HNHacker News
TopNewBestAskShowJobs

lend000

4,111 karma · joined December 17, 2016

Technology, science, and macroeconomics.
submissionscomments
lend000··on Qwen3-Next
Very interesting time to be alive.
lend000··on Qwen3-Next
At the speeds AI is moving, we've effectively used it all; the high quality data you need to make smarter models is coming in at a trickle. We're not getting 10^5 Principia Mathematicas published every day. Maybe I just don't have the vision to understand it, but it seems like AI-generated synthetic data for training shouldn't be able to make a smarter model than whatever produced that data. I can imagine synthetic data would be useful for making models more efficient (that's what quantized models are, after all), but not pushing the frontier.
lend000··on Qwen3-Next
Prediction: AI will become commoditized ~15 IQ points higher than the state of the art models today, and with larger context, within 4 years as the incremental improvements in training from synthetic data plateaus (we've already used all the "real" data out there) and open source models are cheaply trained on the outputs of the big money models. Then AI development stagnates until someone invents an effective way to use competitive reinforcement learning to train generalized intelligence (similar to how AlphaGo was trained), removing the need for vast quantities of training data. Then, we get real AGI.
lend000··on Toothpaste made with keratin may protect and repair damaged teeth: study
I've used these: https://www.amazon.com/dp/B08F2MW3G2 https://www.amazon.com/dp/B078YFQZK7
lend000··on Toothpaste made with keratin may protect and repair damaged teeth: study
It seems to me the two are effectively the same unless you have significantly misshaped teeth (remineralizing vs regenerating). I also use hydroxyapatite, just to reduce my fluoride exposure, although I believe fluoride is supposed to be a more potent remineralizer (and fluorapatite is allegedly stronger than natural hydroxyapatite). But the upside is that I don't mind swishing hydroxyapatite around in my mouth for 10 minutes, twice a day, so whenever I go to the dentist, I'm the healthiest mouth of the day (not the case pre-hydroxyapatite tooth paste/powder).
lend000··on Open models by OpenAI
For me the game changer here is the speed. On my local Mac I'm finally getting token counts that are faster than I can process the output (~96 tok/s), and the quality has been solid. I had previously tried some of the distilled qwen and deepseek models and they were just way too slow for me to seriously use them.
lend000··on Telo MT1
Interesting, thanks for clarifying.
lend000··on Telo MT1
As much as I like the novelty of the design, there isn't much of a crumple zone for a head on collision. I could see the wheel placement making this a fun off-road vehicle, though.
lend000··on The North Korean fake IT worker problem is ubiquitous
Interesting idea! This seems like a natural extension of the coworking space business concept.
lend000··on Context Engineering for Agents
I'm consistently amazed by how great the first response from o3-pro deep research is, and then consistently disappointed by response number 5 or so if I continue the conversation. Better context management is the most important bottleneck in LLMs, and it seems like a robust solution would involve modifying the transformer architecture itself instead of using context limited LLMs to manage the context for other LLMs.
lend000··on What I learned gathering nootropic ratings
I had long covid which manifested as post-exertional malaise and general brain fog, and which was not improving for 2 years, and I essentially solved it with a combination of a low sugar paleo diet (starting as an autoimmune protocol diet for a few weeks to determine some other things that caused flare ups for me like nutritional yeast and capsaicin), lots of walking, light and increasing exercise (especially outdoor exercise), and a dedication to sleep hygiene. Many of which were referenced in the post.

It's a ton of work at first but it's completely worth it -- post-exertional malaise sucks.

Other things that helped me personally: daily cold plunges up to the chin (so you feel it in your vagus nerve), HRV reset breathing exercises, and daily meditation. Wish you the best of luck.

lend000··on Firefox 120 to Firefox 141 Web Browser Benchmarks
This finally convinced me to press the update and restart button. Nice to see that not all commercial software projects tend towards bloat.
lend000··on Starship: The minimal, fast, and customizable prompt for any shell
I'm trying to understand what they are demonstrating with the "false; true" commands in the little video loop. Can someone chime in?
lend000··on Nvidia CEO criticizes Anthropic boss over his statements on AI
Friends and others who have described the details of their non-technical white collar work to me over the last 15 or so years have typically evoked the unspoken response... "Hmm, I could probably automate about 50-80% of your job in a couple weeks." That's pre-AI. And yet years later, they would still have similar jobs with repetitive computer work.

So I'm quite confident the future will be similar with AI. Yes, in theory, it could already replace perhaps 90% of the white collar work in the economy. But in practice? It will be a slow, decades-long transition as old-school / less tech savvy employers adopt the new processes and technologies.

Junior software engineers trying to break into high paying tech jobs will be hit the hardest hit IMO, since employers are tech savvy, the supply of junior developers is as high as ever, and they simply will take too long to add more value than using Claude unless you have a lot of money to burn on training them.

lend000··on Gemma 3n preview: Mobile-first AI
Not a bad idea for next generation models, especially since the state of the art already uses Mixture of Experts.
lend000··on OpenAI o3 and o4-mini
I imagine after GPT-4 / o1, improvements on benchmarks have been increasingly a result of overfitting, because those breakthrough models already used most of the high quality training data that is available on the internet, there haven't been any dramatic architectural changes, we are already melting the world's GPUs, and there simply isn't enough new, high quality data being generated (orders of magnitudes more than what they already used on older models) to enable breakthrough improvements.

What I'd really like to see is the model development companies improving their guardrails so that they are less concerned about doing something offensive or controversial and more concerned about conveying their level of confidence in an answer, i.e. saying I don't know every once in a while. Once we get a couple years of relative stagnation in AI models, I suspect this will become a huge selling point and you will start getting "defense grade", B2B type models where accuracy is king.

lend000··on Lotka–Volterra Equations
You still can! But expect a lot of pushback from midwits who appeal to authority.
lend000··on Why I don't discuss politics with friends
Is it really likely that an intelligent person like yourself could hold 95% intellectual alignment with one of the two lowest common denominators (largest pluralities) in a country on complex political topics? Consider how much each party's platform has changed in the last 20 years, and how much more they will change in the next 20. I would say it's more likely that someone like yourself is quite intelligent and creative and is instead unaware of those deeply ingrained tribal instincts.

Media in the US, especially now via incessant social media feeds, fuels this. It showers us with information showing how the "other side" is bad. So you can have a correct opinion that the other tribe is bad without any quantitative metrics to compare how bad it is compared to your tribe, which is also very bad.

Btw, regarding the basic personality traits thing. I found this paper very interesting [0]. Sort of refutes the "conservatives lack empathy and fear change" angle. On average, I suspect most liberals and conservatives have very similar averages across most personality traits and are mostly just a product of their environment.

[0] https://pubmed.ncbi.nlm.nih.gov/34429211/#:~:text=Our%20meta...

lend000··on A 6-Hour Time-Stretched Version of Brian Eno's Music for Airports
If you like Brian Eno's music, you might enjoy Hiroshi Yoshimura. Wet Land is one of my favorite albums of all time.
lend000··on Finding Signal in the Noise: Machine Learning and the Markets (Jane Street)
From a technical challenge standpoint, trading is extremely difficult. But it provides high upside which only starting a business can exceed. However, starting a business has all the other uncertainty (regulatory, accounting, insurance, licensing, taxes, customers, branding, sales, product-market-fit), making it far more psychologically challenging overall. Startup founders need to learn and juggle a lot of things instead of hyper-fixating on markets. Basically, if you know you are sufficiently intelligent and talented as an engineer, your expected return is likely much higher in trading (especially via a big quant firm), with much lower risk.

If we want more talented traders to become founders building economic value (which we should all agree on), we need to make it less onerous to start a business.

lend000··on GPT-4.5
My understanding is that o1 is a system built on GPT-4o, so this pricing might explain why o3 (the alleged full version) cost so much money to run in the published benchmark tests [0]. It must be using GPT 4.5 or something similar as the underlying model.

[0] https://arcprize.org/blog/oai-o3-pub-breakthrough

lend000··on Warren Buffett amasses more cash and sells more stock
Interesting point -- I wonder if there are variations on this calculation available with global GDP, or ideally, company market cap adjusted by percentage of US revenue/profit to US GDP.
lend000··on The marketing genius of Bryan Johnson
Whoop is definitely not accurate for sleep data, but it's great for exercise and there's certainly value in seeing trends.
lend000··on Musk-led group makes $97B bid for control of OpenAI
Millions of dollars worth of subscriptions for reporting on government press releases with added spin seems like a great example of waste to me... The government shouldn't be paying any media outlets purely due to the conflict of interest, regardless.

> Or are we to believe that competence is just arbitrarily cross-applicable?

I would wager that even you are more competent than the average auditor, assuming you know how to program. I'm seeing a lot of appeals to authority and credentialism, and assumptions that people who are hired by bureaucrats and politicians to audit themselves have better aligned incentives than a wildcard entrepreneur. If we can't agree on the fundamental axioms, I don't see this conversation going anywhere.

lend000··on Musk-led group makes $97B bid for control of OpenAI
> If DOGE was serious about finding cost savings, they would start with the biggest budget hog, the one that has never once passed an audit in its entire history – the Pentagon.

I imagine that's the goal, but we're literally just a few weeks in and a large amount of the DoD is extremely secretive and difficult to access by the elected government (which is concerning enough as it is). From what I understand they are already moving to downsize the FBI, and the more secretive three letter agencies will hopefully follow.

> Preventing the release of existing supplies of medicine and food to literal starving children in third world countries.

This is really your strongest point, but if you look at the US track record abroad, most of that aid is damage control for wars/coups/etc. started directly or indirectly by the US. Most of those people, and certainly US taxpayers, would be better off with no US influence in the long run. Whether that will actually happen under this administration remains to be seen.

lend000··on Musk-led group makes $97B bid for control of OpenAI
This is what I saw most recently: [0]. Whether or not it could be vaguely considered to be congressionally authorized (the executive branch has leeway in the execution of said laws) is more or less irrelevant to whether it's wasteful and/or fraudulent, but it looks like those examples were not documented in law anyway.

I imagine it will be thousands and thousands of examples of "small time corruption" with a few million here and a few tens of millions here that in no sane world should be spent. And for some reason, people prefer to hate billionaires that create wealth when instructed by multimillionaires who siphon wealth from taxpayers but happen to own newspapers.

> You have no reason to believe Elon is either competent, nor independent, nor well-intentioned on any of this.

You could say the last two about anyone, but to deny that Elon is competent, especially coming from an anonymous internet commenter... okay.

[0] https://x.com/ricwe123/status/1887562751364481290

lend000··on Musk-led group makes $97B bid for control of OpenAI
Genuine question: are people genuinely against the work being done by DOGE? I am frequenting both X and HN + other liberal leaning news sources, and it's like the people in the comments are living in different solar systems.

I see a post on X with receipts showing millions of dollars of pretty obvious waste and/or corruption, many involving news outlet grants and subscriptions, including names, amounts, etc. Then I see some tearjerker story about families of USAID workers on one of the beneficiary news outlets. And in the comments, people lamenting that Elon Musk's success is only due to luck and his estranged father's defunct gem mine. Or worrying Elon will use their SSN (which is already available on the dark web) to open a credit card.

You're pretty smart people. Do we really think auditing government spending by a super-competent outsider is a bad thing? The last time it was done was by Clinton, btw, who managed to create a surplus (granted, the debt situation was nothing compared to today). And before that, Truman, back before hyper-partisan politics. I can't imagine the level of grift back then was anywhere close to today in our bloated budget. I'm optimistic, and I think y'all need to relax.

lend000··on 20k federal workers take "buyout" so far, official says
The quality of the comments is really disappointing. I'll add an optimistic one -- this is the first time in my lifetime that I've felt like the federal government was actually making serious progress on an important problem without being forced to by some emergency. I'm very curious to see what things look like a year from now.
lend000··on Don't make fun of renowned author Dan Brown (2013)
Agree. For example, Dragon's Egg by Forward is not a well written book by most of those metrics, and yet I very much enjoyed the novel due to the creative story.
lend000··on LA wildfires force thousands to evacuate, NASA JPL closed
CO2 is already used in many fire extinguishers, especially in industrial settings for class B and C fires. And we trust the average person with huge containers of explosive propane.

More is possible than you think, but obviously it comes with a lot of caveats because of the danger potential. But as a professional firefighting tool, "no potential" is a pessimistic take IMO.

Fire roads are often used as firebreaks, and in SoCal many follow the ridges of mountains. This is really the ideal case for CO2 that I envision, as the cold CO2 released from a compressed container will flow downhill into an evacuated area and smother the fire attempting to cross the ridge. Curious if you believe the current state of the art cannot be improved upon.

The tarps would have to be huge and ideally quite thin, hence why I wondered if modern materials science was up to the task. An advantage is that they could preserve foliage while still creating a firebreak, but again, mostly a materials science question. Probably infeasible, but just spitballing.

← PreviousPage 2 of 34Next →