parameters

I just built my own voice assistant with my Pebble Time 2 watch and it uses a VPS hosted TTS (Piper) and Hermes agent.

I learned about all of these projects on HN at one point or another.

SamPatton Are we offloading too much of our thinking to AI?

In a sense I agree, since GDP is a bit of a silly metric.

But it's a decent enough proxy for the parts of human progress that I think are meaningful that it illustrates the point well.

That website is an excellent resource and you can find other more compelling charts there.

SamPatton Are we offloading too much of our thinking to AI?

>There was no default "state", at least by the time you're talking about humans, we probably already learned from each other by observation. Sure, politics and wars and religion and feudal dynasties and whatever make bumps in the line, but generally it goes up.

No, that's not the shape of the line. You can view it here:

https://ourworldindata.org/grapher/global-average-gdp-per-ca...

The hockey stick upturn in recent centuries needs explanation.

I'm pushing back so hard on this comparative definition of wealth (which I notice is the secondary definition you found behind large possessions) because I think it's silly to do that when we have an unbelievably dramatic increase in wealth in recent centuries which needs understanding first.

Hand-waving away human progress and forcing zero sum definitions of wealth feels unserious. You have terms for things like inequality, why not use those instead of distorting what wealth means?

SamPatton Are we offloading too much of our thinking to AI?

>A wealthy person is not someone with a lot of things.

No one made that claim.

By nearly any sane metric the world is wealthier. You can verify this on sites like Our World in Data.

We are healthier, have more leisure time, are less likely to die in natural disasters, etc.

It's willful ignorance to define wealth in a way that wouldn't show humanity as having gotten more wealthy. You're doing harm by overlooking massive gains we've made and understanding why we made them and instead redefining the common sense understanding of basic terms.

SamPatton Are we offloading too much of our thinking to AI?

>You are talking about access to resources and technologies, which is basically constantly rising, we couldn't stop that if we wanted to

We could very obviously stop our progress in resources and technology. It has happened throughout history many times and we see plenty of modern examples.

You seem to be treating the accumulated products of civilization as though they are geological facts rather than fragile outputs of maintained institutions.

Maybe you should ask yourself why you believe that 1) access to resources and technology are "basically constantly rising" and 2) why the relative difference in wealth today is more important than the (magical?) existence of a system were everyone has continually gotten more access to resources and technology.

SamPatton Are we offloading too much of our thinking to AI?

No. The world has absolutely gotten wealthier over the last few centuries.

Wealth is obviously not zero sum. Humanity is far wealthier today than before the industrial revolution, and the trend is still towards increasing wealth.

SamPatton Punch yourself in the face with reality

Yes but what would they learn from their mistake?

"I should run my purchase decisions past an AI in the future."

SamPatton Protect your right to run local AI

The entire point is to stop it from needing to be a cause in the first place.

Trying to establish the right to run models locally while we still can makes more sense than waiting to see if they do try, then organizing too late.

SamPatton 45°C cooling design cuts data center water use to near zero

As the earth has warmed over the last century, has food production increased or decreased?

If production increased (dramatically) while temperatures increased in the past, could this trend continue?

SamPatton 45°C cooling design cuts data center water use to near zero

Deaths from all natural disasters have dramatically decreased over the last century.

I won't bother sharing the source because you'll find some reason to dismiss it I'm sure.

It's interesting you claim my comment is propaganda when I cite a scientific source and yet have nothing to say about the parent comment which claims we no longer need to worry about heating due to climate change. Which of those comments seems more propagandistic?

SamPatton 45°C cooling design cuts data center water use to near zero

>the way we heat up the earth right now, i don't think you have to worry about heating

Nearly 10x more people die from the cold than from the heat.

"...9.43% of global deaths were attributable to non-optimal temperatures, with 8.52% from cold and 0.91% from heat."

https://www.thelancet.com/journals/lanplh/article/PIIS2542-5...

SamPatton Statement on US government directive to suspend access to Fable 5 and Mythos 5

LLMs don't think for you. Just like any other text you read, you can accept or reject it.

Discernment still exists.

SamPatton India's surprise baby bust

>I can't in good conscious bring a child into a world that so many are focused on absolutely destroying.

Who is focused on destroying the world?

I don't think hardly any super villains exist. People might have a different assessment of what destroying the world means than you do.

SamPatton It is time to give up the dualism introduced by the debate on consciousness

Emotions do seem to act as signaling, but is that the same as an attempt at manipulation for the benefit of the individual?

It seems conceivable in social groups that having an honest accounting of how people are feeling (via emotions) available to the group might benefit the group in achieving their goals while not always benefiting the individual.

SamPatton Today I've made the difficult decision to reduce the size of Coinbase by ~14%

Many comments are mocking the "Non-technical teams are now shipping production code" line as an obvious disaster waiting to happen.

I think this will be commonplace in the not too distant future.

Some disasters will happen, just like they did before AI. Skeptics will gleefully point out these failures while more and more non-technical teams ship code.